Research On State Trend Prediction Of Essential Electrical Equipment Of Hydropower Station Based On Deep Learning Algorithm

With the advancement of China’s energy structure reform,hydropower stations have increasingly become an important part of the power system.Transformers,generator sets and pump turbines are the essential electrical equipment for hydropower conversion in hydropower stations,and are developing more intelligently.Actively carrying out condition maintenance of essential electrical equipment in hydropower stations,and predicting equipment status trends will improve the safety and stability of hydropower station operation.This article focuses on the state trend prediction of essential electrical equipment in hydropower stations.The main research contents are as follows:Firstly,aiming at the problem of lack of pre-decision and parameter correlation analysis ability for temperature situation during condition maintenance of transformers and generator sets of hydropower stations,the temperature situation prediction model of transformers and generator sets based on the fusion of conditional mutual information and deep learning algorithm is proposed.The model considers the influence of different working conditions parameters on the temperature parameters to be predicted,and uses the conditional mutual information method to analyze the mapping relationship between them.The dimension of the network parameters is reduced by the maximum pooling method,and finally the cyclic neural network is used to predict the time series data.The proposed method is applied to the research on temperature situation prediction of key components of transformers and generator sets in Xianju Power Station,and compared with other traditional models.Experiments show that the proposed method has high prediction accuracy and long forecast period,and can detect abnormal temperature rises of transformers and generators in the early stage.Secondly,in order to improve the accuracy of the existing hydropower generator set and pump turbine vibration data prediction model,the prediction model of generator set and pump turbine vibration trend based on gated recurrent unit and attention mechanism is proposed.The former is to improve the problem of long-term data dependence,uses the attention mechanism to comprehensively consider the weights of different feature vectors,and predicts new feature vectors obtained by integrating the gated recurrent unit network with the attention mechanism.The proposed method is compared with other traditional models and used in the vibration trend prediction research of the essential components of the generator set and pump turbine of Xianju Power Station.It shows that the method of this article has higher accuracy in predicting the vibration data,and has a long prediction period for abnormal vibration failures of generator sets and pump turbines.Finally,based on the above theories and research results,as well as the actual engineering requirements of the existing hydropower station monitoring system for the state trend prediction and analysis function,the Xianju Power Station data platform and the algorithm in this article are used as the information source and model basis,and the Xianju Power Station operation inspection application service the state trend prediction subsystem of electrical equipment is developed in the center health assessment and fault diagnosis system.Firstly,the data platform of Xianju Power Station and the architecture of this subsystem are briefly introduced,and then the system functions,databases,back-end programs and front-end development are designed according to the requirements.The system can perform trend prediction,fault warning and parameter management.Finally this chapter shows the application page of the system.