The steady and rapid growth of renewables in the U.S. has caused many states to be well ahead of their renewable portfolio standard (RPS) targets. As a result, there is increasing interest in developing renewables as merchant assets not secured under a long-term power purchase agreement (PPA) for their energy or green attributes.
In this article, we demonstrate how renewable generators could use standard electric and gas forward contracts to manage their market risk over mid- to long-term horizons when a utility PPA is not available. The electricity forwards demonstrate how wind volume uncertainty and its correlation with spot prices influence the hedge, while the gas forwards allow more, albeit imperfect, hedging of longer-term risks. Although this article focuses on hedging methods for wind assets in the Electric Reliability Council of Texas (ERCOT) region, many of the analytical concepts also apply to solar energy, and the framework can be used in other regions, subject to adjustments for features such as capacity pricing.
Using electricity forwards
The basic hedge for a wind plant would involve selling forward contracts to receive a fixed price in exchange for a “floating” price set based on average spot price at the settlement point. The deviations in spot revenues would be offset to some degree by the changes in floating contract payments, allowing for a less volatile revenue stream.
The purpose of selling forwards is to lock in prices, but the uncertainty in the total amount and timing of wind output creates a certain amount of irreducible volume risk for wind resources. Because the variance of the wind output is correlated with market outcomes, volume risk should be taken into account when hedge ratios are being determined. Although selling too little forward leaves a long position exposed to spot prices, selling too much forward can create a short position exposed to spot price replacement, as well. An effective hedge requires estimating the hourly output pattern for 1 MW of wind and valuing it at the expected volume-weighted average spot price of electricity in those hours of production. The resulting expected revenue would then be divided by the fixed cost of a standard forward contract to find the number of contracts needed.
For example, suppose that a wind plant is expected to operate at a 50% capacity factor in a given month during just off-peak hours, and its output is randomly and uniformly distributed across all hours. This output would then get the same expected price as the off-peak forward price (assume at $25/MWh), but because it will average at 0.5 MW for each megawatt of installed capacity, you need half as many hedge contract megawatts as you have wind megawatts.
If the wind output is not uniformly distributed but is concentrated in hours with lower spot prices (say, averaging $10/MWh), then you can expect only $5/MWh over all off-peak hours, so you need just 0.2 hedge megawatts per wind megawatt. Thus, the volume hedged depends on the price pattern of likely output. Similar calculations would be done for on- and off-peak hours in each forward month. Electric hedges are intrinsically imperfect for wind, with residual risks arising from the following:
- Uncertainty in wind output quantity per month;
- Hourly pattern of when that wind output quantity will be realized within the month;
- Hourly pattern of spot prices; and
- Congestion risk between the location of the electric forward price settlement and the bus that wind plant is physically connected to.
In general, these volume risks cannot be eliminated, but their expected effects can be used to revise the size of the desired hedge. These factors combine into a net uncertainty surrounding the realized wind revenues versus the hedge payoff.
For instance, if the variances in wind output are negatively correlated with the realized prices of spot power (i.e., less-than-expected wind leading to higher-than-expected spot prices and vice versa), then net gains and losses will not balance out, resulting in an expected loss from the error terms. Figure 1 demonstrates this negative correlation in a plot of the ranges of wind output and spot prices in ERCOT over the past six years versus real-time location marginal pricing. It shows that wind generators historically ran at a high capacity factor of 60% when the hub prices were below $10/MWh but ran only at 20% when prices reached $50/MWh or higher. Of course, the nature of the wind-price correlation will vary over time, but this negative relationship is fairly typical.
Because of such wind-price correlation, the expected revenues collected by a wind plant could be much lower than the expected spot price multiplied by expected wind volume. As shown in Figure 2, the wind output weighted average prices for wind generators in ERCOT have been persistently below the simple average prices over the past six years, with the difference as large as 30%-40% in some months. This means that if hedge quantities were to match expected wind output, the hedge payoff would often exceed changes in spot revenues relative to expected levels. A negative adjustment for wind-price correlation resulting in smaller hedge positions per megawatt of wind would be needed to reduce the associated exposure to spot prices.
Although there is a net “discount” for wind output from simple average prices, the magnitude of this discount varies significantly from one month to another, possibly with some seasonality. Some of this variation is random, but some is probably recurring and predictable, which ought to be taken into account as a part of hedging decisions.
Electric prices often spike due to a combination of extremely high demand and limited supply, during which wind resources tend to run less (e.g., very hot, still days). Such price spikes could magnify the effects of wind-price correlation and should be considered when hedge ratios are being determined. Although price spikes are uncertain, electric forwards typically include a “premium” to account for the effects of price spikes during occasional scarcity events.
For example, as of February, on-peak futures at ERCOT North Hub for August traded at approximately $65/MWh, which is $40/MWh above the marginal cost of generators likely to set prices at $25/MWh under normal conditions. If a wind plant is estimated to run at 40% capacity factor on average but only 10% in scarcity hours, it would collect only one-fourth of the scarcity premium (or $10/MWh) on average, resulting in expected spot revenues of $35/MWh (equal to $25/MWh in normal hours, plus $10/MWh for scarcity premium). In this case, effective hedging would be to sell forwards in the amount of $35 ÷ $65 = 53.8% of the expected average wind output, which translates to approximately 0.215 hedge megawatts per wind megawatt.
As mentioned earlier, the wind-price correlation is largely driven by aggregate wind output in a system. Individual wind plants may have different output patterns compared with aggregate wind, which means that for effective hedging, it is important to understand the correlation of wind output at specific facilities with the aggregate output.
Using natural gas swaps
Electric forwards are only liquid one to two years out, and therefore, they do not readily support long-term hedging. Gas contracts are often available and liquid for longer delivery periods (especially at Henry Hub), so they can serve as potential hedging instruments over longer horizons.
However, instead of swapping real-time spot value of wind output for an electric forward, market revenues are used to buy spot gas scaled by expected market heat rate (HR) to settle against a fixed gas forward purchase. The on-peak versus off-peak distinction in the hedging contract is lost, as gas is not differentiated by time of delivery.
The risk factors listed for electric hedging also apply here, but there are several additional hedge design elements and uncertainties to consider:
- Expected correlation of gas and electric forward prices over time;
- Uncertainty or drift in the expected long-run market implied HRs (the ratio of electric to gas prices), which cannot be observed far forward due to limited trading of electric contracts, so they will have to be forecasted; and
- Gas basis risk between settlement location of gas contracts and delivered gas for plants setting prices at wind production node.
In ERCOT, during 2011-2016, the monthly average market HRs fluctuated considerably within a range of 6-32 MMBtus/MWh. As shown in Figure 3, the bigger changes in market HR were largely driven by price spikes when there was scarcity in the system. When such price spikes are excluded, the average market HR across the remaining hours follows a relatively steady and seasonal pattern, ranging from 7 MMBtus/MWh in winter to 8-9 MMBtus/MWh in summer.
If market HRs were known and fixed in future periods, it would be simple to substitute gas contracts for electric forwards to achieve equivalent hedging with either one: In the example described previously in which the wind plant expects to have a 40% capacity factor and $35/MWh spot revenues during on-peak hours in August (below the $65/MWh forward price due to negative wind-price correlation), the effective hedging strategy would be 0.215 hedge megawatts per wind megawatt. As of February, the gas forwards for August traded at approximately $3.4/MMBtu, which implies an expected on-peak HR of 19.1 MMBtus/MWh, including anticipated effects of scarcity prices. Then the equivalent hedging strategy using gas swaps would be 0.215 × 19.1 = 4.1 hedge MMBtus per wind megawatt. Under this strategy, expected spot revenues would match fixed charges associated with gas swap contracts.
However, market HRs are variable over short and long time frames, creating conversion risk for using gas swaps as a surrogate. Figure 4 summarizes some of the ways in which they can vary that must be anticipated. If/when market HRs change, there will be a corresponding increase or decrease in the quantity of gas contracts needed to replicate an electric hedge.
Hedging decisions should also consider that errors in forecasted HR may be correlated with unexpected changes in gas prices. In coal-heavy regions, very low gas prices (as experienced in 2012) can push gas plants into being infra-marginal and cause coal plants to set prices more often. This would result in the market HR being higher because electric prices would not decrease as much as gas prices. In regions with considerable amounts of oil-fired plants, unexpected gas price spikes can cause a lower market HR because electricity prices do not rise as much as gas prices in extreme conditions (e.g., during the Polar Vortex in 2014). Such examples illustrate that gas prices and market HRs could be negatively correlated at certain price levels, which may affect the optimal hedging strategy for wind plants.
Price differences between gas at the contract hub and marginal gas resources setting electric prices at the wind production site may create an additional “basis” risk that needs to be considered as a part of the hedging strategy.
One of the attractive features of renewables is that they are an engineering hedge against carbon pricing, but getting credit for this in hedges requires being able to anticipate the effects of possible policies such as carbon pricing. The interaction of CO2 penalties or constraints with power prices, though regionally distinct in marginal sensitivity, is fairly well understood, so it is useful to think about what carbon constraints might do to wind hedging practices when/where implemented. It is important to recognize that short- and long-run effects of CO2 prices on markets can be quite different.
Initially, CO2 prices can be positively correlated with gas prices, especially in coal-heavy regions, because higher gas prices would increase CO2 prices needed for coal to gas switching.
However, over a longer horizon, this relationship can change and perhaps even reverse, as higher gas prices raise electricity prices and accelerate future renewables entry, which may reduce CO2 and electricity prices.
Overall, the relationship between gas and CO2 prices can alter appropriate hedge ratios for wind plants using gas contracts. Generally, the more valuable the future carbon avoidance when CO2 prices are also positively correlated with gas, the larger the gas forward position will have to be (because the price of gas, itself, will not reflect that carbon benefit). Such changes can be usefully anticipated via power system analysis tools and then covered with conventional or more complex hedges (e.g., gas options contingent on future market events).
Increasing amounts of renewables, especially wind, are being developed, often in excess of states’ RPS targets and increasingly without PPAs. Such merchant projects are exposed to significant market risk, as they need to absorb price fluctuations and anticipate how their own production outputs may or may not vary favorably in relation to market tightness. This risk can create a barrier for initial financing and perhaps impact the economic feasibility of projects. In response, customized, bilateral hedging arrangements are gaining popularity, but they could be offered at prices with heavy discounts that reduce the value of projects from a developer’s perspective.
Although these examples have focused on a single wind plant in ERCOT, the concepts are applicable to rebalancing the hedges in a more diversified portfolio with some new, unhedged wind. And the principles would also apply to other markets and even other renewables, such as solar plants.
Frank C. Graves and Bente Villadsen are principals and C. Onur Aydin is associate at The Brattle Group, a Cambridge, Mass.-based consultancy. They can be reached at firstname.lastname@example.org, email@example.com, and firstname.lastname@example.org, respectively.