“The research team proposed a fault recovery method for DG distribution networks that combines net load forecasting, designed a CNN BiLSTM hybrid prediction model that integrates meteorological feature attention mechanism, constructed a multi-objective power supply recovery model, and used a genetic quasi Newton hybrid optimization algorithm to solve the islanding partitioning and reconstruction scheme. This provides a solution to the problem of complex fault recovery and insufficient islanding operation stability in distribution networks caused by high penetration of distributed power sources, and provides a new method for rapid self-healing of high proportion new energy distribution networks.”