Deep Learning Based Muti-Objective Reactive Power Optimization of Distribution Network with PV and EVs

With the high penetration of photovoltaic (PV) and electric vehicle (EV) charging and replacement power stations connected to the distribution network, problems such as the increase of line loss and voltage deviation of the distribution network are becoming increasingly prominent. The application of traditional reactive power compensation devices and the change of transformer taps has struggled to meet the needs of reactive power optimization of the distribution network. It is urgent to present new reactive power regulation methods which have a vital impact on the safe operation and cost control of the power grid. Hence, the idea that applying the reactive power regulation potential of PV and EV is proposed to reduce the pressure of reactive power optimization in the distribution network. This paper establishes the reactive power regulation models of PV and EV, and their own dynamic evaluation methods of reactive power adjustable capacity are put forward. The model proposed above is optimized via five different algorithms and approximated through the deep learning when the optimization objective is only set as line loss and voltage deviation. Simulation results show that the prediction of deep learning has an incredible ability to fit the Pareto front that the intelligent algorithms obtain in practical application.