Cloudy Incentives for Solar: How Distortions in Retail Electricity Pricing Affect the Value of Distributed Generation
A package for analyzing the distortionary impact of California's net energy metering (NEM) policies on distributed energy resource (DER) valuation.
- Use PySAM's PVWatts module to generate 8760s for each representative generation profile (by region, mounting and tracking method, azimuth and tilt).
- Use PySAM's UtilityRate module to calculate revenue from each representative generation profile under NEM 1.0 and 2.0 (based on utility service territories).
- Redo logic of assigning sites to a normalized generation profile before scaling by system capacity.
- Download 5-minute real-time LMPs for sub-LAPs in CAISO and take hourly average to compare against 8760 generation profiles.
- Redo logic of assigning sites to a grid node and associated LMPs.
- Compare switch from NEM 1.0 and 2.0 to California's new policy, NEM 3.0 (which is actually net billing).
- Incorporate distribution losses (see MIT's Utility of the Future)?
To understand how the structure of retail electricity tariffs distorts customers’ incentives to install distributed generation, we examine the case of rooftop solar in California. In particular, we first estimate the value of rooftop solar – defined as the rate that system owners receive for the electricity they produce, whether by avoiding a charge for electricity from the grid or earning compensation for generation that exceeds their instantaneous consumption. We then examine how this value differs between the first and second editions of California’s net energy metering (NEM) policy and a hypothetical tariff under which system owners are paid the locational marginal price (LMP) from the nearest grid node at the moment their generation occurs. This approach involves three main components, as described below.
LMPs reflect the wholesale cost of serving load at a specific time and location on the grid. The retail rate that owners of rooftop solar earn for generation that isn’t instantaneously consumed on-site under NEM tends to be higher than the average LMP because it not only incorporates the cost of distribution, but also costs related to maintaining network infrastructure and hedging against price volatility. Although distributed generation can help avoid distribution losses and may even postpone the need to upgrade certain transmission infrastructure, nearly all households with rooftop solar still depend on the grid to provide them with electricity when the sun isn’t shining, so they should not avoid those costs simply because they participate in NEM. Therefore, although LMPs are not a perfect measure of the value that distributed generation provides, they are a reasonable estimate in the context of this paper.
Estimate how much value existing rooftop solar installations in CA generated during 2022 under the prevailing tariffs (NEM 1.0 and 2.0).
To accomplish this, we construct a model of each of the 1.5 million solar photovoltaic (PV) systems installed in California by customers of the state’s three large investor owned utilities (IOUs): Pacific Gas & Electric, Southern California Edison, and San Diego Gas & Electric. The specifications of all systems installed under an NEM program since 1982 are publicly available via the state’s Distributed Generation Interconnection Program Database. We assign a solar resource file to each site based on its location and simulate hourly generation over the course of a year based on its system size, orientation (i.e., tilt and azimuth), and mounting method with the help of the National Renewable Energy Laboratory’s System Advisor Model (SAM).
To reduce the computational complexity of this step, we bin sites based on their characteristics and construct representative generation profiles that can be scaled to the capacity of each system matching that description. Next, we assign each site a tariff based on their electric utility and when the system was installed. For systems installed when NEM 1.0 was in effect, we assume they selected their utility’s default tariff, which is a time-invariant rate structure. For those installed when NEM 2.0 was in effect, we assume they selected their utility’s default time-of-use (TOU) tariff, which was a requirement of the program. The details of each tariff are drawn from the Utility Rate Database. By applying these tariffs to a system’s generation profile, we are able to estimate the value it creates over the course of a typical year.
Calculate how much value the same installations would have generated if their export rate accurately reflected the value of distributed generation, for which we use the locational marginal price (LMP) at the nearest grid node as a proxy.
The LMP that a system owner either avoids paying for electricity or receives for exports to the grid is set by the clearing bid in the day-ahead market at the nearest Load Aggregation Point (LAP) during the hour in question. This data is available via API request from the Open Access Same-Time Information System (OASIS) maintained by the California Independent System Operator (CAISO), which manages the state’s wholesale electricity market. While more granular prices are available, hourly data for just one node takes several minutes to download, so we settle for zonal prices that correspond to each utility’s service territory.
Explore how the value of rooftop solar differs between tariff structures and how the magnitude of the difference depends on the location of the system.
We first explored how switching all customers who were enrolled under NEM 1.0, which we represent as a flat rate, to NEM 2.0, which employs TOU rates, would affect the average value of rooftop solar. Our model suggests that a system installed in California during NEM 1.0 creates $568 of value per kilowatt (kW) of capacity per year, on average. If the same systems were transitioned to NEM 2.0, they would create just $500 of value per kW-year. In other words, adopting a rate that partially accounts for time variations in electricity prices decreases the value of solar by 12.1%. If all customers enrolled in NEM 1.0 or 2.0 were instead transitioned to a tariff that reflected the real-time LMP at the nearest grid node, the value of rooftop solar would decrease from an average of $526 per kW-year to just $103.6, or a decrease of 80.3%.
While this figure may seem striking, it is almost identical to the 75% reduction in the average rate that customers will be paid for excess generation (the “export rate”) under the successor policy to NEM 2.0 that was recently announced by the California Public Utilities Commission (CPUC). The new policy, a form of net billing, is designed to compensate owners of rooftop solar for electricity that they do not instantaneously consume on-site at “avoided cost,” which is the rate the utility would have otherwise paid to procure electricity from another source at the same time and location, instead of the retail rate. The fact that the CPUC’s measure of avoided cost matches our model’s estimate of the average value of distributed generation suggests the new policy will indeed compensate the owners of rooftop solar more accurately than its predecessors. However, the degree to which the new policy is still distortionary will ultimately depend on the level of spatial and temporal granularity at which avoided cost is calculated. For more on this question, please see the commentary in the appendix.
To illustrate when and where distributed generation is most overvalued, we plotted the difference in the average value of 1 kW of rooftop solar capacity between NEM 2.0 and a tariff based on LMPs for the three IOU service territories during each hour of the year. Where gray appears, the system owner earns the same amount per kW of capacity under each tariff. Most of the gray appears at night, since solar panels don’t generate value when the sun is down regardless of the tariff. Where the map is red, each kW of capacity is valued more highly under NEM 2.0 than it would under LMPs. The opposite is true for blue. If we believe that LMPs are a reasonable estimate of the true marginal benefit that distributed generation provides to the grid, as we argue above, then the balance of red and blue represents how much we are currently over- or undervaluing rooftop solar, respectively.
As the heat maps illustrate, we overvalue rooftop far more often than we undervalue it. There is substantial variation in this trend throughout the year, with rooftop solar being more overvalued in the winter than in the summer, presumably because LMPs tend to be lower during the winter in California, where peak system demand is driven by air conditioning usage, meaning the spread between the rate at which system owners are compensated under NEM 2.0 and LMPs is at its highest. Interestingly, looking at the heat maps for each service territory reveals that differences in how distributed generation is valued between tariffs varies more over time than it does by location, suggesting it is more important to adopt more granular TOU rates than it is to differentiate rates based on location. This makes sense because although LMPs do vary between locations on the grid, they vary much more over the course of the year (or even the day).
