Operating Health State Prediction and Evaluation of Excitation Unit Based on GMM and LSTM

The excitation unit is the weak link in the generator operation since it produces internal deterioration under long-term operation. To evaluate the health condition of the key excitation unit and generate warning information in advance, this work proposes a health assessment method based on the Gaussian mixture model (GMM) and LSTM neural network. Firstly, the weights of each type of data will be determined by obtaining the electrical parameter data of the generator excitation unit from the SIS system, using the infrared image data of carbon brush and excitation transformer based on infrared image acquisition system, and the correlation between each electrical parameter data and temperature on account of the MIC correlation analysis method. On the top of it, LSTM neural network prediction model is established. And establish a GMM-based health benchmark model based on the data collected in the field during healthy operation, and design the health decline index HDI using the Marxist distance for evaluating the health status and deterioration degree of the unit. Finally, the method of this paper is validated by the 2022 operation data of a thermal power plant in Weihai, which proved that this method can provide a basis for regular maintenance and early identification of faults in the power plan and quantifying the degradation degree of the excitation unit under long-term operation and making early warning before the abnormal occurrence of the unit.