A Prediction–Optimisation Method for Power Control of Integrated Photovoltaic‐Energy Storage‐Charging Stations

ABSTRACT

With the rapid adoption of electric vehicles (EV), integrated photovoltaic–energy storage–charging stations (PV–storage–charging stations) not only enhance the utilisation of renewable energy and the capacity of charging services but also face challenges such as PV output fluctuations, stochastic charging demand and the complexity of storage scheduling. This paper proposes an intelligent scheduling method for PV–storage–charging stations under a prediction–optimisation framework, establishing a unified multi‐energy flow queuing model that jointly represents PV output, storage buffering, grid procurement and EV charging demand. A dual‐input queuing mechanism based on a Gaussian mixture model is introduced to capture the heterogeneous charging characteristics of fast‐ and slow‐charging users. A CKB‐Net prediction model with an attention fusion mechanism is then developed to achieve high‐precision forecasting of PV generation and vehicle arrivals, enabling the generation of representative daily operation scenarios. Based on these forecasts, a two‐layer optimisation model is proposed, consisting of intraday economic scheduling and full life‐cycle configuration optimisation, to coordinate peak–valley arbitrage with storage degradation cost. Simulation results verify the model's advantages: the proposed queuing model improves charging demand prediction accuracy by 9.04% compared to a single‐input M/M/c model; CKB‐Net reduces the MAE by 55% over an LSTM baseline; and the proposed scheduling framework increases daily profits by 261% compared to a conventional method. These results demonstrate the effectiveness of the prediction–optimisation strategy for achieving intelligent, economic and sustainable operations of PV–storage–charging systems.