Prediction of Temporal Operational Risks in Power System Based on Deep Neural Networks

The stochastic and intermittent nature of renewable energy generation results in challenges for the operational secure of renewable energy power systems, making it particularly necessary to comprehensively assess and forecast the operational risks of the power system. Therefore, drawing upon the theories of power system scheduling, risk assessment, and artificial intelligence, a method for predicting the temporal operation risk of the power system using deep neural networks(DNN) is proposed. Firstly, A second-order cone (SOC) relaxation-based alternating current (AC) power flow dispatch model is developed. Secondly, quantitative assessment indicators are established and fitted using a Gaussian mixture model(GMM) based on historical dispatch data. Then, using a given confidence level and corresponding score table, interval assignment for assessment indicators is implemented. Finally, DNN is employed to train a predictive model, achieving situational awareness of future operational risks in power system.