3.
Please provide your organization’s feedback on the ISO’s “Accounting and Reporting Approach” presentation.
Resource-specific emission rates: To calculate generator-specific emissions, CAISO proposes using static emissions factors, either submitted by the resource owner, published in eGRID, published by EIA, or calculated by CAISO. While the use of static, annual-average, generator-specific emissions factors has been the status quo approach for estimating annual total emissions, our analysis has found that such approaches are not appropriate for accurate, temporally granular allocation of emissions, such as on a 5-minutely basis as proposed. This is due to the fact that generator heat rates (and thus emissions factors) change from interval to interval based on various factors, and that certain generators can fuel switch over time.
Furthermore, the emission factors published by eGRID are only published at the plant level, rather than generator level, meaning that their use may mis-allocate emissions across multiple generators located at the same plant, which may or may not be owned or contracted by the same entity. We compared hourly emissions calculated using eGRID plant-level emissions factors to measured hourly CO2 emissions data reported to EPA CAMPD using CEMS devices, and found that eGRID factors mis-estimated hourly emissions totals by 17% in CAISO, 12% in BANC, and 45% in PGE, for example. Use of generator-specific emissions factors available through public data sources such as the Open Grid Emissions dataset reduced these errors, and use of generator-specific heat rate curve models (which Singularity has calculated based on publicly-available data) proved to most accurately estimate emissions for these regions.
We would recommend using a heat rate curve model approach to most accurately allocate emissions on a 5-minutely basis, or at the very least, generator-specific static emissions factors.
Tracking and allocation of biogenic CO2 emissions: While California and Washington’s GHG pricing programs generally exclude biogenic CO2 emissions (from the combustion of biomass or biogas), consideration should be given to whether such exclusion is universally applicable for all states/LSEs. Even within California, certain biomass generators may or may not qualify for this exclusion.
While the exclusion of biogenic emissions from CAISO’s accounting and reporting approach may be justified in cases by certain state policies, the broader academic literature indicates that a blanket assumption of zero global warming potential for all biomass combustion is not appropriate given the variety of sources. Biomass combustion is one of the most carbon-intensive sources of power generation (from a direct stack emissions perspective), and research by Miller et al (2023) found that the use of biomass-adjusted emissions factors underestimates total direct CO2 emissions from power generation by 12% in CAISO, 40% in IID, 32% in Pacificorp West, and 9% in BPA.
While this may be a policy consideration outside the scope of the design of CAISO’s accounting approach, the choice of default emission factors do affect this accounting. For example, the plant-level emission factors published in eGRID treat all biomass generation as having zero CO2 emissions.
We recommend that CAISO collect more information about the reporting needs of stakeholders concerning the inclusion or exclusion of biogenic CO2 emissions, and ensure that the factors used for calculating resource emissions take into account generator-specific qualification for exclusion under any state GHG pricing policies. One option would be to separately track and report biogenic and non-biogenic CO2 emissions so that these values could be used separately or in combination to meet stakeholders’ needs flexibly.
Mapping of resources to LSEs: CAISO has noted its difficulty identifying how to map generators to LSEs based on ownership since such data is not currently collected by CAISO. However, we understand there to be multiple ways for a default mapping to be determined.
One option would be to first map each generation resource to its unique EIA plant and generator ID, and then utilize publicly available data about generator ownership in EIA Form 860 to perform a default mapping. If EIA identifiers are not already collected by CAISO, we have found through past experience that performing that mapping is straightforward if not somewhat time consuming. However, mapping each resource to its EIA IDs would be a necessary step anyway in order to identify the appropriate emissions factor from emissions databases such as eGRID, the EIA, or the Open Grid Emissions dataset.
Another option would be to utilize CAISO’s own Full Network Model mapping, which identifies the relationship between generation resources and Utility Distribution Companies (UDCs). While UDCs do not represent the full range of LSEs, it could serve as one potential starting point.
While the above approaches could be used for mapping resources based on owner, mapping based on (long-term) contracts is more challenging, as there is no public database that contains that information that we are aware of. However, each LSE’s IRP filing could provide an indication of which resources are contracted for a default mapping.
We also recommend that the system as implemented allows each LSE to check its default mapping, make changes, and have those changes confirmed by the relevant counterparty.
Mapping Jointly Owned Units and multi-offtaker PPAs: When mapping resources to LSEs, there should be an option to be able to reflect not only ownership/contract but the percentage ownership/offtake in the case of JOUS or resources with multiple offtakers. Such arrangements can vary, and be represented as a fixed share of energy, fixed MWh blocks, or other arrangements. CAISO should solicit inputs on the types of arrangements that may need to be accommodated, but a fixed % approach may be a good starting point.
LSE-specific load data: While LSE-specific load data may need to be directly collected from each LSE (or each UDC on behalf of the LSEs in their distribution territory), one potential avenue to explore is whether LSE-specific load data exists as part of CAISO’s state estimator model / full network model. If it is possible to map individual bus IDs or Cnode IDs from the network model to LSEs, then the 5-minutely load data could be pulled from the state estimator data. While this may not match metered, settled load exactly, it may provide a reasonable starting point for the allocation.
It is also possible that this data already exists somewhere in CAISO’s systems. For example, in the data that CAISO reports to the EIA Hourly Electric Grid Monitor (i.e. Survey Form 930), it reports hourly load data for PG&E, SCE, SDG&E, and Valley Electric Association. It is unclear where this data comes from and whether it represents UDC load or LSE load, but somehow CAISO already is reporting this data. See https://www.eia.gov/electricity/gridmonitor/dashboard/electric_overview/balancing_authority/CISO.
While CAISO does not have data at the LSE level, CAISO does have data for individual BAAs, and in many cases, these BAAs may overlap with LSEs (eg PGE, SMUD/BANC, LADWP, IID, etc), so it may be possible in some cases to use BAA load data for certain LSEs.
Alternatives to the “average” approach for allocating generation to the residual mix: In the case that an LSE’s supply exceeds its load in a given interval, CAISO proposes allocating the excess to the residual mix as an average share of the LSE’s total supply position. However, we are concerned about offering this as the only option available for allocating excess generation to the residual mix. This approach would in effect allow for part of an LSE’s renewable energy position used for RPS compliance, or “non-bypassable” nuclear power to be released to the residual mix to then fill other entities short positions, resulting in potential double-counting of this renewable or non-bypassable generation. As an alternative, there should be an option for an LSE’s clean energy portfolio to be allocated to its own load first, and not end up in the residual mix unless the LSE is long in clean energy or otherwise elects to do so.
To facilitate this, LSEs could elect to define a manual stack order for the resources in their portfolio that would be used to determine which resources are released to the residual mix. Such a stack order could be defined on a resource-by-resource basis, or on a fuel-type basis (eg allocate solar to my own load first).
A third alternative allocation approach that could be offered would be based on an economic stack order approach, where the resources with the cheapest offers are allocated to the LSE’s load first, and the most expensive are allocated to the residual mix.
The current formula appears to not allocate “attributed” energy: The current formula used to summarize the accounting approach subtracts attributed generators from an LSE’s supply position if that generation is attributed to a different market. However, the formula does not currently add this attributed energy to the supply mix of LSEs in areas to which it is attributed. For example, if a WA LSE owns generation that is attributed to CA, that generation is subtracted from their (and thus the state of Washington’s) supply, but currently the formula does not reflect that this attributed energy is then added to CA’s supply.
Similarly, while the formula shows contracted generation being added to an LSE’s portfolio, it does not show that contracted generation could also be subtracted (for example if an LSE owns a resource, but another LSE buys the generation from them).
I assume that attributed energy, if part of a bilateral contract, would be reflected in the “dispatched contracts for purchase” adjustment, attributed energy that is not reflected in a bilateral contract should be somehow allocated to LSEs in the area receiving the attributed energy.
There are several options for where attributed energy could show up:
- Attributed energy could be proportionally allocated to all LSEs in the receiving area after accounting for total owned/contracted. It would then be part of the determination of an LSE’s net position relative to its load, and could then potentially in part be released to the residual mix.
- Attributed energy could go directly into the residual mix. In this case, attributed energy would only be allocated to LSEs with a short position in that interval as part of the broader residual mix. (likewise, in figure 2, the attributed energy could first go into the voluntary intra-GHG LSE adjustment first, before getting released to the residual mix).
Accounting for energy storage: The proposed approach does not currently mention how standalone energy storage should be handled in allocation, even though they now serve as a major percentage of CAISO’s supply mix. Here I am referring specifically to standalone, transmission connected storage resources that support the system, not hybrid resources or behind-the-meter resources.
Rather than treating storage as a zero-carbon generator, or as load, standardization efforts around global emissions accounting standards seem to be moving toward treating storage as a transmission-style asset in that they move energy around the grid (although they move energy over time, rather than across space). What this means is that the emissions intensity of the electricity discharged by energy storage would be calculated based on the emissions intensity of the electricity used to charge the storage resource.
There are developing standards on how this would work in practice, but it would mean that CAISO’s accounting approach should include a way to allocate generation to storage charging, which is then re-allocated back out to LSE load when the battery discharges.