Voltage Stability Control Method for Distribution Networks Under Distributed Generation Load Fluctuations Based on DSCN

Aiming at the voltage stability issues in distribution networks caused by load fluctuations under high penetration levels of distributed generation (DG), this paper proposes a voltage stability control method for distribution networks under DG load fluctuations based on the Depthwise Separable Convolutional Network (DSCN). A stochastic output model for distributed generation is established. By characterizing the power controllable range of distributed generation, the load fluctuation values of distributed generation are derived. A voltage stability control objective function for the distribution network is formulated, encompassing a day-ahead optimization and regulation layer and a real-time rolling control layer. The DSCN is introduced to fuse multi-source data, including distributed generation output and node voltages, extracting spatio-temporal features through depthwise separable convolutions and enhancing the perception capability of critical nodes with an attention mechanism. A multi-objective loss function is designed, integrating the voltage stability control objective function of the distribution network and a control smoothness regularization term to achieve voltage stability control. Experimental results demonstrate that the proposed method achieves a maximum voltage fluctuation of 1.5 V, a static voltage stability margin fluctuation range of 21.5% to 24.5%. This method significantly improves the voltage stability control accuracy and robustness of distribution networks in high-proportion DG scenarios while ensuring computational efficiency.