Data-Driven Microgrid Operation Toward Optimized Battery Energy Storage Lifetime Degradation

This paper proposes a new data-driven approach for two-stage operation of a microgrid (MG) towards optimized battery energy storage (BES) lifetime degradation. At the first stage (day-ahead), the BES degradation range over the scheduled day is optimally allocated under day-ahead forecast of random variables. To accurately quantify BES degradation under temperature impact, a convex BES electrothermal-aging model is constructed. Then, a dynamic model constraint coefficient determination method is proposed for optimizing BES lifetime degradation using the historical operation and updated current information. At the second stage (real-time), an approximate dynamic programming method is developed to real-time optimize the MG operation after multiple uncertainties are realized. Considering the temperature dependent constraints of BES, a sequential decomposed value function slope update method is designed for enhanced optimization feasibility, BES aging quantification accuracy, and training efficacy, which can deliver high-quality solutions under uncertainties. Case studies on a MG system demonstrate the effectiveness of the proposed approach.