A Physical-Data Combined Power Grid Dynamic Frequency Prediction Methodology Based on Adaptive Neuro-Fuzzy Inference System

The fast and accurate dynamic frequency prediction of power grid after disturbance contributes to formulate corresponding emergency frequency control strategy and prevent system frequency from crashing. Current physical or data methods for frequency prediction has the contradiction between computation speed, accuracy, and reliability in practical application. In this context, a physical-data combined power system dynamic frequency prediction methodology based on adaptive neuro-fuzzy inference system (ANFIS) is proposed. An improved average system frequency (ASF) response equivalent model is applied to establish the crucial physical relationship between the frequency responses and disturbances of power system integrated with wind farms. Additionally, the datamining-based support vector regression (SVR) is used to fully extract features contained in the power system operation data. Then, the hybrid prediction methodology is realized by utilizing the ANFIS to fully extract and combine features of the ASF and SVR methods. The case study on the New England 39-nodes system with high wind power penetration shows the proposed methodology can predicts the power grid dynamic frequency after disturbance quickly and accurately. Moreover, it effectively improves the accuracy of frequency prediction compared to two individual method and reduces the dependence on sample size.