Prediction Of Stability Degradation Trend Of Pumped Storage Unit Based On Deep Neural Network

Pumped storage power stations are responsible for peak shaving and valley filling,frequency and phase modulation,accident backup and other tasks.In recent years,its share in my country’s energy system has gradually increased.Pumped storage unit is a complex equipment affected by hydraulic factors,mechanical factors and electromagnetic factors.It frequently switches between pumping,power generation,phase modulation,standby,and shutdown conditions.With the increase of operation time,the unit equipment is prone to abnormal vibration,equipment fatigue,equipment deterioration and other situations,threatening the safety of the unit and power station.Accurately analyzing the operation state of the units and carrying out the degradation prediction research of pumped storage units are of great significance for preventing early failures and ensuring the safe and stable operation of the units.This article mainly focuses on the prediction of the stability degradation trend of pumped storage units.Based on the related theories of deep neural network,a method for evaluating the stability degradation state of pumped storage units is proposed.In order to improve the prediction accuracy of the model,a combined depth prediction model of the degradation trend is established.Applying theory to practice,a system for predicting the degradation trend of pumped storage units is designed to effectively improve the efficiency of inspection and maintenance in power station.The main works and innovative achievements of the paper are summarized as follows:(1)Aiming at the problem that the existing unit degradation analysis fails to fully consider the impact of operating condition factors on the unit status,the maximum information coefficient method is studied to analyze the correlation between working condition parameters and unit stability parameters.And a multi-channel DNN health status model is established to accurately analyze the mapping relationship between working condition parameters and unit stability parameters.The stability degradation trend of each channel and the overall degradation trend of the pumped storage unit under different working conditions are extracted to evaluate the unit status.The comparison experiment with the other three algorithms,based on the actual data of the power station,verifies the accuracy of the health model proposed in this paper and the reliability of extracting the degradation trend.(2)Aiming at the problems of low prediction accuracy and short forecast period of the existing degradation trend,a combined depth prediction model is proposed.The variational mode decomposition algorithm is introduced to decompose the fluctuating nonlinear degradation trend.Then,GRU prediction models are established for the decomposed submodes.A correction model based on error prediction is proposed to optimize the prediction results and realize the accurate prediction of the degradation trend of the pumped storage unit.And based on the field data to design a single-step prediction and multi-step prediction comparison experiment,the method proposed in this paper has the highest prediction accuracy and is suitable for predicting the degradation trend.(3)The degradation trend prediction system is designed with a three-tier B/S architecture consisting of data service layer,business logic layer and application service layer.The system has in-depth and multi-faceted warning functions such as degradation trend warning,time series trend warning,analysis model warning,auxiliary analysis,etc.,which fully makes up for the shortcomings of the single static threshold warning of the traditional condition monitoring system,can more comprehensively excavate the main equipment of the unit and provide timely early warning of equipment abnormalities.