Multi-timescale optimal scheduling for photovoltaic-integrated energy storage charging station based on enhanced constraints

To address the issues of high operational costs and considerable grid load changes in photovoltaic (PV)-integrated energy storage charging stations, this research provides a multi-timescale optimization scheduling method based on time-varying boundary constraints. The method employs a feature mode decomposition-long short-term memory-Bootstrap interval prediction strategy to boost forecasting accuracy for electric vehicle (EV) charging loads and PV generation, attaining a 95.74% coverage rate. The scheduling framework implements a bi-layer optimization model: the upper layer maximizes the load margin index for grid stability and minimizes scheduling deviation between layers, while the lower layer optimizes individual EV charging strategies by minimizing operational costs, user target residual energy, and scheduling deviations. The innovation comes in introducing dynamic time-varying boundary restrictions informed by prediction intervals, time-of-use power pricing, and PV feed-in tariffs, enabling adaptive energy flow optimization that adapts to forecast uncertainty in real-time. Comprehensive simulations using IBM ILOG CPLEX Optimization Studio (CPLEX) demonstrate that, compared with the single-timescale scheduling scenario, the proposed multi-timescale strategy reduces the total daily operation cost by 17.56% and decreases the root mean square error of grid load fluctuations by 32.4%, while ensuring that all EV users maintain sufficient state-of-charge above 0.5, thereby validating its effectiveness in improving both economic efficiency and grid stability.