Research of Outage Duration Prediction Based on CNN-SVR Under Gray Wolf Optimization

The outage duration prediction can improve utility decision-making practices and enhance restoration management procedures for electric utilities. For the purpose of precise outage duration prediction, this paper proposed an outage duration feature extraction prediction model based on convolutional neural network (CNN) and support vector regression (SVR). This model can introduce mathematical properties through traditional machine learning over deep feature extraction, which enhances the reliability of the model. Moreover, the optimal model parameter combination was obtained by the gray wolf optimization (GWO) algorithm at a low computational cost. Finally, prediction experiments were conducted on the outage duration dataset, illustrating the superiority of the proposed model through comparison with the latest methods.