Construction of Power System Fault Prediction Model Based on Deep Convolution Neural Network

Traditionally, the power system fault prediction method can reflect the operation state of the power system to a certain extent, but it is not enough to deal with large-scale and high-dimensional data, and it is difficult to capture the weak signal before the fault occurs. In this study, we introduce a novel power system fault prediction model utilizing a deep convolutional neural network (DCNN), comprising an input layer, a series of alternating convolutional and pooling layers, a fully connected layer, and an output layer. The model's input consists of time-series data from power system operations, and it achieves precise fault predictions through the implementation of an optimization algorithm and strategic network architecture design. For model training, we employ a cross-entropy loss function and the Adam optimizer, while optimizing hyperparameters using a combination of grid search and random search techniques. Additionally, an early stopping mechanism is integrated to mitigate overfitting. Experimental findings indicate that our DCNN model outperforms conventional machine learning approaches such as support vector machines (SVM), random forests (RF), K-nearest neighbors (KNN), and logistic regression (LR) in terms of accuracy, recall, and F1 score. The model of this study not only provides technical support for the safe operation of power system, but also provides scientific basis for the decision-making of power enterprises, which has important theoretical significance and practical application value.