Impact of Battery Degradation on Market Participation of Utility-Scale Batteries: Case Studies

In the ULP, the BESS owner who has several battery storage units at different buses maximizes its revenue from the real-time energy, spinning reserve, and pay as performance frequency regulation markets while considering its operating limits, including or not including the battery degradation cost. The ULP of each framework is described below.

As shown in Fig. 1, the LLP models ISOs’ joint market clearing process, and it is similar in both frameworks. The LLP objective function minimizes the total operating cost of real-time energy, reserve, regulation capacity, and regulation mileage markets across a certain time period, considering the operating costs of the BESS and other non-battery market participants. Each cost term is calculated by multiplying each unit’s price offer to its scheduled power. The BESS’s price offers are input parameters from the ULP. It is assumed that BESS can perfectly predict price offers of other participants.

The following six groups of constraints are considered for the LLP, as shown in Fig. 1. 1) Operational limits of BESS: these constraints limit each battery unit’s scheduled power for each market below its corresponding quantity offer. 2) Operational limits of other participants: minimum and maximum generation limits of other participants are maintained in these constraints along with ramping limits. 3) Frequency regulation market model: these constraints model the impact of each unit’s historical performance in regulation services provision on its current dispatch in the regulation market. 4) System-wide reliability requirements: these constraints ensure that, during each market clearing interval, the system ancillary services requirements are satisfied by the scheduled power of BESS and other market participants. The corresponding dual variables are MCPs for reserve, regulation capacity, and regulation mileage provisions. 5) Power balance at each bus: these constraints enforce the Kirchhoff’s current law at each bus. The corresponding dual variables are locational marginal prices (LMPs) at various buses. 6) Transmission network model: these constraints calculate transmission line power flow and enforce line thermal limits.

Using the conversion procedure in [15], both bi-level frameworks are converted to mixed integer linear programming (MILP) to be solved using available commercial solvers.

To have a synthetic test system built upon real-world data, the IEEE reliability test system (RTS), updated by Grid Modernization Laboratory Consortium, i.e., RTS-GLMC, is adopted for the case studies [16]. The simulations are performed on the third area of the RTS-GMLC network. This test system consists of 25 buses, 39 transmission lines, 26 generators, and a battery unit on Bus 13. The simulation horizon is 24 hours with 15-minute market clearing intervals (i.e., 96 market clearing intervals in total). The AGC signals are dispatched every 20 seconds for the framework with the degradation cost model and AGC signal dispatch model.

System load, reserve, and regulation capacity requirements for the simulation horizon are determined by averaging the summer (June to August) load and ancillary services requirements of the third area in the RTS-GMLC. During the 24-hour simulation horizon, the system load varies between 1285 MW and 2345 MW. The system peak and valley loads happen in Hours 11~18 and Hours 1~6, respectively. Hence, it is expected that the energy market price reaches its highest and lowest values in respectively Hours 11~18 and Hours 1~6.

In each market clearing interval, the system’s regulation mileage requirement is set to be 1.5 times the regulation capacity requirement. The sample AGC set points, provided by ISO-NE, are modified for AGC signal modeling [17]. The modified AGC signals have zero-mean over each 15-minute market clearing interval. In each market clearing interval, the total variation of the modified AGC signals is equal to the corresponding system regulation mileage requirement.

It is assumed, in each market clearing interval, each generator’s price offer for the energy market is equal to the generation cost. Generator price offers for reserve, regulation capacity, and regulation mileage markets are 0.15, 0.4, and 0.07 times their energy price offers, respectively. The multipliers used for calculating ancillary services’ price offers are derived from PJM historical data by averaging over the ratio of each ancillary service price to the energy price [18].

The BESS owner has a lithium-ion battery unit at Bus 13 with the operational parameters specified in Table I. The useful life of the battery is 6000 full charge/discharge cycles with cycle depth of 80%, and the battery replacement cost is 200 k$/MWh price. This information is used for incorporating the battery degradation cost into the framework with the degradation cost model. Details on the degradation cost function can be found in [10].

This case compares the BESS’s performance with and without modeling its degradation cost, when the BESS only participates in the energy market. Figs. 2(A) and 2(B) show the BESS’s scheduled power (in MW) and SOC (in MWh) when the BESS participates in the energy market only, without and with the degradation model, respectively.

Fig. 2(A) shows when degradation costs are neglected, energy arbitrage between hours is a profitable bidding strategy for a BESS only participating in the energy market. Hence, the BESS buys energy (with negative scheduled power) and charges itself (with an increase in the SOC) during initial hours with lower energy prices (hours 1~6), in order to sell the stored energy during hours with higher energy prices (hours 11~18). As shown in Fig. 2(A), BESS uses all its available capacity for this inter-temporal energy arbitrage. The BESS also experiences deep charge/discharge cycles for maximizing its revenue, which, in real-world practices, may result in considerable battery life loss and degradation cost.

Fig. 2(B) shows when degradation costs are considered in the BESS’s bidding strategies, the BESS’s participation in the energy market is very limited. It only buys 12 MWh energy once during off-peak hours (hours 3~4) and sells this amount of energy once during peak hours (hours 14~16). In this case, when battery degradation is considered, participating in the energy market does not generate enough revenue to overcome the total degradation cost. Therefore, the BESS limits its participation in the energy market.

This case compares the BESS’s performance with and without modeling its degradation cost, when the BESS participates in both energy and reserve markets. Figs. 3(A) and 3(B) show the BESS’s scheduled power (in MW) and SOC (in MWh) when the BESS participates in the energy and reserve markets, without and with the degradation model, respectively.

In Fig. 3(A), the BESS prefers participating in the reserve market more than the energy market. In most hours, the BESS’s scheduled power in the reserve market reaches its 50 MW charge/discharge limit. The BESS also performs inter-temporal energy arbitrage to gain revenue from the energy market, but the amount of energy arbitrage is limited as the BESS reaches its charge/discharge limit. BESSs with higher charge/discharge limits may participate more in the energy market.

As shown in Fig. 3(B), when the degradation cost is modeled, the reserve market becomes the only source of revenue for BESS, and BESS does not participate in the energy market. In essence, as reserve deployment is not modeled in this work, participation in the reserve market does not incur any degradation cost for the BESS. Hence, BESS provides reserve services as much as it can and does not perform any energy arbitrage between hours as the revenue generated by this activity does not overcome the degradation cost in this test system. In this case, the battery degradation does not have a significant impact on the BESS’s operation in the reserve market, but this degradation reduces the BESS’s energy market participation significantly.

This case compares the BESS’s performance with and without the degradation cost model, when the BESS participates in both energy and regulation markets. Figs. 4(A) and 4(B) show the BESS’s scheduled power (in MW) and SOC (in MWh) when the BESS participates in the energy and regulation markets, without and with the degradation model, respectively.

In this case, the simulation results without (Fig. 4(A)) and with (Fig. 4(B)) the degradation cost model are more similar to each other than those in the previous cases. Under both frameworks, the BESS participates the most in the frequency regulation market. In Fig. 4(A), when battery degradation is neglected, BESS also participates in the energy market and performs energy arbitrage when its scheduled power for regulation capacity provision does not reach its charge/discharge limit at the beginning and end of the day. A BESS with a higher charge/discharge limit may participate more in the energy market. In Fig. 4(B), when battery degradation is considered, the BESS limits its energy market participation to only compensating for the energy discharged due to following AGC signals. Although the AGC signals have zero-mean over each market clearing interval, the battery’s charge/discharge efficiency is not 100%. In the framework with models for the degradation cost and AGC signal dispatch, this fact leads to a SOC reduction when the BESS follows the AGC signals. Therefore, BESS needs to compensate for the energy discharged due to AGC signal following to maintain a similar SOC at the beginning and end of the day.

The SOC curve of Fig. 4(B) has small fluctuations (on top of the overall shape) all over the simulation horizon. These fluctuations are low-depth charge/discharge cycles caused by following AGC signals. As these low-depth charge/discharge cycles will result in low degradation costs, the BESS’s regulation market participation will not cause significant degradation costs. Hence, degradation cost modeling in these simulations does not affect the operation pattern of BESS in the regulation market a lot, and this pattern is similar in 4(A) and 4(B).

This case compares the BESS’s performance with and without modeling its degradation cost, when the BESS participates in all the energy, reserve, and regulation markets. Figs. 3(A) and 3(B) show the BESS’s scheduled power (in MW) and SOC (in MWh) when the BESS participates in these markets, without and with the degradation model, respectively.

Comparison of Fig. 5(A) with Fig. 5(B) shows that similar to the previous cases, the operation of BESS in ancillary services markets does not change a lot by considering degradation costs in this test system. This happens since 1) the reserve deployment is not modeled; and 2) providing regulation services does not cause high degradation costs. However, the BESS’s energy market participation is reduced after considering degradation costs, since doing energy arbitrage in the energy market is less profitable for BESSs with degradation costs.

Fig. 6 shows the BESS’s total revenue from each market and its degradation cost in Cases 1~4, using frameworks with and without the degradation cost model.

Comparison of Bar A with Bar B for each case in Fig. 6 validates our previous observations for this synthetic test case system build upon real-world data. First, considering batteries’ degradation cost in BESS operation scheduling significantly reduces BESS’s participation and revenue in the energy market. Second, BESS’s participation and revenue in reserve and regulation markets do not change significantly by modeling degradation costs. However, modeling reserve deployment in the studies may affect the results for the reserve market.

Fig. 4(B) and Fig. 5(B) show the degradation costs in Cases 3~4 are mostly associated with the BESS’s regulation market participation and AGC signal following. Fig. 6 shows the total degradation costs in Cases 3~4 are around 1% of the BESS’s revenue from the regulation market, which is negligible in comparison to the revenue. Therefore, neglecting the degradation costs caused by AGC signal following should not have a considerable impact on the results. This neglection could significantly reduce the optimization complexity and computation time. It could be adopted in future work when other BESS/market operational details need to be considered.