What happens to electricity prices when variable resources (VREs) like solar and wind become the grid’s dominant source of energy? Energy modelers have ideas, but no clear answers. Wind and solar are relatively unique compared to nuclear, gas, and coal systems because, despite having some upfront capital costs, they do not have fuel costs. This means that VREs, when they are generating, have a marginal price of electricity at or near zero. This is a problem: orthodox economic theory suggests that scarce resources are rationed on the basis of the marginal costs of producing them. In fact, we believe this so much that we have designed markets for electricity in the same manner, where energy producers bid into the market for electricity at some set period of time at a price that corresponds to their marginal cost. This is called the “merit order” system, where the price of the electricity sold to consumers at any given time is determined by the highest-marginal-cost power source needed to meet demand at that time. 

Let’s say we have four bids, each bid worth one megawatt-hour (MWH), two of which are one dollar per MWH, one of which is three dollars per MWH, and the last of which Is four dollars per MWH, and let’s say we have 3 MWH of system demand, enough to call on three out of the four bids. The three-dollar-per-MWH bid therefore sets the spot price of all electricity for the period; the two generators which bid at one dollar will each make three dollars and pocket a bigger spread than the three dollar bidder, the top marginal bidder, which earns revenue at cost. In theory, this merit order mechanism is designed to reward the most cost-efficient producers.

But electricity markets are weird, and VREs make them weirder. Up until now, it has been very hard to store electrons. Storage of some kind is critical to commodity markets, and to marginal pricing as an economic phenomena, because storage enables the buildup or discharge of an “inventory” if supply and demand experience unpredictable mismatches that existing agents can absorb onto their own inventories.  Alfred Marshall, one of the founders of “marginalism,” pointed out, marginal pricing is only efficient when the auction involves well-informed dealers. Dealers are unique agents in auctions because they are two-sided entities who earn a spread from buying and selling a commodity at different prices. A dealer’s goal is to off-load extra supply from those who have it and sell to those who need it. This means they rely on some kind of implied storage mechanism and pay their provider of storage some “carrying cost” to hold the commodity between purchase and sale. In other words, the dealers must finance their inventory to make efficient marginal pricing possible. 

Until recently, electricity markets didn’t have storage. That has not been a major issue:  electrons move at the speed of light, meaning that supply and demand can settle very quickly even absent reserves. To be sure, any mismatches between demand and supply, however temporary, could quite literally endanger the physical functioning of the grid itself. But traditional fossil-fuel based generation used other methods in lieu of storage to modulate their power output, such as  increasing or decreasing the output of steam turbines.. They also utilized storage substitutes like load sinks, spinning reserves, and–in the last resort–blackouts and brownouts. Thus, in this fossil-dominant system,merit order works to create a marginal pricing system that keeps the grid functioning, if not always effectively.

Enter VREs. Without fuel, they will always be first in the merit order and undercut other resources in price when the sun shines, wind blows, or water flows. So much so that, at a certain level of penetration, they will lower the clearing marginal price (the marginal price of the last unit of power required to meet demand at any one time) below the point that allows each individual project to recover its upfront investment costs. 

However, VREs are, by definition, variable, which means that they are not always available—when the sun sets, winds stall, or streams trickle. In those moments, prices will do the opposite—spike!—because any resources that can fill the gap must make all of their money at those very few moments where they aren’t being competed out of existence by the nearly zero marginal cost of VREs. Renewables, by virtue of their variability, end up compressing the period over which fossil resources that remain on the grid can recoup their capital costs. 

Batteries help tackle this problem: they introduce storage to the market for electrons. In market terms, battery suppliers act as dealers, insofar as they buy electricity when it’s generated cheaply and keep it until they can sell it later, when it’s worth more to grid participants, and earn a spread in the process. However, they too might encounter a situation where the very low cost of electricity in periods of abundance does not allow them to recoup their initial investment even when electricity demand is high. This risk is especially relevant for “marginal” batteries that either lack consistent charge- or discharge- schedules or bid into the system for the increment of load that fluctuates most from day-to-day or season-to-season.

To solve this problem, some authors have suggested that we begin to introduce payments for producers based not just on the price of electricity but also on the capacity that they provide. In other words, we do not just allow a source of electricity to make money when it is in the merit order, but also allow it to receive payments for offering spare capacity to make up for gaps in VRE availability. This kind of market structure exists, but has its own issues. For example, it is very difficult to separate out the difference between spot and capacity uses of a resource. A series of papers by Ramteen Sioshansi has argued that, in practice, one cannot predictably tell between the best use of a battery’s storage capacity between providing for the resilience sought by capacity markets and arbitraging in electricity spot markets (offering one’s resource in the merit order). Some batteries will be able to operate as purely generation resources—always charging and discharging at predictable and consistent times. At a system level, the restraints on batteries that cannot operate consistently will also limit storage’s ability to balance a system without some kind of capacity subsidy (a means of income not based on generation discharge). 

This means that, without the right set of institutional ownership arrangements, batteries will either be over-built or under-built—raising real concerns for system reliability and consumer costs. There is already some evidence of this problem in the real world, something we’ve actually written about in this newsletter: Texas’ ERCOT created a capacity payment scheme which allowed gas turbines to receive payments for standing idle if they could rapidly spin up to provide “peaker” electricity when there were gaps in VRE generation. The scheme took the most efficient gas turbines out of the merit order, resulting in higher spot electricity prices—all while those idling gas plants made a killing in capacity payments. Turning batteries into a grid resource via capacity payments would present ratepayers with a similar problem: the infrastructure costs would have to be allocated and paid for. 

However, a new paper by Tom Brown, Fabian Neumanna, and Iegor Riepina of the Technische Universitaet Berlin (TUB) recently argued that a VRE-dominated system with long-term storage could work without price spikes and disruptions because electricity consumers have more demand elasticity than we have previously assumed. Using historical data on weather and electricity consumption, they show that wholesale consumers such as utilities adjust their demand, even with very limited foresight. The 5% elasticity that they find in German data is enough to compensate battery operators without the need for capacity payments. 

A feature of this paper is the presence of long duration storage enabled by either hydrogen or district heating. Long-duration storage works because its depth and duration might allow it to absorb a much higher percentage of daily or seasonal fluctuations in VRE output and thus optimize its charging and discharging decisions efficiently in response to a flexible demand curve—i.e., a system with longer duration elements provides us with a class of dealers who can be much more flexible in how to manage their inventory. Longer duration is like providing better refrigeration to a milk market: if you aren’t as worried about spoilage, you can be more flexible in your inventory management. This feature, paired with strong empirical data on a much more elastic demand than assumed by previous studies allow the authors to question whether the problem of pricing under VREs won’t solve itself? 

We might get lucky pairing long and short duration storage to match exactly what we need—but our experiences with short-duration storage, increasing VRE penetration, and fossil fuel-dominant systems has not produced encouraging results. Moving toward the kind of system envisioned by the authors of the paper takes capital expenditures to not just install new capacities – some of which are still maturing technologies like long duration storage – but retire old ones. This is a piecemeal process that happens in historical time. Moreover, it is a process which will ideally happen as quickly as possible to help prevent the worst outcomes of the climate crisis. There will always be some increment of storage capacity that cannot perfectly match even the most elastic of demand since there are other sources and uses coming online at heterogeneous rates. And that is before we consider how much the system should keep in reserve.

In these conditions, there will have to be some tool that insures against uncertainty about the final shape of the system as a whole. Capacity payments paid for by ratepayers are one such instrument. However, they are not the only tool. The state has long served this backstop role in a variety of markets. In this case, however, the state shouldn’t limit itself to one option for incentivizing and supplementing storage in a VRE dominant grid. It should build clean firm generation like geothermal and nuclear. It should build additional VRE resources. It should build load sinks that can take on excess power. It should also build additional storage capacity. The role of public options in a transitional energy system is to provide several paths to enable grid balancing as the shape of the grid changes by allowing the public sector to do the job it does best – shouldering the cost of uncertainty.  In the absence of long duration storage, increasingly variable electricity systems are turning to gas as the final, marginal, resource to balance the grid and let the merit order operate. If storage is going to displace gas and anchor our grid in the coming years, some institution must provide the overage—large or small—of required reserve capacity. This is a job for the state. Good policy allows the state to mediate between the ability of ratepayers to shoulder the mismatches and costs that come with financing complex capital investment in clean energy and the imperative to decarbonize the grid. In doing so, the state can facilitate the increased capital expenditure,balance the grid, and limit the price shocks from energy shortages–stabilizing prices and enabling effective consumer protection. Crucially, the state has played this systemic protection role in electricity before; it is time to transform that role towards cleaner technologies, stabilization of consumer costs, and a firm commitment to reliability.