Exploration of Grid Faults Based on CVPFI Mechanism and CNN-LSTM Models

In this paper, a grid fault identification, prediction and self-healing system based on CVPFI (Conditional Variance Partial Feature Importance) mechanism and CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory Network) model is proposed. The system utilizes advanced deep learning and big data analytics, aiming to improve the accuracy and response speed of fault prediction. First, the system monitors the state of the grid in real time to quickly locate fault areas. It identifies abnormal data in real time through the fault recognition system, analyzes the alarm information of protection devices and the state change information of circuit breakers, and provides a two-dimensional graphical representation of the grid operation status. Second, the system predicts the future operating conditions of the equipment based on the historical and real-time status data of the equipment by incorporating deep learning algorithms, reserving a safe time window for operation and maintenance personnel and effectively reducing the probability of fault occurrence. Finally, based on the results of fault location and prediction, an efficient self-healing system is designed to isolate the faulty area and quickly restore the grid at the lowest cost by automatically adjusting the grid operating parameters or enabling standby solutions, optimizing the strategy to adapt to the diversity and variability of the grid environment and demand, and ultimately realizing the self-healing of the grid. The advantages of this system are its high efficiency, accuracy and automation, which greatly enhance the operational efficiency and safety of the grid. In this way, this paper provides a new solution for rapid localization, prediction and self-healing of grid faults.