Optimal Autonomous Control for Distribution Transformer Area with High Photovoltaic Penetration Based on CSBO-LSTM

With the large-scale deployment of distributed photovoltaic (PV) and energy storage system (ESS) in distribution network transformer area, problems such as low management efficiency and difficult collaborative regulation have become increasingly prominent. To enhance the autonomous control capability of transformer area with high PV penetration, this paper proposes a “data-model” dual-driven collaborative optimization model that integrates the Circulatory System Based Optimization (CSBO) algorithm and Long Short-Term Memory (LSTM), constructing an architecture of “offline optimization-knowledge migration-online inference”. The model takes maximizing PV power consumption as the core objective, while considering voltage stability regulation and minimum network loss, to construct a multi-objective optimization function. Firstly, the improved CSBO is used to solve the multi-objective optimization problem, accurately formulate the PV reactive power regulation and ESS charging/discharging strategies, and generate the optimal operation dataset of the transformer area through multi-scenario simulation. Furthermore, leveraging the advantage of LSTM in processing time-series data, a real-time response model is constructed through deep training to achieve rapid perception of grid status and dynamic control decisions. Experimental results show that the model maintains the voltage of the transformer area near 1.00 p.u., while significantly reducing network losses. The organic combination of CSBO and LSTM effectively improves data completeness construction, model complexity control, and real-time decision response, providing new ideas and implementation solutions for the autonomous control of transformer areas with high PV penetration.