Summary
Optimal market bidding for battery energy storage systems (BESS) is a challenging problem, mainly due to the unknown market outcome and resulting market clearing prices. To manage the uncertainty, stochastic optimization [1] can be used along various machine learning-based decision support tools. In this paper, we focus on stochastic optimization due to its ability to consider electricity price uncertainty and potential to achieve optimal solutions with correct input parameters. The mixed-integer linear programming (MILP)-based approach is flexible and can adapt to any environment without a need for historical data, training or an initial learning phase. Open questions are how to efficiently handle uncertain forecast data and use the heterogenous market landscape to ensure reliable delivery of reserves when called upon. In an earlier contribution, we studied a system comprising of renewable energy sources and a BESS and deployed self-scheduling for the energy markets [2]. This approach provided an effective framework for securing the BESS state-of-charge (SoC) levels, while aiming to successfully participate in the reserve markets. However, it had two main fallbacks: 1) ideal conditions were assumed with neither ancillary events nor any SoC inaccuracies taking place, and 2) due to selfscheduling there is little prospect for arbitrage trading.
Read more Read lessIn this paper, we mainly focus on the first aspect. To support the feasibility of the system with more risky energy bidding scenarios, we include the imbalance markets to account for lost bids.
One effect of imbalance markets could be that it drives the system operation towards an intended infeasibility in the quest for expected higher profits. In multi-energy markets [4], the decision chain must therefore be well considered to secure feasible operations. Another challenge with defining a bidding strategy for the next days, is how to verify the robustness of the derived solution. For this, we propose two complementary methodologies to evaluate the robustness of the proposed approach:
• Add deviations to the calculated SoC-values to account for non-ideal battery operations • Simulate ancillary events, where accepted bids are either partially or fully consumed The SoC-behavior is randomly perturbed, leading to efficiency deviations on the order of 12%. The timing of ancillary events is selected randomly but on a realistic level (assuming around 1-2 daily events). In a multi-day simulation, this enforces the system to recover itself through consecutive energy bidding schemes or imbalance markets. We systematically test and compare the benefits of using a stochastic bidding scheme towards using a deterministic approach, which is based both on known market prices and on the 50% quantile and evaluate the effect of disturbances to the realized profits. The algorithm used follows the main principles of the work reported in [5]. We conclude the paper by highlighting the findings and the future challenges on market bidding.
Additional informations
| Publication type | Session Materials |
|---|---|
| Reference | C5_12488_2026 |
| Publication year | |
| Publisher | CIGRE |
| Country | Germany |
| Study committees | |
| File size | 1 MB |
| Price for non member | 30 € |
| Price for member | 30 € |
Authors
HARJUNKOSKI Iiro - Hitachi Energy Research Germany; GHAWASH Faiq - Hitachi Energy Research Germany
