Parameter Identification of Electromechanical Oscillation Mode in Power Systems Driven by Data: A Quasi-Real-Time Method Based on Randomized-DMD-Multilayer Artificial Neural Networks

Abstract

With the increase in the power system scale, the identification of electromechanical oscillation mode parameters by traditional numerical methods can no longer meet the requirements of complete real-time analysis. Therefore, a method based on machine learning (multilayer artificial neural networks) is proposed to identify the electromechanical oscillation mode parameters of the power system. This method can take the monitorable variables of the WAMS as the input of the model and the key characteristic information such as frequency and damping ratio as the output. After processing the input and output data with randomized dynamic mode decomposition (randomized-DMD), their mapping relationship can be analyzed by using the multilayer neuron architecture. The simulation results of the 4-generator 2-area system and the IEEE 16-generator 5-area system show that this method can accurately calculate the key characteristic parameters of the system without considering the change in the control parameters and after the offline training of historical data, which shows higher accuracy, stronger robustness, and sensitive online tracking ability.