Data-Driven Parameter Calibration of Power System EMT Model Based on Sobol Sensitivity Analysis and Gaussian Mixture Model
Yuhong Wang, Bingjie Zhai, Shilin Gao, Yitan Guo, Chen Shen, Ying Chen, Zongsheng Zheng, Yankan Song
Abstract
The parameters of power system electromagnetic transient (EMT) model have great influences on the accuracy of EMT simulation. This paper proposes a data-driven parameter calibration method based on Sobol sensitivity analysis and Gaussian mixture model (GMM) to calibrate the parameters of the power system EMT models. First, the dominant parameters of the power system EMT model are derived based on the derivative-free Sobol sensitivity analysis method. Then, the GMM that describes the relationship between the dominant parameters and the EMT simulation errors is established and solved. Finally, an improved particle swarm optimization algorithm is adopted to optimize the EMT simulation errors and the values of the parameters are obtained according to the minimum error and the conditional probability invariance of GMM. The test results on four different systems show that the proposed method can accurately calibrate all dominant parameters of the EMT models of the various power systems.