Topology design of medium voltage distribution network based on nonlinear SVM algorithm and kernel function decision

The medium-voltage (MV) distribution systems possess considerable potential for hosting high-penetration distributed energy resources (DERs), leveraging their inherent power adaptation flexibility and digitally enabled transformer configurations. Increasing the renewable energy hosting capacity in regional grids places higher demands on dynamic power balancing mechanisms. Therefore, it is imperative to establish accurate source-load management models and fast-response regulation frameworks. This paper proposes a kernel function-based topology quality assessment model. The methodological procedure is as follows: First, a composite evaluation function is formulated to characterize both the nodal connection density within individual topologies and the degree of isolation among different topological regions. Second, constraint conditions are explicitly defined, addressing both Topology Scale Balance (TSB) and overall connectivity requirements. Finally, the optimization objective is achieved through a nonlinear mapping technique using Support Vector Machines (SVM) with Radial Basis Function (RBF) kernels. Simulation results demonstrate that the proposed topology method effectively enhances several key grid performance metrics: voltage regulation responsiveness improves by over 37%, the utilization rate of reactive power compensation equipment increases by 22%, and fault isolation time is reduced by 27%. These technical outcomes reinforce the system’s transient stability control capability. Unlike traditional approaches that rely on fixed kernel functions or static weighting schemes, the proposed method employs a Dynamic Weight Adaptation-Kernel Parameter Coordination (DWA-KPC) closed-loop decision-making framework. This framework enables the real-time, coordinated adjustment of both the Radial Basis Function (RBF) kernel width and the penalty factor. For complex challenges in distribution network topology optimization-characterized by high dimensionality, nonlinearity, and multi-objective trade-offs-this method demonstrates superior comprehensive performance and enhanced engineering applicability compared to conventional techniques.