In the second part of the Braiins series on mining and the electric grid, this article explores transmission, curtailment, and behind the meter with applications to bitcoin mining.
This article is the second in a series about bitcoin mining and energy infrastructure. Each article offers an introductory level explanation of electric grids and their relationship with mining to better educate miners and other bitcoin investors.
Read Part 1: Generators here.
Part 1 of this series covered a general understanding of power generation, how grid operators balance generation with load, frequency regulation, and how bitcoin mining can fit into the puzzle.
As mining continues to integrate into the energy industry, it is also important for miners and bitcoin investors to learn the basics of power transmission, what terms like curtailment and behind-the-meter mean, and how bitcoin mining can fit in.
Feedback from Part 1 of this series was exceptional, and it’s this author’s hope that the information shared in this Part 2 will also be helpful.
This article is long. Even longer than Part 1. And the goal for this article is to strike a tone somewhere between light reading that’s not fully reflective of the system's complexities, and dense, truly-reflective-of-the-system reading that would be hard to digest for someone who doesn't work in power systems on a daily basis.
This author hopes to provide enough detail for those looking for it, while also not adding so much detail that readers lose interest. In striving for this goal, some representations made in this article will be missing precision or nuance. But the general concepts should be accurate.
Like the previous article, this one also uses the Electric Reliability Council of Texas (ERCOT) electric grid as example material. Some of the concepts explained throughout this article would map onto other grids, while others would not.
Transmission lines are a combination of wood or steel transmission towers and the conductors they carry. These lines act as power pathways that connect generator substations to transmission level switching stations and load serving substations.
Substations are structures where power is transformed from high levels to lower levels of voltage (or vice versa), and/or where transmission lines terminate and others start.
Switching stations are a type of substation used to sectionalize transmission lines, or allow for certain lines to be switched out while allowing power to flow through other paths. Switching stations don't normally have different voltage levels, and therefore don't have transformers. At a high level, substations have different configurations of breakers that make connecting to them easier or more difficult. Substations can also be built to have extra room for future expansions.
That’s enough of a substation vocabulary lesson for now.
Because of certain physical qualities of electricity, moving power at higher voltage levels minimizes losses. For this purpose, when moving power over long distances power systems engineers try to use the highest voltage level practicable. Generator substations step voltage up to transmission levels, and load serving substations step the voltage down to a level useful for loads.
Transmission lines, substations, and generators in their aggregate form a high voltage transmission network, which is monitored and managed for reliability by a central grid operator (like ERCOT).
The decision of what type of transmission line to build is normally dependent on the purpose of the line, the local geography, vertical clearance (meaning, the amount of space available below the conductors), right-of-way, local climate, and cost.
A fun rule of thumb: Voltage levels are normalized to a set. Some common levels in the 60 HZ world are 13.8 kV, 34.5 kV, 69 kV, 115 kV, 138 kV, 220 kV, and 345 kV. Count the insulators (small ceramic saucers between the conductors and the towers) to estimate the voltage level of a line (roughly 10 or 11 kV per insulator). See below for some images of different tower types. This author's wife loves when he calls out tower type and voltage levels as they drive past lines.
The series of images below show H-Frame (often wood or steel), Monopole (often steel), Lattice (often steel), and Monopole (with underslung distribution).
Transmission lines are different from the wooden pole lines seen around residential neighborhoods. Those are distribution lines, and they are an integral part of a power system’s distribution infrastructure. Distribution lines are used to distribute power from transmission level substations to distribution level loads. These lines are normally monitored and managed by a local utility, not the main “headline” grid operator. But sometimes it can be the same entity. It just depends on where you are.
In ERCOT, transmission level equipment is normally considered 69 kV and above. This level of equipment has a different level of scrutiny and operation than distribution equipment, since impacts on this higher voltage level would cause subsequent problems for downstream systems. 345 kV lines and higher have their own designation as “Extra High Voltage” and receive even more scrutiny.
In the image below, transmission lines are shown in the middle, gray section between power generators and distribution lines.
Before continuing even deeper in discussing power transmission, it will be helpful to briefly revisit some of the power generation information from the previous article of this series on mining and the grid and add some graphics.
Readers will recall from the previous article that generators on a power grid with a nodal market such as ERCOT are dispatched in ascending order from least marginal cost of production (cheapest fuel cost) upward towards the most expensive marginal cost of production (most expensive fuel). The most expensive megawatt (MW) that matches the total system generation with load then sets the price for all generators that were dispatched in that interval.
This means that generators in a nodal market want to:
Solar and wind generators, since they have no fuel cost to account for, are $0 marginal cost generators, which means that they bid $0 into the market and are subsequently by default online. The grid operator accounts for the forecasted megawatt contribution from these zero-cost generators first before dispatching more expensive generation.
What does this look like?
The diagrams below illustrate a few scenarios with different fuel types competing to supply power to a city. Each example diagram is followed by a scenario explanation. These figures are helpful to demonstrate a couple concepts explained in this article.
A caveat for other power engineers: These diagrams and scenarios ignore constraints for now in order to absolutely hammer down the concept of marginal pricing.
Example 1: Wind Or Solar Is Marginal
Explanation for Example 1
Solar and Wind Node Price: $0
City Node Price: $0
The grid operator is informed that wind and solar forecasts call for the facilities to produce 60 MW & 40 MW respectively, totaling 100 MW, which perfectly matches the 100 MW forecasted demand of the city for the next five minute interval (unrealistic... but theoretically possible).
Since the wind and solar generators both bid the market at $0 and they provide the last MW needed to match the city demand, they set the marginal price for the system to $0, and they subsequently get paid $0 for each MW they produced in that interval. The grid operator does not need to move upward from $0 to dispatch more expensive generation, and so the peaker and combined cycle plants remain offline and don't get paid at all (for energy).
ERCOT uses ancillary services that it procured in the day ahead market to manage fluctuations in frequency between this interval and the next. An example of this might be that the peaker plant sold non-spinning reserve to ERCOT yesterday, such that if ERCOT needed them to start up really fast, they could.
This is also an illustration of how wind and solar don’t ever plan on actually being the marginal unit, because if they did, bidding $0 wouldn't make any sense. They would never make any money!
Example 2: Combined Cycle (Natural Gas) Is Marginal
Explanation for Example 2
Solar and Wind Node Price: $20
City Node Price: $20
Combined Cycle Node Price: $20
The grid operator is informed that wind and solar forecasts call for the facilities to produce 50 MW and 40 MW respectively, totaling 90 MW, and the city load for the next five minutes is still 100 MW, so the grid operator looks to the available, least expensive generation.
The peaker plant says it can provide 10 MW for $100/MWh, and the combined cycle plant says it can provide 10 MW for only $20/MWh. Since $20 is cheaper than $100, the grid operator dispatches the combined cycle plant instead of the more expensive peaker.
The combined cycle plant ramps to 10 MW at its marginal rate of $20/MWh, and it therefore sets the price at $20/MWh for everyone.
Example 3: Peaker Plant (Natural Gas) Is Marginal
Explanation for Example 3
Solar and Wind Node Price: $100
City Node Price: $100
Peaker Price: $100
The grid operator is informed that wind and solar forecasts call for the facilities to produce 50 MW and 40 MW respectively, totaling 90 MW, and the city load for the next five minutes is still 100 MW, so the grid operator looks to the available, least expensive generation.
The peaker plant says it can provide 10 MW for $100/MWh. The combined cycle is actually offline for maintenance as part of a planned outage, so the grid operator has no choice but to dispatch the more expensive peaker plant.
The peaker turns on for 10 MW at its marginal rate of $100/MWh, and it sets the price at $100/MWh for everyone.
After considering the above examples, hopefully readers can see just how important marginal pricing is for power generators. Generators plan their multi-decade lifespan around variables like how often their marginal price will enable them to be online, how expensive their fuel may become, and how often they will face outages for repairs or maintenance.
On a related note, it’s interesting to observe that natural gas facilities like the Combined Cycle Plants and Peaker Plants in the above examples are the primary price-setters in today’s markets. And what do they use to set their prices? Their fuel costs. Because of this, the wholesale electricity market is really a pseudo-index of natural gas prices.
Congratulations for progressing through the basics of transmission lines and a refresher (with examples) on power generation. Let’s keep going!
The previous section showed diagrams of how much generators are paid as their $/MWh nodal price, which is set by the marginal unit. But what rate does the 100 MW city load in the diagram pay?
There are various strategies for these situations, especially if the load is a municipality or utility that needs to physically take ownership of the power. But for loads like bitcoin miners, ERCOT has a Load Zone structure where all the nodes in a certain geography (North, West, East, Austin, LCRA) are load-weight-averaged together with delivery and other charges added in some fashion.
Bitcoin miners or other load types normally hedge their variable Load Zone rate by entering into fancy power purchase agreements (PPAs) with nearby generators. This is usually a financial “swap” whereby when the generator’s nodal or average geographic area (called a hub) price goes up, the bitcoin miner is exposed to that increased price, which then offsets their commensurate increased Load Zone rate (and vice versa for the downside, when the generator would suffer from decreased pricing).
Provided that the generator's pricing correlates with the bitcoin miners’ load zone pricing, the bitcoin miner should be safe. This type of agreement can set both the load and generator to an effectively fixed price that is dependent upon the modeling assumptions that went into the PPA structure, plus whatever risks still exist (basis, volume, etc.)
Even more sophisticated loads and generators employ power traders to execute optimal strategies here on top of their PPAs. This gets complicated rather quickly.. but the important part to understand is that there are many creative financial instruments used in these energy markets to hedge your floating wholesale price.
Now, back to transmission.
The transmission lines in our diagrams above were a bit unrealistic, because in the real world, transmission lines can only carry so much power. If a grid operator sent more power down a line than it could handle, they would damage the equipment. To stop this from happening, an entire field of electrical engineering called protection engineering uses coordination and combinations of circuit breakers, fuses, relays, and other equipment to isolate equipment if they sense too much current.
Also to prevent too much power being sent on these lines, grid operators like ERCOT spend a majority of their computing power solving the math problem of how to match generation to load not only at the least cost, but at the least cost without overloading any equipment.
This means that ERCOT’s dispatch for generation should never result in overloads, so the aforementioned protection engineering is primarily concerned with isolating faults or trips, what are called contingencies. But that is a complex topic for another article. Back to the fact that transmission lines have ratings…
Revisiting the first example diagram shared in this article, notice the inclusion of a 50 MW transmission line rating for the West to East line.
Warning: these next few diagrams are going to be doozies, but the reader should know they won’t be too bad. And any reader that fully understands them will be ready to apply for a Market Engineer position at ERCOT!)
Example 4: Wind and Solar Curtailment
Explanation for Example 4
Solar and Wind Node Price: $0
City Node Price: $36
Combined Cycle Node Price: $36
This example is a modification of the first scenario used in this article, but with a 50 MW limit on the transmission line. In the original example, the solar and wind produced 100 MW, all of which flowed to the city and set the clearing price for the system & their nodes at $0. In this modified example, some key things are different:
The solar irradiance and wind speed of that day could have resulted in the plants producing a combined 100 MW, which would have caused an overload on our 50 MW transmission line. To prevent this from happening, the grid operator had to partially curtail the wind and solar facilities. Simply put, there wasn’t enough room on the transmission lines between the generation and load to handle that amount of power. In other markets, this is solved by forcing generators to pay for upgrades before they come online (PJM). In ERCOT, it’s solved by just curtailing plants in real-time.
The grid operator performs this curtailment by ordering the generators to produce less power by sending them instructions (base points), which the wind and solar have to follow or else they face fines. Along with ordering the facilities down, the grid operator also lowers their nodal price using some fancy math described below.
Readers will notice in the modified diagram that the wind and solar farm are paid $0/MWh instead of the clearing price of $36/MWh set by the Combined Cycle plant. In this new scenario, the grid operator takes the clearing price set by the combined cycle plant ($36) and subtracts something called a “Shadow Price” ($36) multiplied by a “Shift Factor” (1.0).
Yes, even more vocabulary to learn, so let’s pause briefly and define these terms:
Notice also that since the transmission line could not carry all 100 MW of that sweet $0/MWh solar and wind to the load, more expensive generation needed to be turned on to make sure the city load was met.
As a result, the combined cycle plant was dispatched up to 50 MW and the clearing price was set by its marginal cost to produce that amount - $36/MWh. This marginal price then feeds into the price for all generators on the grid. More room on transmission would’ve kept the grid operator from having to go that far up the bid stack, and thus would have resulted in a lower clearing price, and an overall lower cost to run the system. In other words, curtailment results in a less efficient, more costly system.
In power system terms what’s happening here is called a Binding Constraint. The constraint was the limit on the line (50 MW). It’s binding because there’s no more room on the line (it’s full, and was at risk of overloading), and so the grid operator had to act on it in order to prevent an overload. Each Binding Constraint on a system will have its own marginal price-setting generator. Yes, the author knows they previously said there was only one price setter in a system.. But the author lied!
In some five minute intervals through the day there are no binding constraints, and thus there is in fact only one price setter. Otherwise, each binding constraint will have its own price setter, in addition to the clearing price.
A small caveat: the flows that determine if a constraint is binding or not are “post-contingency” flows. Which means the grid is actually operated such that if it loses any single line or element due to a fault or trip, the flows that were going along that element could redirect elsewhere and not overload anything else after being redirected. This method of dispatching is done to ensure operators can isolate single contingencies, such that one fault does not cause a domino effect tripping other elements and causing a blackout. The example used above is a bit too simple to include this nuance, but it's still good to know.
Simple enough, right?
Even though the big headline name for what was described in the last example is curtailment– where solar and wind are ordered to produce less power than they otherwise could–the same scenario also describes congestion. And congestion isn’t unique to renewables. It happens any time that cheaper generation can’t get to load due to transmission bottlenecks.
And just to complete the full definition, curtailment could also happen if there was more solar and wind generation available than load. For instance, in our first ever example with the infinite transmission line, if there was actually 110 MW of solar and wind but still only 100 MW of city load, the operator would still have to partially curtail those facilities. Back to our congested line:
In power systems jargon, the congestion scenario is described like this:
“There was congestion along the line going into the city, which resulted in 50 megawatts of curtailment in the west & higher pricing for city residents.”
In the real world, congestion is a huge component of wholesale electricity pricing, and it’s the reason why pricing differs between nodes in a network. If congestion never happened, all nodes would be the same price.
The chart below shows real-time market pricing in ERCOT at the time of writing this article. Readers will notice how some price pockets have significantly higher or lower prices than other pockets. These price differences are due to congestion that prevents generation within those low priced areas from getting to loads in the higher priced areas. Thus, more expensive generation is turned on in the higher priced areas.
At the risk of being annoyingly repetitive, it’s important to emphasize that this article’s example power network is just a four bus system with one transmission line and four generators. The simplicity of the diagram is also why it worked out perfectly that the nodes ended up clearing at the price they bid.
In the real world, ERCOT, for example, has 700+ generators, thousands of buses, thousands of lines, and the algorithm that determines pricing is re-run every five minutes, ordering all these elements in a thousand by thousand matrix to perform the correct optimizations of flows. In simplifying all this complexity, this article’s example is really only useful for the topics discussed here: describing congestion, curtailment, and marginal pricing generally.
What is negative pricing, and how does it work?
Wind Production Tax Credits (PTCs) have been a part of the power generation industry for many decades. The original PTC began in 1992, it has lapsed numerous times since, but has always been reinstated.
The PTC is a $/MWh credit for production for a wind facility that meets the eligibility requirements for the tax credit program. Depending on what year a wind generator comes online, they either receive the full credit amount or a phased proportion of the full amount. This means that some wind facilities may get the full amount of $26/MWh, and some may get only 60% (~$16/MWh), etc. After a project is online for ten years, the PTC runs out, after which the generator only gets grid pricing.
So, for every megawatt hour a wind facility produces while eligible for PTC, it will receive tax credit payments. Readers will remember previous discussions about the “bid stack” where grid operators move from cheapest generation upwards. Since some wind generators receive this credit, they can bid into the market at a negative price since they know they will receive the $/MWh credit for their production anyways. Bidding into the negative ensures that these generators are curtailed after the competing $0/MWh solar or newer wind that receives less PTC..
Let’s revisit the power system diagram again, this time with the wind bid changed from $0 down to $-25. Continue reading for an explanation of this diagram.
Negative Pricing Example
Explanation for Negative Pricing Example
Solar and Wind Node Price: $-25
City Node Price: $36
Combined Cycle Node Price: $36
In the example, because wind bids $-25, significantly lower than solar ($0), the grid operator fully curtails the solar facility before curtailing the wind facility at all.
Since the 50 megawatts of wind power is still unable to meet the city load, the combined cycle plant turns on to meet the demand, and sets the clearing price at $36/MWh.
The grid operator then takes the clearing price set by the combined cycle plant ($36) and again subtracts the “Shadow Price” ($61) multiplied by the “Shift Factor” (1.0). To revisit shadow prices quickly:
What all of this means is that the price at the wind farm is now $-25, and the price at the combined cycle facility is $36. Note that the wind farm bidding negative doesn't actually have to pay to produce, because the PTC payment brings them up to positive.
Here are a few high-level observations about curtailment due to transmission and congestion:
So what’s the solution?
A quick way to eliminate curtailment and congestion would be to increase the transmission line rating of every line on the grid by a factor of 10 and add synchronous condensers everywhere that there are stability concerns.
But that wouldn't make sense.. How much would that cost!? These are the questions grid operators are tasked with balancing, and the influx of intermittent generation and bitcoin mining loads don't make the job any easier.
This is what normally happens.
If congestion in a specific area is identified as a costly problem, transmission planning engineers like this author will analyze models of the system to understand the expected cost to upgrade a congested transmission line and compare that cost to the total savings to the consumer.
Normally there is a rule that sets expectations for a number of years wherein the cost of the upgrade is paid back. For example, if a transmission line upgrade costs $200 million but the consumer (system cost) saves $50 million per year, the consumer breaks even and begins saving money after four years.
Transmission planning engineers then say, “This makes economic sense. Let’s propose it!”
Curtailment introduces risks and disadvantages to generators.
For generation developers, assumptions about a prospective project location’s future grid pricing and curtailment are identified early on using power system models. Most of the time, generators that actually get built have modeled future revenue that looks something like a low risk, low return with contracted payback after a few decades.
Generator developers secure this contracted payback by selling their future projects power as a PPA. In a PPA, a generator contracts some percentage of the power they will produce to either a utility that needs the capacity, an industrial customer looking to hedge their exposure to real-time rates, or an industrial customer looking for renewable energy credits (RECs) to offset their use and claim carbon neutrality. Normally these PPA counterparties, called “offtakers” put out request for proposals (RFPs), which generation developers competitively bid to win. The generator will retain some smaller uncontracted percentage of their future production to sell at the market price (merchant strip).
A signed PPA secures contracted revenue for the generator, which is then used to market the project to a buyer. The buyers of these projects are usually investment funds looking for long-term, low-risk returns which the PPA provides.
Since generators don’t get paid for their curtailed energy, generation developers will incorporate their expected curtailment into their PPA price that they submit to RFPs. For example, if a generator expects to be curtailed 50% of the time, they would need to double the PPA price to meet the same returns as if they had no curtailment. But if a generator only expects to be curtailed 1% of the time, the PPA only needs a strike price increase of 1% to achieve those same economics. This is why selecting a good spot on the grid with low curtailment is important, as it will allow a generator to price PPAs low and win RFPs.
This is where a lot of prospective generation projects die. No project is going to win an RFP and secure financing if their PPA strike price is 20% higher than their competitors. Curtailment risk of any significant size is a project killer. Tangentially, this author often wonders how many projects in the ERCOT generation queue are just a few curtailment percentage points away from becoming real projects, but are just sitting on the shelf.
And with increased renewable buildout, wind and solar facilities are increasingly cannibalizing their own pricing. Since these generators are all online during the same hours of the day, every project pushes their own grid pricing downward and increases their own curtailment (see: Duck Curve in California, Armadillo Curve in ERCOT).
While it may seem obvious, curtailment is a big waste of energy. If you have a wind or solar farm that could produce 200 MW, but is curtailed to 0 MW, there is a huge opportunity cost for you. For generators, being curtailed is akin to throwing money down the drain, even if they already included curtailment assumptions into their PPA and won an RFP. The generator would still prefer to monetize that curtailed energy.
Even if a wind or solar farm looks great on paper, stability concerns could force the grid operator to limit exports from their area. Now, all of a sudden, the new generator is being curtailed a lot more than was expected, and their revenue is terrible. To be specific:
In the case of a fixed volume PPA where the facility owes an offtaker a certain volume of megawatt hours per month, the generator isn’t able to deliver the contracted amount of power because of curtailment. Even if the fixed volume amount is perfectly aligned with what the generator could be producing, since they can’t deliver those MW’s to the grid, the generator still has to buy the balance of the owed energy on the market and make the offtaker whole.
In a different scenario with an as-generated PPA, the generator isn’t able to sell the amount of power they previously planned to sell, so they end up simply missing revenue targets.
And if a generator does not have a PPA and is just selling energy to the market at the wholesale rate, being curtailed is still just wasting energy they could be generating and doing something with.
Transmission upgrades take years, and in a 30 year project those years really matter. For a wind project, those are also years that the facility could be earning PTCs for the megawatt hours that are curtailed. Remember, a qualifying project only has 10 years to collect PTCs before they run out, tick tock!
It is no small feat reading through all of the diagrams, jargon, scenarios, and explanations from the past 15 pages or so. This author promises this article is getting close to applying these concepts to bitcoin mining.
Mining advocates often tout their industry as fitting into this picture of curtailment and generation growth somehow. The concept of metering will help to clarify the potential alignment.
Metering Explained
A generation or load project that connects to the transmission level network must connect through something called a meter, which are very similar to the residential meters most readers are probably familiar with. These meters have different names in different systems, but in ERCOT they are called EPS (ERCOT Polled Settlement) meters.
EPS meters are normally connected at the Point Of Interconnect (POI), where the generator or load owned equipment ends and the transmission operator equipment begins. Wherever they are connected, the meters must be able to read the total megawatts being injected to or withdrawn from the transmission network. In other words, the meter is what actually matters when it comes to paying or being paid for electricity.
Aside from settlement (getting paid or paying), the meter is also how ERCOT monitors and dispatches generators or loads as resources.
Hopefully that quick explanation of metering and overview of curtailment helps shed light on what “Behind-The-Meter” implies. If a load is connected behind the meter of a facility, what changes is the net megawatt flow to the transmission network.
For example, if a miner connects a 10 MW bitcoin mine behind the meter at a 100 MW solar facility, the grid operator sees -10 MW at night, 0 MW when the solar panels are actually producing 10 MW, and 90 MW during peak production when the panels are producing 100 MW.
So, what impacts does a behind-the-meter bitcoin mine have?
Example: Bitcoin Mine At Curtailed Facility
Explanation of Bitcoin Mine Example
Solar and Wind Node Price: $0
City Node Price: $36
Combined Cycle Node Price: $36
In this example, a 15 MW and 10 MW bitcoin mine are connected behind-the-meter at the wind and solar farms. Note that while the wind and solar farms produce 60 MW & 40 MW respectively, the grid operator only sees in the market the net of their output, which is 45 MW (60 MW Wind - 15 MW BTC) and 30 MW (40 MW Solar - 10 MW BTC). This example uses a $30/MWh and $25/MWh bus-bar PPA for the behind-the-meter mines.
This behind-the-meter scheme has some immediate impacts to the original examples:
Keeping with our example diagram, the generators are already suffering heavy regular curtailment, so they have no financial downside to this arrangement. Their options are to be curtailed by 30 MW, or be curtailed by 40 MW, but either way they are otherwise making $0. Adding a behind-the-meter load to improve revenue is a no-brainer.
The tradeoff comes at other times of day, when maybe they aren’t curtailed. The miners want to run 24/7. Does the miner also take the generator's lucrative hours as well as their curtailed hours? If so, how does that math shake out?
As a behind-the-meter load, bitcoin mining can improve revenue for prospective new generators facing curtailment risk, in addition to improving revenue for existing curtailed projects.
With increased penetration of intermittent energy resources (e.g., solar and wind) and storage, the grid is no longer a simple, lucrative market for generation. For generation developers, modeling the future grid pricing and curtailment is increasingly sensitive to assumptions on how much of an impact increased renewable penetration will have on pricing & curtailment, what the fuel type and cost of the price-setter of the future will be, what renewable credits will exist, just how much storage will inevitably come online, and how renewables will price when they become price setters. That’s a lot of new variables to include in modeling assumptions!
The traditional risk profile for curtailment and low grid pricing (avoiding it) made sense. It isn’t prudent to add new generation facilities to locations where there is already enough of it. But if the stated goal of the Energy Transition is to replace thermals with intermittent sources like wind and solar, new builds need a new financial model that can handle curtailment figures in the double digits. Just because a wind farm is curtailed 15% of the time doesn't mean the other 85% of energy isn't useful for the grid.
But forcing the offtaker or consumer to pay a premium for that wasted energy doesn't make any sense. Why not find a way to use that curtailed energy? Storage is one answer here that surely has a place. The other is something called Power-to-X and is already a big deal in Europe.
Power-to-X, or direct behind-the-meter loads, can fix a lot of the previously mentioned buildout and revenue issues, all while still permitting, shoring up, or even subsidizing the traditional grid facing offtake. These behind-the-meter industrial loads can offer the generator a bus-bar PPA pricing scheme that is curtailment and basis (a complicated PPA term) free.
In certain Power-to-X structures like joint ventures or PPAs with upside share, these schemes also expose the generator to a revenue stream that is uncorrelated to their normal source. Imagine a generator joint venture with a colocated bitcoin mine that takes some portion of their revenue share in bitcoin. The generator may enjoy a bit of fun bitcoin exposure and pass that business line to their trading desk.
An important tradeoff here is that most loads don’t want to only run during curtailed hours. So, finding how much of a generator's total energy to give away to the co-located load in order to capitalize on reducing curtailment takes some nuanced financial modeling and strategy.
So why is bitcoin mining special? Couldn't the generator opt for green hydrogen, ammonia, or concrete?
Bitcoin is a special behind-the-meter load because it:
Before moving on, here’s some napkin math to show real numbers.
Whatsminer M3's are a model of older generation ASIC rated for 12 TH/s that pull 2 kW each. One MW of M3's would be 500 machines (1 MW / 2 KW). 500 machines would be 6000 TH. With a hashprice of $0.20, this would be 6000*0.2=$1200/day revenue. To get this to a MWh figure I divide by 24, since this is daily revenue. 1 MWh = 1200/(24hrs*1MW) = $50.
So, M3's break even at $50/MWh, or 5 cents/KWh. Much cheaper than what residential rates or industrial miners could probably handle. But still higher than what generators usually sign PPAs at. Taking the example further, Kaboomracks, a trusted second-hand ASIC retailer, was recently advertising M3's in bulk for $100/machine, a steep discount due to the fact that few operators can run them profitably.
Now, imagine buying 1 MW of M3's at $100/machine. That's $50,000. And pretend the miner owns or has connections with a solar or wind farm that already has electrical infrastructure to support one additional MW of load on site.
To breakeven on the investment while pretending infrastructure costs, maintenance, and used ASIC risks don’t exist, the generator or miner would only need to run their M3s for $50,000/$50 = 1000 hours. About 41 days.
For a generator that has $0 fuel cost and could be suffering from curtailment where their generation is actively being wasted, 41 days to breakeven on an investment that would then churn out $50/MWh doesn't seem like a bad deal. Way different than waiting out the next few years as the climate world tosses around the green hydrogen idea. And as of now, this asymmetry between older generation ASIC pricing compared to their gross revenue is huge.
In a scenario where miners don't operate behind-the-meter, but simply co-locate in front of the meter or locate as standalone facilities near curtailed generators, it is true that they provide some lift on the grid economics of the location. But this author thinks these scenarios are less interesting.
There is no curtailment protection provided to the generator, and no real savings on shared interconnection facilities or equipment. These arrangements also do not have as much alignment potential for joint venture upside sharing.
There is more clarity, however, for participating in ancillary services as a standalone or front-of-meter load (explained more in the next section). But this author isn't sure how that compares to the other benefits a miner gets going behind-the-meter. The real innovative opportunity, in this author’s opinion, is the behind-the-meter optionality and curtailment derisking for new generation development.
Readers will recall from the previous article in this series that bitcoin miners can sell ancillary services to grid operators if they qualify (meaning, if miners prove they can actually deploy the services when ordered).
This ability could be another feather in the behind-the-meter bitcoin miner’s hat, since miners could attach themselves to generators that are not able to provide the ancillary services themselves, adding to the flexibility of the grid at that location.
But remember the metering paradigm described above: grid operators are only able to dispatch these services based on the miner’s metered flow. This means that for a miner operating as a behind-the-meter load, while they could perform ancillary services, if the meter is spinning positive (for example, because the co-located generator is producing more power than the miner is consuming), the miner is unable to sell those ancillary services.
ERCOT has been working to solve these issues, and there is already a carveout in the rules for batteries to sell Ancillary Services behind-the-meter. But this would be one way to unlock value for the system at a low cost of some acceptable metering configurations.
It’s important to explore some of the downsides of bitcoin mining’s ability to improve generator revenue and underwrite new generation.
If bitcoin mining can improve revenue for curtailed facilities, could it not also improve revenue for non-curtailed facilities, and therefore consume cheap energy that would otherwise go to residents?
Yes. At least for now, bitcoin mining is still highly profitable with eye-popping revenue.
But a strong argument can be made for the network difficulty adjustment eventually bringing mining revenue so low that it only becomes lucrative for facilities to be mining bitcoin when fed with extremely low cost energy, or if the grid pricing is lower than simple payback of the generation asset. But arguing and proving this game theory on the network difficulty algorithm is a level of discussion just barely bubbling up into the landscape.
For now, it is true that mining bitcoin is an extremely lucrative endeavor for those with access to new generation machines at bulk rates, so long as miners also have access to relatively low electricity cost.
And if a miner builds a 100 MW wind or solar farm, but funds the generation by building a 100 MW bitcoin mine behind its meter, technically they aren’t adding energy to the grid and are still a net load on the system, raising costs.
To be clear, this isn’t exactly a problem in and of itself, although climate activists will say it is, since new renewable build out is meant to displace fossils. But it’s definitely not leveraging bitcoin mining to its full potential, and it isn’t difficult to imagine the problems that could arise from being able to add hundreds of megawatts of load twice as fast as a commensurate level of generation.
There are also downsides to bitcoin mining’s ability to shield generators from curtailment and congestion.
Readers will recall in the previous section on congestion how grid operators and planners use congestion as a signal that transmission upgrades may be required. This means that if generators suffering from curtailment and congestion suffer enough, they may eventually be rewarded by having their congestion relieved by a transmission upgrade.
But if bitcoin miners soak up low priced or curtailed hours, won’t they prevent that congestion from being noticed as an issue? It’s definitely possible.
If bitcoin miners espouse that they turn off their operation during energy scarcity or high pricing, ensuring that the energy they free up by turning off is actually able to make it to load is extremely important. If a miner’s operation obfuscates the need for a congestion upgrade, during scarcity events the power that the miner frees up by turning off may never be able to make it out of that congested area.
What’s the solution to this issue? Honestly, this author isn’t sure. It may take grid operator collaboration to identify how best to upgrade transmission infrastructure to provide ample energy during scarcity events, while balancing the idea that spending money on ‘non-existant-during-non-scarcity’ congestion may be a tough sell using traditional transmission planning paradigms.
For bitcoin mining to be an irrefutable good for the grid, this author proposes a couple ideals to strive toward. But to be clear, these are by no means a ‘minimum bar’ or 'requirement' that bitcoin should have to hurdle. But some miners can and should achieve these things, both because they align with the miners’ self-interest and because it would be a net benefit to society. For comparison, AWS data centers don't have to prove their irrefutable value to the grid, but they also aren't scaling at the speed of huge mining operations.
While mining is extremely profitable, bitcoin miners should strive to add additionality of energy to the power system. This mainly targets the mega mines with big balance sheets, but the concept is that if a grid adds 1 GWh of bitcoin mining, 1+ GWh of generation should be added as well. Ideally this would take the form of on-grid mega mines becoming vertically integrated, developing & owning their own generation assets, and offering their surplus to the grid while still offering to curtail bitcoin mining operations during times of scarcity. But it could also happen by strategic partnerships with generation developers, or with certain curtailment saving joint ventures that keep plants from retiring. This should also happen in regions of the world that have had problems financing their own energy. In this author’s opinion, leveraging mining to bring electricity to those without is one of the best stories bitcoin could write.
Bitcoin miners should coordinate with grid operators very closely to ensure that they provide maximal benefit to the grid during times of scarcity. This may mean coordinating with the operator such that they (the miners) are not added to models where congestion is analyzed for planning (system peak) purposes. Again, the system would not benefit if miners obfuscated congestion problems that would still exist after they turned themselves off.
Bitcoin miners should spend time educating grid operators on their function and advocating for a stronger grid generally. With a product based on pure energy, leveraging the current mining high margin to harden grid infrastructure is a perfect alignment. This author doesn’t know what form this could or should take. But bitcoin miners 'taking up' energy on the grid is such an easy argument to make, having a strong proactive strategy for providing a benefit is a must.
Congratulations and thanks to each reader who made it this far. Even though many of the concepts in this article were simplified, understanding even the basics of transmission, curtailment, congestion, and metering is not simple. And adding more complexity through analyzing how bitcoin can present some risks and value for the system is an even larger intellectual load.
After writing these several thousand words about bitcoin mining and the grid, this author thinks it's important to emphasize that while these concepts demonstrate some utility to the grid, they are still only a peripheral benefit of bitcoin mining. Bitcoin miners have one job: mine bitcoin. In this regard, comparing bitcoin miners to battery storage or green hydrogen is not a legitimate comparison because none of those things produce bitcoin.
No one writes articles about how steel mills solve grid problems or how much energy Facebook or TikTok pulls on an annual basis. These conversations center on mining because of Bitcoin’s lofty and abstract value proposition. Thus, miners become relegated to focusing on this secondary aspect (benefits to the grid) instead of having the primary purpose accepted as enough justification for mining to be a load on the system. Of course, it may not seem fair that bitcoin mining receives this type of scrutiny, but it’s our job to change this perception.
Bitcoin mining is threatening to add between 10% and 20% to ERCOT base load. People are going to have questions. Especially when they do not understand or value the function miners perform. When it comes to people who operate the world’s energy infrastructure, bitcoin miners need to find a way to include them in their circle.
This author does not have the answers on how to design or operate an energy market perfectly, nor how to operate or model bitcoin mines to maximize their value to the system. But hopefully, by gaining a better base of understanding of their possible interactions, miners can have more quality conversations about how to do just those things.
It’s also important to note that a lot of these concepts are in flux. With bitcoin miners popping up all over the grid, grid operators are scrambling to identify new rules, incentives, and methods for incorporating them effectively. It would not be surprising if some concepts in this article become outdated rather quickly. For more on that, stay tuned.
Read Part 1: Generators here.
This article was written for the Braiins blog by Blake King. Blake is a power engineer who builds and analyzes software models of electric grids. His views here do not reflect those of any of his past, present, or future employers. Follow Blake on Twitter.
Blake would like to thank his coworkers and industry colleagues for sharing their knowledge of power systems with him. Especially Joe Weatherly, his former manager in the Network Modeling department at ERCOT.
| Konference | Web | X | Where | When | Confirmed |
|---|---|---|---|---|---|
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell | ||
| Cell | Cell | Cell | Cell |
.svg){kind=small}
Bitcoin mining software company: Braiins Pool, Braiins OS & Stratum V2.
By miners, for miners.
Increase hashrate on your Bitcoin ASICs, improve efficiency as much as 25%, and mine on any pool or get 0% pool fees on Braiins Pool.
Reduce data transmission between your farm and pool by 95%. Configure parallel usage of multiple pools. Set a backup pool for the whole farm.
It focuses on making data transfers more efficient, reducing physical infrastructure requirements for mining operations, and increasing security
Industry leaders in transparency and innovation, with more than 1.25 million BTC mined since 2010.
Published
25.4.2022
In the second part of the Braiins series on mining and the electric grid, this article explores transmission, curtailment, and behind the meter with applications to bitcoin mining.
Table of Contents
This article is the second in a series about bitcoin mining and energy infrastructure. Each article offers an introductory level explanation of electric grids and their relationship with mining to better educate miners and other bitcoin investors.
Read Part 1: Generators here.
Part 1 of this series covered a general understanding of power generation, how grid operators balance generation with load, frequency regulation, and how bitcoin mining can fit into the puzzle.
As mining continues to integrate into the energy industry, it is also important for miners and bitcoin investors to learn the basics of power transmission, what terms like curtailment and behind-the-meter mean, and how bitcoin mining can fit in.
Feedback from Part 1 of this series was exceptional, and it’s this author’s hope that the information shared in this Part 2 will also be helpful.
This article is long. Even longer than Part 1. And the goal for this article is to strike a tone somewhere between light reading that’s not fully reflective of the system's complexities, and dense, truly-reflective-of-the-system reading that would be hard to digest for someone who doesn't work in power systems on a daily basis.
This author hopes to provide enough detail for those looking for it, while also not adding so much detail that readers lose interest. In striving for this goal, some representations made in this article will be missing precision or nuance. But the general concepts should be accurate.
Like the previous article, this one also uses the Electric Reliability Council of Texas (ERCOT) electric grid as example material. Some of the concepts explained throughout this article would map onto other grids, while others would not.
Transmission lines are a combination of wood or steel transmission towers and the conductors they carry. These lines act as power pathways that connect generator substations to transmission level switching stations and load serving substations.
Substations are structures where power is transformed from high levels to lower levels of voltage (or vice versa), and/or where transmission lines terminate and others start.
Switching stations are a type of substation used to sectionalize transmission lines, or allow for certain lines to be switched out while allowing power to flow through other paths. Switching stations don't normally have different voltage levels, and therefore don't have transformers. At a high level, substations have different configurations of breakers that make connecting to them easier or more difficult. Substations can also be built to have extra room for future expansions.
That’s enough of a substation vocabulary lesson for now.
Because of certain physical qualities of electricity, moving power at higher voltage levels minimizes losses. For this purpose, when moving power over long distances power systems engineers try to use the highest voltage level practicable. Generator substations step voltage up to transmission levels, and load serving substations step the voltage down to a level useful for loads.
Transmission lines, substations, and generators in their aggregate form a high voltage transmission network, which is monitored and managed for reliability by a central grid operator (like ERCOT).
The decision of what type of transmission line to build is normally dependent on the purpose of the line, the local geography, vertical clearance (meaning, the amount of space available below the conductors), right-of-way, local climate, and cost.
A fun rule of thumb: Voltage levels are normalized to a set. Some common levels in the 60 HZ world are 13.8 kV, 34.5 kV, 69 kV, 115 kV, 138 kV, 220 kV, and 345 kV. Count the insulators (small ceramic saucers between the conductors and the towers) to estimate the voltage level of a line (roughly 10 or 11 kV per insulator). See below for some images of different tower types. This author's wife loves when he calls out tower type and voltage levels as they drive past lines.
The series of images below show H-Frame (often wood or steel), Monopole (often steel), Lattice (often steel), and Monopole (with underslung distribution).
Transmission lines are different from the wooden pole lines seen around residential neighborhoods. Those are distribution lines, and they are an integral part of a power system’s distribution infrastructure. Distribution lines are used to distribute power from transmission level substations to distribution level loads. These lines are normally monitored and managed by a local utility, not the main “headline” grid operator. But sometimes it can be the same entity. It just depends on where you are.
In ERCOT, transmission level equipment is normally considered 69 kV and above. This level of equipment has a different level of scrutiny and operation than distribution equipment, since impacts on this higher voltage level would cause subsequent problems for downstream systems. 345 kV lines and higher have their own designation as “Extra High Voltage” and receive even more scrutiny.
In the image below, transmission lines are shown in the middle, gray section between power generators and distribution lines.
Before continuing even deeper in discussing power transmission, it will be helpful to briefly revisit some of the power generation information from the previous article of this series on mining and the grid and add some graphics.
Readers will recall from the previous article that generators on a power grid with a nodal market such as ERCOT are dispatched in ascending order from least marginal cost of production (cheapest fuel cost) upward towards the most expensive marginal cost of production (most expensive fuel). The most expensive megawatt (MW) that matches the total system generation with load then sets the price for all generators that were dispatched in that interval.
This means that generators in a nodal market want to:
Solar and wind generators, since they have no fuel cost to account for, are $0 marginal cost generators, which means that they bid $0 into the market and are subsequently by default online. The grid operator accounts for the forecasted megawatt contribution from these zero-cost generators first before dispatching more expensive generation.
What does this look like?
The diagrams below illustrate a few scenarios with different fuel types competing to supply power to a city. Each example diagram is followed by a scenario explanation. These figures are helpful to demonstrate a couple concepts explained in this article.
A caveat for other power engineers: These diagrams and scenarios ignore constraints for now in order to absolutely hammer down the concept of marginal pricing.
Example 1: Wind Or Solar Is Marginal
Explanation for Example 1
Solar and Wind Node Price: $0
City Node Price: $0
The grid operator is informed that wind and solar forecasts call for the facilities to produce 60 MW & 40 MW respectively, totaling 100 MW, which perfectly matches the 100 MW forecasted demand of the city for the next five minute interval (unrealistic... but theoretically possible).
Since the wind and solar generators both bid the market at $0 and they provide the last MW needed to match the city demand, they set the marginal price for the system to $0, and they subsequently get paid $0 for each MW they produced in that interval. The grid operator does not need to move upward from $0 to dispatch more expensive generation, and so the peaker and combined cycle plants remain offline and don't get paid at all (for energy).
ERCOT uses ancillary services that it procured in the day ahead market to manage fluctuations in frequency between this interval and the next. An example of this might be that the peaker plant sold non-spinning reserve to ERCOT yesterday, such that if ERCOT needed them to start up really fast, they could.
This is also an illustration of how wind and solar don’t ever plan on actually being the marginal unit, because if they did, bidding $0 wouldn't make any sense. They would never make any money!
Example 2: Combined Cycle (Natural Gas) Is Marginal
Explanation for Example 2
Solar and Wind Node Price: $20
City Node Price: $20
Combined Cycle Node Price: $20
The grid operator is informed that wind and solar forecasts call for the facilities to produce 50 MW and 40 MW respectively, totaling 90 MW, and the city load for the next five minutes is still 100 MW, so the grid operator looks to the available, least expensive generation.
The peaker plant says it can provide 10 MW for $100/MWh, and the combined cycle plant says it can provide 10 MW for only $20/MWh. Since $20 is cheaper than $100, the grid operator dispatches the combined cycle plant instead of the more expensive peaker.
The combined cycle plant ramps to 10 MW at its marginal rate of $20/MWh, and it therefore sets the price at $20/MWh for everyone.
Example 3: Peaker Plant (Natural Gas) Is Marginal
Explanation for Example 3
Solar and Wind Node Price: $100
City Node Price: $100
Peaker Price: $100
The grid operator is informed that wind and solar forecasts call for the facilities to produce 50 MW and 40 MW respectively, totaling 90 MW, and the city load for the next five minutes is still 100 MW, so the grid operator looks to the available, least expensive generation.
The peaker plant says it can provide 10 MW for $100/MWh. The combined cycle is actually offline for maintenance as part of a planned outage, so the grid operator has no choice but to dispatch the more expensive peaker plant.
The peaker turns on for 10 MW at its marginal rate of $100/MWh, and it sets the price at $100/MWh for everyone.
After considering the above examples, hopefully readers can see just how important marginal pricing is for power generators. Generators plan their multi-decade lifespan around variables like how often their marginal price will enable them to be online, how expensive their fuel may become, and how often they will face outages for repairs or maintenance.
On a related note, it’s interesting to observe that natural gas facilities like the Combined Cycle Plants and Peaker Plants in the above examples are the primary price-setters in today’s markets. And what do they use to set their prices? Their fuel costs. Because of this, the wholesale electricity market is really a pseudo-index of natural gas prices.
Congratulations for progressing through the basics of transmission lines and a refresher (with examples) on power generation. Let’s keep going!
The previous section showed diagrams of how much generators are paid as their $/MWh nodal price, which is set by the marginal unit. But what rate does the 100 MW city load in the diagram pay?
There are various strategies for these situations, especially if the load is a municipality or utility that needs to physically take ownership of the power. But for loads like bitcoin miners, ERCOT has a Load Zone structure where all the nodes in a certain geography (North, West, East, Austin, LCRA) are load-weight-averaged together with delivery and other charges added in some fashion.
Bitcoin miners or other load types normally hedge their variable Load Zone rate by entering into fancy power purchase agreements (PPAs) with nearby generators. This is usually a financial “swap” whereby when the generator’s nodal or average geographic area (called a hub) price goes up, the bitcoin miner is exposed to that increased price, which then offsets their commensurate increased Load Zone rate (and vice versa for the downside, when the generator would suffer from decreased pricing).
Provided that the generator's pricing correlates with the bitcoin miners’ load zone pricing, the bitcoin miner should be safe. This type of agreement can set both the load and generator to an effectively fixed price that is dependent upon the modeling assumptions that went into the PPA structure, plus whatever risks still exist (basis, volume, etc.)
Even more sophisticated loads and generators employ power traders to execute optimal strategies here on top of their PPAs. This gets complicated rather quickly.. but the important part to understand is that there are many creative financial instruments used in these energy markets to hedge your floating wholesale price.
Now, back to transmission.
The transmission lines in our diagrams above were a bit unrealistic, because in the real world, transmission lines can only carry so much power. If a grid operator sent more power down a line than it could handle, they would damage the equipment. To stop this from happening, an entire field of electrical engineering called protection engineering uses coordination and combinations of circuit breakers, fuses, relays, and other equipment to isolate equipment if they sense too much current.
Also to prevent too much power being sent on these lines, grid operators like ERCOT spend a majority of their computing power solving the math problem of how to match generation to load not only at the least cost, but at the least cost without overloading any equipment.
This means that ERCOT’s dispatch for generation should never result in overloads, so the aforementioned protection engineering is primarily concerned with isolating faults or trips, what are called contingencies. But that is a complex topic for another article. Back to the fact that transmission lines have ratings…
Revisiting the first example diagram shared in this article, notice the inclusion of a 50 MW transmission line rating for the West to East line.
Warning: these next few diagrams are going to be doozies, but the reader should know they won’t be too bad. And any reader that fully understands them will be ready to apply for a Market Engineer position at ERCOT!)
Example 4: Wind and Solar Curtailment
Explanation for Example 4
Solar and Wind Node Price: $0
City Node Price: $36
Combined Cycle Node Price: $36
This example is a modification of the first scenario used in this article, but with a 50 MW limit on the transmission line. In the original example, the solar and wind produced 100 MW, all of which flowed to the city and set the clearing price for the system & their nodes at $0. In this modified example, some key things are different:
The solar irradiance and wind speed of that day could have resulted in the plants producing a combined 100 MW, which would have caused an overload on our 50 MW transmission line. To prevent this from happening, the grid operator had to partially curtail the wind and solar facilities. Simply put, there wasn’t enough room on the transmission lines between the generation and load to handle that amount of power. In other markets, this is solved by forcing generators to pay for upgrades before they come online (PJM). In ERCOT, it’s solved by just curtailing plants in real-time.
The grid operator performs this curtailment by ordering the generators to produce less power by sending them instructions (base points), which the wind and solar have to follow or else they face fines. Along with ordering the facilities down, the grid operator also lowers their nodal price using some fancy math described below.
Readers will notice in the modified diagram that the wind and solar farm are paid $0/MWh instead of the clearing price of $36/MWh set by the Combined Cycle plant. In this new scenario, the grid operator takes the clearing price set by the combined cycle plant ($36) and subtracts something called a “Shadow Price” ($36) multiplied by a “Shift Factor” (1.0).
Yes, even more vocabulary to learn, so let’s pause briefly and define these terms:
Notice also that since the transmission line could not carry all 100 MW of that sweet $0/MWh solar and wind to the load, more expensive generation needed to be turned on to make sure the city load was met.
As a result, the combined cycle plant was dispatched up to 50 MW and the clearing price was set by its marginal cost to produce that amount - $36/MWh. This marginal price then feeds into the price for all generators on the grid. More room on transmission would’ve kept the grid operator from having to go that far up the bid stack, and thus would have resulted in a lower clearing price, and an overall lower cost to run the system. In other words, curtailment results in a less efficient, more costly system.
In power system terms what’s happening here is called a Binding Constraint. The constraint was the limit on the line (50 MW). It’s binding because there’s no more room on the line (it’s full, and was at risk of overloading), and so the grid operator had to act on it in order to prevent an overload. Each Binding Constraint on a system will have its own marginal price-setting generator. Yes, the author knows they previously said there was only one price setter in a system.. But the author lied!
In some five minute intervals through the day there are no binding constraints, and thus there is in fact only one price setter. Otherwise, each binding constraint will have its own price setter, in addition to the clearing price.
A small caveat: the flows that determine if a constraint is binding or not are “post-contingency” flows. Which means the grid is actually operated such that if it loses any single line or element due to a fault or trip, the flows that were going along that element could redirect elsewhere and not overload anything else after being redirected. This method of dispatching is done to ensure operators can isolate single contingencies, such that one fault does not cause a domino effect tripping other elements and causing a blackout. The example used above is a bit too simple to include this nuance, but it's still good to know.
Simple enough, right?
Even though the big headline name for what was described in the last example is curtailment– where solar and wind are ordered to produce less power than they otherwise could–the same scenario also describes congestion. And congestion isn’t unique to renewables. It happens any time that cheaper generation can’t get to load due to transmission bottlenecks.
And just to complete the full definition, curtailment could also happen if there was more solar and wind generation available than load. For instance, in our first ever example with the infinite transmission line, if there was actually 110 MW of solar and wind but still only 100 MW of city load, the operator would still have to partially curtail those facilities. Back to our congested line:
In power systems jargon, the congestion scenario is described like this:
“There was congestion along the line going into the city, which resulted in 50 megawatts of curtailment in the west & higher pricing for city residents.”
In the real world, congestion is a huge component of wholesale electricity pricing, and it’s the reason why pricing differs between nodes in a network. If congestion never happened, all nodes would be the same price.
The chart below shows real-time market pricing in ERCOT at the time of writing this article. Readers will notice how some price pockets have significantly higher or lower prices than other pockets. These price differences are due to congestion that prevents generation within those low priced areas from getting to loads in the higher priced areas. Thus, more expensive generation is turned on in the higher priced areas.
At the risk of being annoyingly repetitive, it’s important to emphasize that this article’s example power network is just a four bus system with one transmission line and four generators. The simplicity of the diagram is also why it worked out perfectly that the nodes ended up clearing at the price they bid.
In the real world, ERCOT, for example, has 700+ generators, thousands of buses, thousands of lines, and the algorithm that determines pricing is re-run every five minutes, ordering all these elements in a thousand by thousand matrix to perform the correct optimizations of flows. In simplifying all this complexity, this article’s example is really only useful for the topics discussed here: describing congestion, curtailment, and marginal pricing generally.
What is negative pricing, and how does it work?
Wind Production Tax Credits (PTCs) have been a part of the power generation industry for many decades. The original PTC began in 1992, it has lapsed numerous times since, but has always been reinstated.
The PTC is a $/MWh credit for production for a wind facility that meets the eligibility requirements for the tax credit program. Depending on what year a wind generator comes online, they either receive the full credit amount or a phased proportion of the full amount. This means that some wind facilities may get the full amount of $26/MWh, and some may get only 60% (~$16/MWh), etc. After a project is online for ten years, the PTC runs out, after which the generator only gets grid pricing.
So, for every megawatt hour a wind facility produces while eligible for PTC, it will receive tax credit payments. Readers will remember previous discussions about the “bid stack” where grid operators move from cheapest generation upwards. Since some wind generators receive this credit, they can bid into the market at a negative price since they know they will receive the $/MWh credit for their production anyways. Bidding into the negative ensures that these generators are curtailed after the competing $0/MWh solar or newer wind that receives less PTC..
Let’s revisit the power system diagram again, this time with the wind bid changed from $0 down to $-25. Continue reading for an explanation of this diagram.
Negative Pricing Example
Explanation for Negative Pricing Example
Solar and Wind Node Price: $-25
City Node Price: $36
Combined Cycle Node Price: $36
In the example, because wind bids $-25, significantly lower than solar ($0), the grid operator fully curtails the solar facility before curtailing the wind facility at all.
Since the 50 megawatts of wind power is still unable to meet the city load, the combined cycle plant turns on to meet the demand, and sets the clearing price at $36/MWh.
The grid operator then takes the clearing price set by the combined cycle plant ($36) and again subtracts the “Shadow Price” ($61) multiplied by the “Shift Factor” (1.0). To revisit shadow prices quickly:
What all of this means is that the price at the wind farm is now $-25, and the price at the combined cycle facility is $36. Note that the wind farm bidding negative doesn't actually have to pay to produce, because the PTC payment brings them up to positive.
Here are a few high-level observations about curtailment due to transmission and congestion:
So what’s the solution?
A quick way to eliminate curtailment and congestion would be to increase the transmission line rating of every line on the grid by a factor of 10 and add synchronous condensers everywhere that there are stability concerns.
But that wouldn't make sense.. How much would that cost!? These are the questions grid operators are tasked with balancing, and the influx of intermittent generation and bitcoin mining loads don't make the job any easier.
This is what normally happens.
If congestion in a specific area is identified as a costly problem, transmission planning engineers like this author will analyze models of the system to understand the expected cost to upgrade a congested transmission line and compare that cost to the total savings to the consumer.
Normally there is a rule that sets expectations for a number of years wherein the cost of the upgrade is paid back. For example, if a transmission line upgrade costs $200 million but the consumer (system cost) saves $50 million per year, the consumer breaks even and begins saving money after four years.
Transmission planning engineers then say, “This makes economic sense. Let’s propose it!”
Curtailment introduces risks and disadvantages to generators.
For generation developers, assumptions about a prospective project location’s future grid pricing and curtailment are identified early on using power system models. Most of the time, generators that actually get built have modeled future revenue that looks something like a low risk, low return with contracted payback after a few decades.
Generator developers secure this contracted payback by selling their future projects power as a PPA. In a PPA, a generator contracts some percentage of the power they will produce to either a utility that needs the capacity, an industrial customer looking to hedge their exposure to real-time rates, or an industrial customer looking for renewable energy credits (RECs) to offset their use and claim carbon neutrality. Normally these PPA counterparties, called “offtakers” put out request for proposals (RFPs), which generation developers competitively bid to win. The generator will retain some smaller uncontracted percentage of their future production to sell at the market price (merchant strip).
A signed PPA secures contracted revenue for the generator, which is then used to market the project to a buyer. The buyers of these projects are usually investment funds looking for long-term, low-risk returns which the PPA provides.
Since generators don’t get paid for their curtailed energy, generation developers will incorporate their expected curtailment into their PPA price that they submit to RFPs. For example, if a generator expects to be curtailed 50% of the time, they would need to double the PPA price to meet the same returns as if they had no curtailment. But if a generator only expects to be curtailed 1% of the time, the PPA only needs a strike price increase of 1% to achieve those same economics. This is why selecting a good spot on the grid with low curtailment is important, as it will allow a generator to price PPAs low and win RFPs.
This is where a lot of prospective generation projects die. No project is going to win an RFP and secure financing if their PPA strike price is 20% higher than their competitors. Curtailment risk of any significant size is a project killer. Tangentially, this author often wonders how many projects in the ERCOT generation queue are just a few curtailment percentage points away from becoming real projects, but are just sitting on the shelf.
And with increased renewable buildout, wind and solar facilities are increasingly cannibalizing their own pricing. Since these generators are all online during the same hours of the day, every project pushes their own grid pricing downward and increases their own curtailment (see: Duck Curve in California, Armadillo Curve in ERCOT).
While it may seem obvious, curtailment is a big waste of energy. If you have a wind or solar farm that could produce 200 MW, but is curtailed to 0 MW, there is a huge opportunity cost for you. For generators, being curtailed is akin to throwing money down the drain, even if they already included curtailment assumptions into their PPA and won an RFP. The generator would still prefer to monetize that curtailed energy.
Even if a wind or solar farm looks great on paper, stability concerns could force the grid operator to limit exports from their area. Now, all of a sudden, the new generator is being curtailed a lot more than was expected, and their revenue is terrible. To be specific:
In the case of a fixed volume PPA where the facility owes an offtaker a certain volume of megawatt hours per month, the generator isn’t able to deliver the contracted amount of power because of curtailment. Even if the fixed volume amount is perfectly aligned with what the generator could be producing, since they can’t deliver those MW’s to the grid, the generator still has to buy the balance of the owed energy on the market and make the offtaker whole.
In a different scenario with an as-generated PPA, the generator isn’t able to sell the amount of power they previously planned to sell, so they end up simply missing revenue targets.
And if a generator does not have a PPA and is just selling energy to the market at the wholesale rate, being curtailed is still just wasting energy they could be generating and doing something with.
Transmission upgrades take years, and in a 30 year project those years really matter. For a wind project, those are also years that the facility could be earning PTCs for the megawatt hours that are curtailed. Remember, a qualifying project only has 10 years to collect PTCs before they run out, tick tock!
It is no small feat reading through all of the diagrams, jargon, scenarios, and explanations from the past 15 pages or so. This author promises this article is getting close to applying these concepts to bitcoin mining.
Mining advocates often tout their industry as fitting into this picture of curtailment and generation growth somehow. The concept of metering will help to clarify the potential alignment.
Metering Explained
A generation or load project that connects to the transmission level network must connect through something called a meter, which are very similar to the residential meters most readers are probably familiar with. These meters have different names in different systems, but in ERCOT they are called EPS (ERCOT Polled Settlement) meters.
EPS meters are normally connected at the Point Of Interconnect (POI), where the generator or load owned equipment ends and the transmission operator equipment begins. Wherever they are connected, the meters must be able to read the total megawatts being injected to or withdrawn from the transmission network. In other words, the meter is what actually matters when it comes to paying or being paid for electricity.
Aside from settlement (getting paid or paying), the meter is also how ERCOT monitors and dispatches generators or loads as resources.
Hopefully that quick explanation of metering and overview of curtailment helps shed light on what “Behind-The-Meter” implies. If a load is connected behind the meter of a facility, what changes is the net megawatt flow to the transmission network.
For example, if a miner connects a 10 MW bitcoin mine behind the meter at a 100 MW solar facility, the grid operator sees -10 MW at night, 0 MW when the solar panels are actually producing 10 MW, and 90 MW during peak production when the panels are producing 100 MW.
So, what impacts does a behind-the-meter bitcoin mine have?
Example: Bitcoin Mine At Curtailed Facility
Explanation of Bitcoin Mine Example
Solar and Wind Node Price: $0
City Node Price: $36
Combined Cycle Node Price: $36
In this example, a 15 MW and 10 MW bitcoin mine are connected behind-the-meter at the wind and solar farms. Note that while the wind and solar farms produce 60 MW & 40 MW respectively, the grid operator only sees in the market the net of their output, which is 45 MW (60 MW Wind - 15 MW BTC) and 30 MW (40 MW Solar - 10 MW BTC). This example uses a $30/MWh and $25/MWh bus-bar PPA for the behind-the-meter mines.
This behind-the-meter scheme has some immediate impacts to the original examples:
Keeping with our example diagram, the generators are already suffering heavy regular curtailment, so they have no financial downside to this arrangement. Their options are to be curtailed by 30 MW, or be curtailed by 40 MW, but either way they are otherwise making $0. Adding a behind-the-meter load to improve revenue is a no-brainer.
The tradeoff comes at other times of day, when maybe they aren’t curtailed. The miners want to run 24/7. Does the miner also take the generator's lucrative hours as well as their curtailed hours? If so, how does that math shake out?
As a behind-the-meter load, bitcoin mining can improve revenue for prospective new generators facing curtailment risk, in addition to improving revenue for existing curtailed projects.
With increased penetration of intermittent energy resources (e.g., solar and wind) and storage, the grid is no longer a simple, lucrative market for generation. For generation developers, modeling the future grid pricing and curtailment is increasingly sensitive to assumptions on how much of an impact increased renewable penetration will have on pricing & curtailment, what the fuel type and cost of the price-setter of the future will be, what renewable credits will exist, just how much storage will inevitably come online, and how renewables will price when they become price setters. That’s a lot of new variables to include in modeling assumptions!
The traditional risk profile for curtailment and low grid pricing (avoiding it) made sense. It isn’t prudent to add new generation facilities to locations where there is already enough of it. But if the stated goal of the Energy Transition is to replace thermals with intermittent sources like wind and solar, new builds need a new financial model that can handle curtailment figures in the double digits. Just because a wind farm is curtailed 15% of the time doesn't mean the other 85% of energy isn't useful for the grid.
But forcing the offtaker or consumer to pay a premium for that wasted energy doesn't make any sense. Why not find a way to use that curtailed energy? Storage is one answer here that surely has a place. The other is something called Power-to-X and is already a big deal in Europe.
Power-to-X, or direct behind-the-meter loads, can fix a lot of the previously mentioned buildout and revenue issues, all while still permitting, shoring up, or even subsidizing the traditional grid facing offtake. These behind-the-meter industrial loads can offer the generator a bus-bar PPA pricing scheme that is curtailment and basis (a complicated PPA term) free.
In certain Power-to-X structures like joint ventures or PPAs with upside share, these schemes also expose the generator to a revenue stream that is uncorrelated to their normal source. Imagine a generator joint venture with a colocated bitcoin mine that takes some portion of their revenue share in bitcoin. The generator may enjoy a bit of fun bitcoin exposure and pass that business line to their trading desk.
An important tradeoff here is that most loads don’t want to only run during curtailed hours. So, finding how much of a generator's total energy to give away to the co-located load in order to capitalize on reducing curtailment takes some nuanced financial modeling and strategy.
So why is bitcoin mining special? Couldn't the generator opt for green hydrogen, ammonia, or concrete?
Bitcoin is a special behind-the-meter load because it:
Before moving on, here’s some napkin math to show real numbers.
Whatsminer M3's are a model of older generation ASIC rated for 12 TH/s that pull 2 kW each. One MW of M3's would be 500 machines (1 MW / 2 KW). 500 machines would be 6000 TH. With a hashprice of $0.20, this would be 6000*0.2=$1200/day revenue. To get this to a MWh figure I divide by 24, since this is daily revenue. 1 MWh = 1200/(24hrs*1MW) = $50.
So, M3's break even at $50/MWh, or 5 cents/KWh. Much cheaper than what residential rates or industrial miners could probably handle. But still higher than what generators usually sign PPAs at. Taking the example further, Kaboomracks, a trusted second-hand ASIC retailer, was recently advertising M3's in bulk for $100/machine, a steep discount due to the fact that few operators can run them profitably.
Now, imagine buying 1 MW of M3's at $100/machine. That's $50,000. And pretend the miner owns or has connections with a solar or wind farm that already has electrical infrastructure to support one additional MW of load on site.
To breakeven on the investment while pretending infrastructure costs, maintenance, and used ASIC risks don’t exist, the generator or miner would only need to run their M3s for $50,000/$50 = 1000 hours. About 41 days.
For a generator that has $0 fuel cost and could be suffering from curtailment where their generation is actively being wasted, 41 days to breakeven on an investment that would then churn out $50/MWh doesn't seem like a bad deal. Way different than waiting out the next few years as the climate world tosses around the green hydrogen idea. And as of now, this asymmetry between older generation ASIC pricing compared to their gross revenue is huge.
In a scenario where miners don't operate behind-the-meter, but simply co-locate in front of the meter or locate as standalone facilities near curtailed generators, it is true that they provide some lift on the grid economics of the location. But this author thinks these scenarios are less interesting.
There is no curtailment protection provided to the generator, and no real savings on shared interconnection facilities or equipment. These arrangements also do not have as much alignment potential for joint venture upside sharing.
There is more clarity, however, for participating in ancillary services as a standalone or front-of-meter load (explained more in the next section). But this author isn't sure how that compares to the other benefits a miner gets going behind-the-meter. The real innovative opportunity, in this author’s opinion, is the behind-the-meter optionality and curtailment derisking for new generation development.
Readers will recall from the previous article in this series that bitcoin miners can sell ancillary services to grid operators if they qualify (meaning, if miners prove they can actually deploy the services when ordered).
This ability could be another feather in the behind-the-meter bitcoin miner’s hat, since miners could attach themselves to generators that are not able to provide the ancillary services themselves, adding to the flexibility of the grid at that location.
But remember the metering paradigm described above: grid operators are only able to dispatch these services based on the miner’s metered flow. This means that for a miner operating as a behind-the-meter load, while they could perform ancillary services, if the meter is spinning positive (for example, because the co-located generator is producing more power than the miner is consuming), the miner is unable to sell those ancillary services.
ERCOT has been working to solve these issues, and there is already a carveout in the rules for batteries to sell Ancillary Services behind-the-meter. But this would be one way to unlock value for the system at a low cost of some acceptable metering configurations.
It’s important to explore some of the downsides of bitcoin mining’s ability to improve generator revenue and underwrite new generation.
If bitcoin mining can improve revenue for curtailed facilities, could it not also improve revenue for non-curtailed facilities, and therefore consume cheap energy that would otherwise go to residents?
Yes. At least for now, bitcoin mining is still highly profitable with eye-popping revenue.
But a strong argument can be made for the network difficulty adjustment eventually bringing mining revenue so low that it only becomes lucrative for facilities to be mining bitcoin when fed with extremely low cost energy, or if the grid pricing is lower than simple payback of the generation asset. But arguing and proving this game theory on the network difficulty algorithm is a level of discussion just barely bubbling up into the landscape.
For now, it is true that mining bitcoin is an extremely lucrative endeavor for those with access to new generation machines at bulk rates, so long as miners also have access to relatively low electricity cost.
And if a miner builds a 100 MW wind or solar farm, but funds the generation by building a 100 MW bitcoin mine behind its meter, technically they aren’t adding energy to the grid and are still a net load on the system, raising costs.
To be clear, this isn’t exactly a problem in and of itself, although climate activists will say it is, since new renewable build out is meant to displace fossils. But it’s definitely not leveraging bitcoin mining to its full potential, and it isn’t difficult to imagine the problems that could arise from being able to add hundreds of megawatts of load twice as fast as a commensurate level of generation.
There are also downsides to bitcoin mining’s ability to shield generators from curtailment and congestion.
Readers will recall in the previous section on congestion how grid operators and planners use congestion as a signal that transmission upgrades may be required. This means that if generators suffering from curtailment and congestion suffer enough, they may eventually be rewarded by having their congestion relieved by a transmission upgrade.
But if bitcoin miners soak up low priced or curtailed hours, won’t they prevent that congestion from being noticed as an issue? It’s definitely possible.
If bitcoin miners espouse that they turn off their operation during energy scarcity or high pricing, ensuring that the energy they free up by turning off is actually able to make it to load is extremely important. If a miner’s operation obfuscates the need for a congestion upgrade, during scarcity events the power that the miner frees up by turning off may never be able to make it out of that congested area.
What’s the solution to this issue? Honestly, this author isn’t sure. It may take grid operator collaboration to identify how best to upgrade transmission infrastructure to provide ample energy during scarcity events, while balancing the idea that spending money on ‘non-existant-during-non-scarcity’ congestion may be a tough sell using traditional transmission planning paradigms.
For bitcoin mining to be an irrefutable good for the grid, this author proposes a couple ideals to strive toward. But to be clear, these are by no means a ‘minimum bar’ or 'requirement' that bitcoin should have to hurdle. But some miners can and should achieve these things, both because they align with the miners’ self-interest and because it would be a net benefit to society. For comparison, AWS data centers don't have to prove their irrefutable value to the grid, but they also aren't scaling at the speed of huge mining operations.
While mining is extremely profitable, bitcoin miners should strive to add additionality of energy to the power system. This mainly targets the mega mines with big balance sheets, but the concept is that if a grid adds 1 GWh of bitcoin mining, 1+ GWh of generation should be added as well. Ideally this would take the form of on-grid mega mines becoming vertically integrated, developing & owning their own generation assets, and offering their surplus to the grid while still offering to curtail bitcoin mining operations during times of scarcity. But it could also happen by strategic partnerships with generation developers, or with certain curtailment saving joint ventures that keep plants from retiring. This should also happen in regions of the world that have had problems financing their own energy. In this author’s opinion, leveraging mining to bring electricity to those without is one of the best stories bitcoin could write.
Bitcoin miners should coordinate with grid operators very closely to ensure that they provide maximal benefit to the grid during times of scarcity. This may mean coordinating with the operator such that they (the miners) are not added to models where congestion is analyzed for planning (system peak) purposes. Again, the system would not benefit if miners obfuscated congestion problems that would still exist after they turned themselves off.
Bitcoin miners should spend time educating grid operators on their function and advocating for a stronger grid generally. With a product based on pure energy, leveraging the current mining high margin to harden grid infrastructure is a perfect alignment. This author doesn’t know what form this could or should take. But bitcoin miners 'taking up' energy on the grid is such an easy argument to make, having a strong proactive strategy for providing a benefit is a must.
Congratulations and thanks to each reader who made it this far. Even though many of the concepts in this article were simplified, understanding even the basics of transmission, curtailment, congestion, and metering is not simple. And adding more complexity through analyzing how bitcoin can present some risks and value for the system is an even larger intellectual load.
After writing these several thousand words about bitcoin mining and the grid, this author thinks it's important to emphasize that while these concepts demonstrate some utility to the grid, they are still only a peripheral benefit of bitcoin mining. Bitcoin miners have one job: mine bitcoin. In this regard, comparing bitcoin miners to battery storage or green hydrogen is not a legitimate comparison because none of those things produce bitcoin.
No one writes articles about how steel mills solve grid problems or how much energy Facebook or TikTok pulls on an annual basis. These conversations center on mining because of Bitcoin’s lofty and abstract value proposition. Thus, miners become relegated to focusing on this secondary aspect (benefits to the grid) instead of having the primary purpose accepted as enough justification for mining to be a load on the system. Of course, it may not seem fair that bitcoin mining receives this type of scrutiny, but it’s our job to change this perception.
Bitcoin mining is threatening to add between 10% and 20% to ERCOT base load. People are going to have questions. Especially when they do not understand or value the function miners perform. When it comes to people who operate the world’s energy infrastructure, bitcoin miners need to find a way to include them in their circle.
This author does not have the answers on how to design or operate an energy market perfectly, nor how to operate or model bitcoin mines to maximize their value to the system. But hopefully, by gaining a better base of understanding of their possible interactions, miners can have more quality conversations about how to do just those things.
It’s also important to note that a lot of these concepts are in flux. With bitcoin miners popping up all over the grid, grid operators are scrambling to identify new rules, incentives, and methods for incorporating them effectively. It would not be surprising if some concepts in this article become outdated rather quickly. For more on that, stay tuned.
Read Part 1: Generators here.
This article was written for the Braiins blog by Blake King. Blake is a power engineer who builds and analyzes software models of electric grids. His views here do not reflect those of any of his past, present, or future employers. Follow Blake on Twitter.
Blake would like to thank his coworkers and industry colleagues for sharing their knowledge of power systems with him. Especially Joe Weatherly, his former manager in the Network Modeling department at ERCOT.
Read Privacy Policy.
12.12.2024
4.12.2024
11.11.2024
Read Privacy Policy.
.svg)