Power System Fault Maintenance System Based on Deep Learning Algorithm

Malfunctions in the power system may lead to power outages and even equipment damage. In order to improve the reliability of the power system, this study constructs a power system fault maintenance system based on CNN-LSTM. The system monitors the operation data of the power system in real time and uses the CNN-LSTM model for fault detection and diagnosis to achieve fast and accurate fault location. In the research process, real-time operational data such as current, voltage, and frequency of the power system are first collected through sensors and data acquisition systems, and then denoised, normalized, and feature extracted to meet the input requirements of the model. Next, the article designs a deep learning model that combines CNN and LSTM. CNN is used to extract spatial features from power signals, while LSTM processes time series data and dynamically changing power signals. In terms of fault diagnosis, models are used to analyze real-time data, detect and locate faults in a timely manner, thereby improving the efficiency and accuracy of power system fault handling. The experimental results show that the response time of the system in the first dataset is 585ms, while the response times of the CNN based and LSTM based fault diagnosis systems are 607ms and 640ms, respectively. In terms of accuracy, the system’s fault detection accuracy ranges from 90% to 100%, while the detection accuracy of fault repair systems based on CNN and LSTM is both between 85% and 98%. These data indicate that the fault diagnosis system combining CNN and LSTM has significant advantages in response time and also performs well in accuracy.