A battery degradation-aware energy management system for agricultural microgrids

The integration of renewable energy sources (RESs) into power grids underscores the necessity for efficient energy storage solutions to ensure power balance and increase grid reliability. Although battery energy storage systems (BESSs) are pivotal for storing excess energy from RESs and mitigating peak demand periods, their chemical nature poses limitations, particularly in microgrid (MG) applications, due to degradation concerns that can lead to reduced performance over time. This necessitates careful consideration of degradation effects in optimizing system design and operation. This paper addresses this issue through developing a novel methodology aimed at optimizing the operation of renewable-based MGs while accounting for the degradation mechanisms of the battery storage systems. A machine learning model based on the XGBoost strategy is developed to predict the remaining useful life (RUL) of Lithium-ion (Li-ion) batteries, leveraging initial battery characteristics. This data-driven model is then incorporated into the day-ahead scheduling problem of an agricultural MG as a use-case to assess the impact of battery degradation modeling on the MG operation in both grid-connected and island operation modes. The proposed methodology utilizes the Coati Optimization Approach (COA) to determine optimal battery charging and discharging policy related to battery cycle limitations. The Monte Carlo Simulation (MCS) approach is employed to generate different scenarios reflecting varying power generation from RESs, electricity price volatility, and load demand variations. Case studies conducted on a real-world agricultural MG in Ankara, Turkey, demonstrate the effectiveness of the proposed methodology in reducing the total MG costs including both operational and degradation costs. Sensitivity analyses underscore the robustness of the methodology across various RES penetration levels and market conditions. Results reveal a reduction of 55.30% and 41.23% in the degradation cost of the agricultural MG in grid-connected and island modes, respectively, through the integration of the proposed data-driven-based battery degradation modeling.