New EV Capacity Credit Method
The global transition toward sustainable energy systems is accelerating, driven by urgent climate goals and the rapid electrification of transport. However, this shift introduces complex challenges for power grid stability, particularly as intermittent renewable energy sources replace traditional fossil fuel generation. Ensuring resource adequacy—the ability of the power system to meet demand at all times—has become a critical priority for grid operators and policymakers worldwide. In this context, electric vehicles are increasingly viewed not merely as transportation tools but as vital distributed energy resources capable of supporting grid reliability. A groundbreaking study from Tsinghua University has introduced a sophisticated new framework for evaluating the capacity credit of generic energy storage, including electric vehicles, offering a more accurate and credible assessment method for future capacity markets.
The research, conducted by Qi Ning, Cheng Lin, and Liu Feng from the Department of Electrical Engineering at Tsinghua University, addresses a significant gap in how grid planners value flexible resources. As high proportions of renewable energy integrate into the grid, the variability of wind and solar power can lead to severe capacity deficiencies. To mitigate this, many regions, including the United Kingdom, Ireland, and parts of the United States, have established capacity markets. These markets recruit resources to guarantee system reliability. While traditional generators were the primary participants, energy storage and demand response resources are now essential contenders. These resources, often termed generic energy storage, possess energy throughput capabilities similar to conventional batteries. Accurately assessing their capacity, or the amount of conventional generation they can reliably replace, is paramount for the safe operation of new power systems and the stable functioning of capacity markets.
Historically, capacity credit evaluation focused on renewable generation, assessing their ability to replace traditional plants at equivalent reliability levels. As storage and demand response technologies matured, evaluation methods evolved. Early approaches often utilized derating methods, where storage capacity was reduced by a fixed percentage based on discharge duration. However, research has shown that such methods lack general applicability because storage credibility depends on multiple factors, including power capacity, energy capacity, load coefficients, and efficiency. More advanced techniques employ Monte Carlo simulation to calculate reliability levels under different configurations, using indicators like Effective Load Carrying Capacity or Equivalent Firm Capacity. While these simulation methods are more accurate, they face significant hurdles in handling optimization scheduling oriented toward reliability.
The core difficulty lies in how the resources are scheduled and modeled during the simulation. Existing studies have predominantly relied on fixed or greedy scheduling methods. Fixed scheduling determines the operating curve of the storage resource before the simulation begins. For instance, a fixed peak-shaving strategy might be applied to reduce peak load. While this simplifies the assessment, it fails to account for real-time system faults, leading to an underestimation of the resource’s capacity. Conversely, greedy scheduling assumes the storage is fully charged during normal states and discharges maximally during capacity shortages. This approach assumes the resource can participate in the capacity market at full capacity, ignoring the resource’s own energy management needs and uncertainties, thereby overestimating its value. Neither method provides a realistic picture of how these assets behave in a real-world market environment where they must balance self-interest with grid support.
The Tsinghua team identified that a major limitation in existing research is the modeling of generic energy storage. Previous models often treated storage as idealized and deterministic, assuming rated capacity equals available capacity in the market. In reality, the available capacity of generic energy storage is time-varying and uncertain. This uncertainty stems from baseline energy behavior; for example, a battery might need to smooth renewable fluctuations, or a thermostatically controlled load must maintain user comfort. Furthermore, market factors, user behavior, and equipment reliability introduce multiple layers of uncertainty. While some studies have considered forced outage rates, few have accounted for capacity degradation or the dynamic nature of user response.
A key innovation in this study is the distinction between two types of uncertainty: Decision-Independent Uncertainty and Decision-Dependent Uncertainty. Most prior research modeled uncertainty as decision-independent, meaning the randomness does not couple with system decisions. However, in reality, many stochastic processes are influenced by scheduling strategies and price factors. This manifests as a dynamic, endogenous uncertainty known as Decision-Dependent Uncertainty. For example, the response capacity of a virtual energy storage resource, such as an electric vehicle fleet, can change based on the incentive prices offered by the grid operator or the dispatch strategy employed. Ignoring this dependency can lead to significant errors in reliability assessment, potentially ing the true reliability level of the system.
To address these challenges, the researchers proposed a unified generic energy storage model that achieves homogeneous representation of battery storage, thermostatically controlled loads, and electric vehicles. This model incorporates both decision-independent uncertainty regarding operation status and baseline consumption, and decision-dependent uncertainty regarding available energy capacity. By treating these diverse resources under a single modeling framework, the study allows for a more aggregated and coordinated evaluation, which is essential for virtual power plants and aggregators managing flexibility resources.
The proposed model extends the traditional battery storage framework to include virtual storage characteristics. It constrains charging and discharging power, state of charge, and ensures the sustainability of schedulable energy on a daily basis. For electric vehicles, the state of charge represents the battery’s energy level, while for thermostatically controlled loads, it represents the relative indoor temperature. Crucially, the model accounts for time-varying parameters. Unlike actual storage where parameters are often fixed based on device specifications, virtual storage parameters like power boundaries and state of charge boundaries are time-varying due to baseline energy behavior and user preferences. The model also introduces correction terms related to baseline energy use, acknowledging that virtual storage must meet its primary function, such as keeping a home cool or ensuring an EV has enough range for the next trip, before it can support the grid.
The study further details how available capacity is modeled under decision-dependent uncertainty. Physically, available capacity might seem fixed, the rated capacity. However, due to baseline energy behavior, the available capacity is often less than the rated capacity and exhibits time-varying characteristics. Moreover, market incentives and dispatch strategies alter the probability distribution of this capacity. This is described as a trade-off between expected revenue effects from capacity market incentives and comfort loss effects from demand response discomfort. The researchers used available capacity boundaries to describe this phenomenon, where the boundaries expand or contract based on whether the revenue effect outweighs the comfort loss effect. This mathematical formulation captures the dynamic nature of user willingness to participate, which is critical for electric vehicles where driver comfort and range anxiety are significant factors.
Building on this modeling foundation, the team developed a capacity credit evaluation method that includes sequential coordinated dispatch and response unavailability assessment. The sequential coordinated dispatch is designed to achieve a trade-off between day-ahead self-energy management and real-time adjustment to system capacity deficiency. In normal system states, where capacity is sufficient, the generic energy storage follows a day-ahead strategy, such as peak-valley arbitrage in the energy market. This ensures the resource owner maintains economic viability. However, when the system enters an emergency state with capacity shortage, the storage discharges to reduce load loss. When the system recovers, the storage recharges to restore its capacity, continuing its day-ahead strategy. This approach mimics real-world behavior much more closely than fixed or greedy methods, as it respects the resource’s need to manage its own energy while remaining available for grid support.
The evaluation process also incorporates a response unavailability assessment to enhance credibility. Even with high confidence levels in scheduling, there remains a risk that the generic energy storage may fail to respond due to the inherent uncertainties modeled. The method assesses this risk by sampling the true values of the state of charge boundaries under decision-dependent uncertainty and correcting the dispatch instructions accordingly. If an instruction exceeds the actual available boundary, it is corrected to the boundary value. This allows for the calculation of actual reliability indicators versus theoretical ones. The results show that system reliability levels become decision-dependent, exhibiting typical operational reliability characteristics. This step is crucial for preventing the overestimation of capacity credit, ensuring that the values assigned in capacity markets are trustworthy and do not compromise system security.
The effectiveness of this new method was verified using the IEEE RTS-79 test system combined with historical operation data. The case study compared the proposed sequential coordinated dispatch against fixed and greedy scheduling methods. The results highlighted significant differences in how storage operation strategies affect system reliability. Under fixed scheduling, storage followed a rigid charge-discharge pattern that could not respond to real-time generator faults, resulting in the lowest theoretical reliability. Greedy scheduling maintained the storage in a full state to respond to any issue, yielding the highest theoretical reliability but ignoring baseline energy needs and uncertainties. The proposed sequential coordinated dispatch achieved a middle ground in reliability results but offered the highest credibility. It supported all system capacity deficiency events while recharging after load loss states ended, balancing self-management with grid support.
An analysis of actual versus theoretical reliability revealed that fixed and greedy methods suffer from significant gaps due to ignored factors like storage faults, self-discharge losses, and capacity degradation. Self-discharge losses alone contributed significantly to the reduction in actual reliability. As storage rated power increased, the gap between actual and theoretical reliability widened for the traditional methods. In contrast, under the proposed sequential coordinated dispatch, the actual and theoretical reliability remained consistent, validating the high credibility of the new method. This finding is particularly relevant for automakers and grid operators planning large-scale EV integration, as it suggests that simplistic valuation models could lead to either undervaluing asset potential or overpromising grid support capabilities.
Economic implications were also a key part of the analysis. Storage revenue comes from day-ahead energy market arbitrage and real-time capacity market support. The study assumed a specific price for reliability improvement. Results showed that both greedy and sequential scheduling captured significant profits from the capacity market by actively responding to real-time faults. Fixed scheduling, targeting only peak loads, generated minimal profit, accounting for only about ten percent of the other methods. Furthermore, because the sequential method also captures day-ahead market profits, it demonstrated higher overall economics than the greedy approach, with the advantage growing as capacity increased. This suggests that a balanced strategy is not only more reliable but also more profitable for asset owners, including EV fleet operators.
The study also examined the health of the storage assets, defined as the State of Health. While the proposed method showed a faster decline in State of Health compared to existing methods due to frequent switching between real-time and day-ahead strategies, this trade-off reduces idle capacity and increases overall utilization. For electric vehicles, this highlights the need for battery management systems that can handle frequent cycling without excessive degradation, or for market mechanisms that compensate for battery wear.
Regarding capacity credit indicators, the study evaluated four types: Equivalent Firm Capacity, Equivalent Conventional Capacity, Equivalent Generation Capacity Substituted, and Effective Load Carrying Capacity. While numerical values differed, all indicators showed a trend where capacity credit decreased as power capacity increased but increased significantly as energy capacity improved. This indicates that for systems with high storage power penetration, building new storage to improve reliability may not be economical. However, long-duration storage offers stronger reliability support. This is a critical insight for the automotive industry, suggesting that EVs with larger battery packs capable of longer discharge durations may hold significantly higher value in capacity markets than those with smaller batteries, potentially influencing future vehicle design and marketing strategies.
The research also explored core influencing factors such as renewable energy penetration, charging efficiency, storage reliability, access location, load levels, confidence levels, and the level of decision-dependent uncertainty. In scenarios with low renewable penetration, storage capacity credit varied widely, but in high penetration scenarios, long-duration storage maintained high capacity credit, reaching up to sixty percent. This underscores the indispensable role of long-duration storage in new power systems. Charging efficiency had a minimal impact on short-duration storage but a significant impact on long-duration storage, indicating that improving efficiency for long-discharge applications is crucial. Storage reliability parameters, such as forced outage rates, had a greater impact on long-duration storage than short-duration storage.
Location also mattered. Distributed access of storage at load nodes improved capacity credit compared to centralized access at renewable generation nodes, as it avoided network transmission losses. This supports the case for decentralized EV charging infrastructure that can support local grid needs. Virtual storage, such as electric vehicles and thermostatically controlled loads, generally showed lower capacity credit than actual battery storage, heavily influenced by load levels and correlation levels. Improving these factors could significantly enhance their value. Additionally, the level of decision-dependent uncertainty played a major role. Reducing user discomfort levels had a more significant positive impact on capacity credit than increasing incentive prices. This suggests that recruiting resources with lower discomfort characteristics is more beneficial than simply ing more funds into incentives, a finding that could shape how aggregators design user engagement programs for EV owners.
The authors acknowledge certain limitations. In a competitive market environment, storage entities may adopt optimized capacity allocation strategies to participate in different electricity markets to maximize profit, which could affect the assessment. Future work will continue to explore capacity credit evaluation under complex electricity market environments. Despite this, the proposed method provides a robust foundation for future capacity market design.
This research represents a significant step forward in integrating electric vehicles and other flexible resources into the power grid. By providing a more accurate and credible method for assessing capacity value, it helps mitigate the risk of theoretical assessment results failing in actual capacity markets. For the automotive sector, this means a clearer pathway to monetizing vehicle batteries without compromising grid stability. It validates the concept of vehicles as grid assets but adds the necessary nuance regarding user behavior and uncertainty that previous models lacked. As the industry moves toward vehicle-to-grid technologies, frameworks like the one developed by the Tsinghua team will be essential for establishing fair compensation mechanisms and ensuring that the promise of electric vehicles as stabilizers for the renewable energy grid can be reliably fulfilled.
The study concludes that uncertainty modeling and scheduling strategies are critical factors influencing capacity credit assessment. The proposed generic energy storage model successfully achieves homogeneous representation of diverse resources. The sequential coordinated dispatch method balances baseline energy behavior with system adequacy support tasks, providing a credible reference for capacity value. While actual storage is influenced by efficiency and capacity, virtual storage is heavily impacted by load levels and uncertainty characteristics. Both see reduced capacity credit with increased power capacity but increased credit with energy capacity. Long-duration storage holds higher value in high renewable systems. These insights provide a comprehensive guide for policymakers, grid operators, and automotive manufacturers navigating the complex intersection of transportation and energy infrastructure.
Authors: Qi Ning, Cheng Lin, Liu Feng Affiliation: Department of Electrical Engineering, Tsinghua University; State Key Laboratory of Power System Operation and Control (Department of Electrical Engineering, Tsinghua University) Journal: Power System Technology DOI: 10.13335/j.1000-3673.pst.2023.1328