Power Grid Repair Time Prediction and Scheduling Using GA‑BP for Emergency Power Supply Operations

Abstract

Accurate prediction of power grid repair time and efficient allocation of resources are critical to improving the response capability of emergency power supply operations and infrastructure resilience under extreme events. However, current methods still show clear limitations in prediction accuracy, scheduling efficiency, and the coordination between these two aspects. This manuscript presents a prediction and scheduling strategy for emergency power supply scenarios. It develops a repair-time prediction method by combining a Genetic Algorithm (GA) with a Backpropagation (BP) neural network, and constructs a coordinated scheduling model using Improved Ant Colony Optimization (ACO). Experimental results show that the prediction method achieves a coefficient of determination of 0.958 for the repair time task, with a Mean Absolute Error of only 0.181 h. The prediction accuracy is higher than that of the comparison models. In addition, the scheduling model restores the load in 5.98 h under severe fault conditions. The average repair radius stays within 8.46 km, the resource use rate reaches 76.39%, and the load loss is more than 41.43% lower than that of the comparison models. These results show that the method proposed in this manuscript provides stable prediction performance and effective scheduling ability in emergency power supply operations. It offers a reliable technical path for an intelligent emergency command system and helps improve response speed and resource utilization efficiency in complex power grid environments.