Parameters Identification of Park Three-Circuit Generator Model

Abstract

The research is devoted to the problem, associated with the development of a parameter identification method of a Park three-circuit mathematical model formalized in the Order of Energy Ministry No. 102 during current operation. The article provides a brief review of methods for generator parameters identification and a machine learning method based on convolutional neural networks (CNN) that is implemented. To prepare a training sample, an analysis of the real generator under study is carried out, in particular, characteristic modes of its operation, specifics and variability of disturbances. Based on this analysis, the equivalent of a real electrical network is modeled with the option of simulating similar real events. A deep learning approach using a CNN is used to identify parameters. The final result of this work is the test of a machine learning model based on synthetic data. As a result, the average error in parameters identification according to mean absolute percentage error indicator was 2.8%. This research would be interested to organizations that own and manage generation equipment.