After years of debate and study, California’s Self-Generation Incentive Program is on the verge of setting new rules to manage a key part of its mission: making sure that its fleet of state-funded behind-the-meter batteries are subtracting from, not adding to, the state’s greenhouse gas emissions.
This may seem like a relatively simple problem to solve from an electrochemical perspective, but that's not quite the case. Every energy storage system loses some energy from the time it’s stored to the time it’s discharged, with a typical modern lithium-ion battery system averaging about 80 to 85 percent round-trip efficiency. That puts the onus on battery operators to ensure they’re only charging with the cleanest power, and discharging to replace the dirtiest power, to make up for that lost energy. But according to multiple studies, today’s primary use cases for behind-the-meter batteries, such as demand-charge management and backup power, fail to meet those needs.
The California Public Utilities Commission has been studying this problem since 2015. Evaluations conducted for the CPUC by Itron in 2016 and 2017 have indicated that today’s Self-Generation Incentive Program-funded fleet has actually slightly increased greenhouse gas (GHG) emissions compared to if they hadn’t been added to the grid, largely because they’re not being used in ways that maximize their consumption of renewable energy. After an initial CPUC staff proposal released last year to fix this issue was widely panned by utilities and battery vendors alike, the CPUC came back with a revised staff proposal in December (PDF), one that is seen as more likely to win commission approval later this year.
But as this Self-Generation Incentive Program (SGIP) debate has shown, making sure that batteries reduce carbon emissions remains a devilishly complicated problem to solve from a policy perspective. First, as we’ve noted in past coverage of the SGIP issue, today’s economic incentives are misaligned with carbon reduction. It's a problem the CPUC’s new proposal seeks to remedy through penalties for not hitting carbon dioxide reduction thresholds — a solution that’s drawn opposition from energy storage advocates, but which is supported by utilities and ratepayer advocates.
Second, in order to optimize batteries to meet these new carbon-reduction thresholds, SGIP projects will need some way to pull the data on the marginal GHG impact of every kilowatt-hour of energy stored up and discharged, at various points on the grid, in sub-hourly increments. This concept, dubbed a “GHG signal,” is the centerpiece of the new SGIP regime. And while the CPUC staff proposal doesn’t make it explicit, it’s largely the product of one nonprofit software developer that’s been working on this challenge for the past five years.
The background on SGIP’s carbon-reducing mission
The SGIP was launched in 2006 to support peak load reduction by funding solar, biomass generation, and other on-site power. But in the past decade it has become a primary source of funding for behind-the-meter batteries, with companies including Tesla, Stem, Green Charge Networks, Sunverge and others taking advantage of its incentives for systems under 30 kilowatts in size.
In 2017, the CPUC approved an SGIP plan that would dedicate 75 percent of SGIP funding for energy storage, and last year the California legislature extended the program to spend about $830 million through 2023.
SGIP has also faced its share of controversies, such as the flaws in its first-come, first-served online submission process that allowed some companies to game the system, as well as its funding of fuel cells from Bloom Energy, despite concerns that the natural-gas-fueled generators weren’t reducing fossil fuel consumption. But for the past three years, attempts to come up with a new system to replace SGIP’s current round-trip efficiency proxy for carbon emissions — a makeshift method disliked by utilities and battery vendors alike — have dominated the program’s development.
Last year, CPUC staff proposed a fix that would force SGIP projects to yield carbon dioxide reductions equivalent to 25 kilograms per kilowatt-hour of energy stored, or face reductions in their inventive payments as a result. This drew the ire of battery vendors that pointed out it would punish systems even if they’re reducing GHG emissions, but not enough to hit that 25-kilogram threshold. SGIP participants also complained that they wouldn’t be able to hit the target without forcing their systems to operate uneconomically — arguments that CPUC staff confirmed through software modeling.
To correct for that, in the revised proposal now under consideration, CPUC has reduced its SGIP carbon dioxide reduction requirement to at least 5 kilograms per kilowatt-hour. But that’s still too much, according to most of the energy storage industry groups involved.
The Center for Sustainable Energy filed comments asking for the threshold to be “set to zero,” writing that “penalizing systems that do not reach an arbitrarily set threshold, but are nonetheless reducing GHG emissions, will negatively impact program participation and discourage good-faith efforts to legitimately reduce GHG emissions.” Tesla agreed, writing in its reply comments that “thresholds are inherently arbitrary and appear to be unnecessary to fulfill the statutory intent of the program.”
As for the punishment for not meeting those targets, the CPUC staff proposes a penalty of $1 per kilogram, or $1,000 per ton, of carbon dioxide emitted beyond that threshold. That means projects emitting 15 kilograms per kilowatt-hour would have their annual payments reduced by about 57 percent, and projects emitting 30 kilograms per kilowatt-hour would have their annual payments completely erased. This figure has also come under fire from storage advocates, who have said it appears arbitrarily to be set at a much higher cost than other real-world values for carbon, or even societal cost valuations, such as the $150 per ton referenced in the CPUC’s Integrated Resource Planning proceeding.
But according to the CPUC staff proposal, this combination of thresholds and penalties is the best way to “ensure GHG reductions and encourage future projects to find the combination of rate, operational algorithm, customer load profile, system efficiency, and other factors needed to reduce GHGs.” In fact, according to its computer modeling, it’s better than setting a target of zero carbon emissions, as many storage advocates have proposed.
CPUC reached this conclusion by using the Open-Source Energy Storage Model, software developed by the CPUC’s SGIP working group to simulate “optimal dispatch” of batteries to maximize economic returns, reduce carbon emissions, or a combination of the two. Using “commercial model runs” simulating the SGIP fleet, the CPUC staff was able to show that only about 5 percent of projects were able to meet the 25 kilograms per kilowatt-hour threshold, even with the highest “carbon-adder” plugged into the model. But 50 percent of runs were able to meet the 5 kilograms per kilowatt-hour threshold “without significantly impacting bill savings” for the customers hosting the batteries in question.
What’s more, setting the threshold to 5 kilograms per kilowatt-hour yielded no results that led to an increase in greenhouse gas emissions, as well as an average reduction of about 8.1 kilograms per kilowatt-hour of energy stored and discharged.
Meanwhile, according to the Open-Source Energy Storage Model simulation, setting the threshold to zero kilograms per kilowatt-hour, as energy storage advocates have proposed, ended up leading to a net increase in carbon emissions in more than half the model runs. Results show an average addition of 4 kilograms per kilowatt-hour into the atmosphere in this scenario, as a larger number of systems choose economic optimization over carbon reduction.
The GHG signal and WattTime: The marginal carbon intensity calculator for California’s grid
All of these modeled and real-world impacts and benefits can only be achieved if SGIP battery system operators can match their charging and discharging decisions to some kind of real-world measurements of their carbon emissions impacts. This kind of data has not been available throughout most of the SGIP’s history. But over the past year and a half, the CPUC and stakeholders in the SGIP proceeding have been working on a system to deliver those measurements, called a GHG signal.
CPUC’s new proposal defines the GHG signal as a “digitally accessible, real-time, marginal GHG emissions factor” for every pricing zone operated by California grid operator CAISO, in 5-minute and 15-minute intervals. In simple terms, it’s a measure of the carbon intensity of the grid’s generation mix in real time, which in California means calculating the emissions profile of the natural-gas-fired power plants — or possibly, the solar or wind power — providing that period’s marginal energy needs.
For planning purposes, the GHG signal will also include a series of forecasts, including 15-minute-ahead forecasts that are updated constantly, 72-hour-ahead forecasts that are updated hourly, month-ahead forecasts updated daily, and year-ahead forecasts updated monthly.
The key technology player in developing this GHG signal has been WattTime, a nonprofit founded in 2014 by two UC-Berkeley Ph.D.s working on technology to combine historical and real-time power plant emissions data to yield marginal values for various parts of the grid. The initiative started as an app for homeowners who wanted to improve their energy habits. It has since grown to delivering its data to university energy management systems, smart EV chargers, smart thermostats, and Department of Energy-funded smart grid research projects, as well as being more broadly available via APIs for other initiatives that align with the nonprofit’s mission. In 2017, the Rocky Mountain Institute absorbed WattTime as its subsidiary, expanding the nonprofit's universe of potential partners.
“I jokingly call it mission-sourced,” instead of open-sourced, software, is how co-founder Gavin McCormick described WattTime's technology in an interview this week. “There are some companies we won’t share our data with for mission reasons,” such as oil companies that might want to use it to sell more oil.
But beyond those restraints, its data and software are widely available for use by public and private parties — including the CPUC, the state’s investor-owned utilities, and SGIP vendors, which have been working closely with WattTime as part of the GHG Signal Working Group.
The CPUC’s latest staff proposal doesn’t explicitly name WattTime as the provider for companies seeking to comply with its new rules, as its original proposal did. But it does note that companies “may wish to contract with WattTime,” given its role in developing the GHG signal, and noted that WattTime “would be able to deploy a GHG signal four to six months after being commissioned.”
As McCormick put it, “It’s not a done deal that WattTime would be the provider of this technology” if the CPUC ends up approving the current staff proposal. “I want to be very careful about that.”
Even so, McCormick said, “We can calculate the carbon emission impact of charging and discharging a battery right now, as well as provide forecasts” across CAISO territory. “Having been in the field for five years, I am doubtful that anyone else is capable of providing that at this point.”
In fact, it was to solve problems like California’s SGIP carbon emission imbalance that WattTime was founded, he said. “It wasn’t that battery companies were getting this wrong,” he said of the data showing a slight emissions increase from SGIP-funded projects. “There was no data set for them to check.”
How WattTime works
WattTime’s platform pulls data from two key sources. The first is the Continuous Emission Monitoring System that all power plants maintain under U.S. Environmental Protection Agency rules, which provides emissions data for individual power plants under various operating conditions. The second is real-time data from the Open Access Same-Time Information System, the federally mandated platform that allows data exchange between regional grid operators like CAISO. (Here's a link to WattTime's map that tracks average marginal emissions in grid operator territories where it has available data.)
California’s emissions profile can swing significantly from hour to hour and from region to region, said McCormick. “It’s not just on-peak versus off-peak. Every single plant has a different level of efficiency, and that’s going up and down all day long.” And of course, during the increasingly common times when the state is generating more renewable energy than it can use, and would otherwise be forced to curtail solar or wind farms to maintain the balance between supply and demand, “if you switch on a light or charge a battery, there’s no carbon footprint at all," he said.
In terms of providing some guidance to SGIP participants worried about hitting the proposed 5 kilograms per kilowatt-hour target, McCormick noted that it boils down to better managing when their batteries charge up. Most energy storage systems designed to meet certain business needs, like demand-charge management or backup power, don’t have much flexibility in terms of when they’re asked to discharge their energy, he noted. But they tend to have much more latitude in deciding when to charge — and under today’s economic structure, they’ll logically choose the times when grid power is cheapest to do so.
Add a GHG signal, and these batteries will finally have data to tell them when the cleanest grid power is available for recharging as well, he said. Of course, the times of the cleanest grid power may not match the times when it’s cheapest, he noted. But battery vendors have optimization software that can compare the costs and benefits of these kinds of tradeoffs. “I can put that into my software and treat the carbon savings like money,” he said. If the operator decides to optimize for hitting the emissions target over maximizing revenues, “it will pick a time that’s slightly different, and you’ll have slightly less profit.”
What’s important, however, is that the GHG signal provides these SGIP vendors some kind of certainty about how they’ll manage the costs of GHG emissions compliance for years to come. Some battery companies have been voluntarily using WattTime's software to achieve their own, or their clients,’ desire for systems that didn’t increase emissions, McCormick noted.
While the economic modeling does indicate that the current 5 kilograms per kilowatt-hour threshold “may be a little bit aggressive” in terms of the economic tradeoffs it will ask of SGIP-funded battery projects, he said, most of the parties involved in the proceeding were assuming there would be some kind of tradeoff between profitability and emissions reduction.
WattTime has been investing in extending the capabilities of its software, both to ease integration for utilities, demand response aggregators, battery fleet operators and other grid-scale players, and to extend its real-time emissions data to internet-of-things-enabled devices via its cloud platform. The company calls this its Automated Emissions Reduction platform, and California’s SGIP could be a “game-changer” in terms of providing a large-scale test of its capabilities.
“A giant question mark for me is, is this a one-off, or is this a sign of things to come for the entire industry?” McCormick said. The energy storage industry has been wary of acknowledging the growing body of research showing that it could be contributing to, rather than reducing, the grid’s carbon intensity, since they’ve lacked options for dealing with the problem. “But now there’s a recognition of what the science can do. It’s one thing to have a problem — it’s another thing to have a solution.”