Fault Prediction and Rapid Self-Healing Control Method for Smart Distribution Networks Based on Adaptive Neural Networks

With the increasing complexity of the smart distribution community shape, the conventional fault detection and healing mechanism is difficult to meet the requirements of actual-time and stability of the gadget. As a way to remedy this trouble, this paper proposes a fault prediction and fast self-healing manage method for smart distribution community based on adaptive double-branch recurrent neural community (ADBRNN). Primarily based on the conventional gated recurrent unit (GRU) fault predictor and static rule recovery strategy, a temporal interest mechanism is designed, which could dynamically weight the sensor information move at multiple time scales to enhance the accuracy and timeliness of fault caution. A self-recovery controller of the converged graph convolutional community is constructed, which could dynamically optimize the isolation and recovery transfer techniques of faulty nodes by means of combining the actual-time topology, power waft constraints and cargo precedence facts of the distribution community. Similarly, the model education adopts the joint optimization framework of supervised mastering and reinforcement getting to know, and realizes the synergistic enhancement of prediction and manage via alternating the mixed loss feature built by means of minimizing the failure prediction mistakes and the self-recuperation manage reward. The fault vicinity accuracy of the proposed version is improved with the aid of $17.3 \%$, and the common recuperation time is reduced with the aid of $36.5 \%$, that is better than the existing mainstream comparison methods.