BLACKOUTS

California's electrical grid operator (CAISO) declared a stage 3 emergency last Friday, initiating a series of rolling blackouts across the state. Stage 3 emergencies are not trivial events; the last stage 3 emergency was in 2001 during the California energy crisis, which forced a halt to the state's deregulation process and initiated 15 years of litigation and regulatory proceedings. 

The initial media reporting on the blackouts fell into two buckets. The first focused on the idea that increasing renewable energy in California was to blame for the outages. This idea is not entirely inaccurate, but the argument fails to consider the complexity of the situation. It is also an argument that makes for poorly written, bombastic, and politically motivated stories when the reality is we are dealing with a systems engineering problem. The second story, stranger in our minds, was a narrative about electrical under-supply. The state's independent system operator did not say there were congestion issues or burned wires, but merely a lack of power necessary to meet demand – an extremely uncommon cause for initiating rolling blackouts. 

Our research indicates that demand response programs and energy storage have not scaled sufficiently to accommodate the number of renewable resources brought online over the last several years. 

Dynamic load flexibility

Power markets are (traditionally) characterized by load-following generation assets. Production of electricity follows demand, not the other way around. This works when the system is built with nuclear, gas, coal (assets you can turn on/off[1]), but not as well with renewables. Variable resources cannot respond as well to changes in demand. 

As California increases the share of renewable energy production, the amount of production capacity that cannot follow demand increases. This lack of flexibility is exacerbated when retail markets are disconnected from wholesale markets. When prices rise in wholesale markets (as they did last weekend in California[2]), the increases are not passed on to retail markets, so customers keep consuming the same amount of energy.[3]

Having load (demand) be flexible relative to the underlying cost of delivering energy can be very powerful. It is likely necessary for a system with increasing variable supply. When supply is abundant, in the presence of dynamic pricing, electricity users, specifically heavy users such as industry, are incentivized to shift their demand. When there is a deficiency, load falls to accommodate the shortfall. Dynamic retail pricing can thus help rationalize the market by giving price signals to both the demand and supply. [4]

The rolling blackouts this past weekend were a form of demand response, just a very costly example. In essence, the market failed to clear, and the system operator had to/choose to (forcefully) remove demand from the grid. Barring an increase in demand response capacity, and in the presence of a grid increasingly powered by renewables, we should expect this to happen again unless energy storage capacity increases (which allows utilities to shift the supply capacity of renewable generating assets through time). 

This is transition risk. Yes, putting more solar on the grid is a component of an energy transition. Still, the most significant aspect of the transition is realizing the nature by which consumers interact with their utility (and the market) will completely change. The concept of value on the grid needs to become bi-directional; generators get paid to provide energy, and consumer's use of electricity, or lack thereof, will also be priced or incentivized. The lines become even more blurred as traditional consumers become generators with either excess solar or storage they choose to offer to the market. 

Supply constrained? Not necessarily

Confusion around what triggered the blackouts raised enough noise that the New York Times published a piece on Sunday outlining the skepticism. The article notes that blackouts were initiated when operating reserves (oversupply in the system relative to demand) was at 8.9% and that perhaps grid operators were too trigger happy to initiate blackouts.[5] The article raises the question of whether "something else is going on here" and that the "lack of supply" narrative does not seem to add up. It's possible, but there are a few things to consider. 

First, actual operating reserves for Friday dipped below 6% by 4:00 pm, not 8.9%. At that point, there was still ~ 3 GW of spare capacity in the system. CAISO operating procedures call for a stage 3 emergency when operating reserves are "anticipated to be below the Contingency Reserve requirement," which are set at 3%. So, it is fair to question whether CAISO's initiation of emergency response procedures was necessary. 

Second, California has a lot of rooftop solar. This is where it gets interesting. In the absence of rooftop solar, the state could not have handled the peak electricity demand on Friday, but at sunset, when PV assets come offline, the grid experiences a load (demand) increase. Rooftop solar solves one problem and creates another. Complicating matters further, CAISO has little visibility into the state of rooftop assets. On Friday evening, all CAISO knew was that an indeterminate amount of generating capacity was going to come offline in an uncoordinated fashion. 

Imagine you are sitting in the control room trying to make decisions in real-time to balance a market. It is 109 degrees outside, and you have just seen the state's supply reserves drop from 10% that morning to 6%. A stage 2 emergency has already been declared (something you have never seen unless you have worked at the grid operator for more than 14 years), and you know that solar generation is coming offline within the next hour. 

How confident are you in the 6% reserve margin? If you initiate rolling blackouts, you guarantee that some portion of the state (we now know ~3 million people) will lose power for a period of time. If you don't, and demand exceeds the power supply capability of the network, you risk a total blackout of the power system, leaving ~ 39 million people without power for an unknown period. Depending on the configuration of the electrical network and the damage caused by the system' tripping', outages of this nature can take weeks to recover from. Furthermore, a protective fuse on a circuit breaker may fail during a forced shutdown, causing a forest fire. 

What would your decision be, in real-time, given the above scenario? 

In edge cases, grid operations often come down to human decision making, not necessarily an algorithm or procedural manual. It is easy to question the decisions made by grid operators after the fact, but real-time decision making of this nature is not a simple matter.   

Here are some outstanding questions and loosely held observations we have:

1. How much did a reduction in power imports (electricity from other states) relative to expectations impact market conditions? Data suggests that CAISO had very optimistic import availability assumptions. The hourly net imports observed on Friday and Saturday were 27% and 57% below base case import assumptions that feed into their assessment of staged emergencies for the summer. We don't have all the information, but it looks like CAISO should no longer depend on 9-11 GW of imports during days of scarcity.

2.  How much power was available, but choose not to bid into the system? By which we mean how much potential power was available in theory but decided not to participate in the market. Loretta Lynch, the former president of the California Public Utilities Commission, told a local radio station host over the weekend that she believes sellers of electricity were withholding power from the market until the price rose; an Enron 2.0. There is no evidence for this right now; however, it's worth noting that FERC allowed some of the protections in place to prevent market manipulation to expire last year. Theoretically, the conditions are more conducive for such an event. [6]

3.  What role are home energy storage systems playing in these events? It appears that due to the heatwave, Tesla home storage units went into 'storm mode' on Saturday, which means that the battery attempts to keep itself as close to 100% charged as possible to ensure full power in case of a blackout. In storm mode, instead of reducing your load profile on the grid, the battery was drawing power from the grid.  Now multiple that by the number of energy storage systems that are (or will) be installed. There is likely a simple software fix, but clearly, an unintended consequence of that may have made the problem worse, not better. 

The blackouts serve as a clear demonstration that decarbonization strategies are subject to trade-offs, and those trade-offs require more careful consideration. In power markets, that means the system needs to be optimized to reduce atmospheric emissions, subject to the degree of reliability deemed acceptable. Initiating rolling blackouts is not a feasible solution to increasing renewable energy, so demand response and storage must be scaled accordingly.

[1] This is highly simplified. Generation assets have vastly different characteristics that make the “on/off” switch a more complicated problem. 

[2] Day ahead and real time pricing in the wholesale markets can have significant pricing volatility, with prices at any given moment being 10x that of prices at other parts of the day based on the need to incentivize any and all supply to enter the market. 

[3] This is slowly changing with the introduction and rollout of time of use (TOU) rates across the country where the $/kWh charge for your electricity will change throughout the day to better reflect the cost (or value) of delivering that energy. 

[4] There are many different types of ‘demand response’ (DR) programs out there, dynamic retail pricing is simply one mechanism. 

[5] This is not an unreasonable assertion. Texas, for instance, has routinely hit operating reserve margins less than 8.9% without triggering rolling blackouts. 

[6] FERC Orders nominally set CAISO price capes at $2,000/MWh. However, as cost verification is required for all bids above $1,000/MWh, it’s very possible that price signals above $1,000/MWh have no real effect in incentivizing more supply to be made available. Absent further investigation, this would appear to be a much more plausible reason why there may have been excess capacity sitting on the sidelines.