The rapid integration of photovoltaic (PV) generation and energy storage systems has significantly increased the operational complexity of low-voltage direct current (LVDC) distribution networks in zero-carbon parks. Under highly variable operating conditions, conventional DC protection schemes relying on fixed overcurrent thresholds often suffer from maloperation or failure to trip, particularly during fluctuations in PV power, load switching, and changes in network topology. To address these challenges, this paper proposes an adaptive DC protection strategy based on an artificial neural network (ANN)-driven dynamic threshold optimization mechanism. The proposed method replaces static protection settings with an adaptive threshold that is continuously updated according to real-time system operating conditions. A dual-layer ANN architecture is developed to capture the nonlinear relationship between grid parameter variations and optimal protection thresholds. To enhance learning accuracy and convergence performance, the backpropagation neural network is further optimized using an improved grey wolf optimizer with a nonlinear convergence factor and mutation operator. The optimized ANN enables rapid and reliable threshold adjustment without relying on high-speed communication, making the scheme suitable for decentralized and edge-computing-based protection architectures. A comprehensive simulation platform for a PV-energy storage LVDC distribution system is established in MATLAB/Simulink to generate training and testing datasets under diverse scenarios, including variations in PV output, load shedding, different fault types, and measurement uncertainties. Simulation results demonstrate that the proposed adaptive protection strategy effectively eliminates threshold mismatch problems observed in fixed-setting methods. The results confirm that the proposed ANN-based adaptive protection strategy provides a robust, fast, and communication-independent solution for reliable protection of LVDC distribution networks with high penetration of renewable energy sources.
Tan et al. (2026) studied this question.