Deterioration Trend Prediction Model of Hydropower Unit Based on Improved SVM-GRU

Under the background of ’double carbon’ goal and ’building a new power system with new energy as the main body’ large-scale new energy access, due to the strong volatility of new energy, makes hydropower units frequently start and stop and carry out power regulation. However, frequent start-stop and power regulation will adversely affect the operating state and life of hydropower units. With the long-term operation of hydropower units, the problem of unit deterioration is becoming more and more serious. In order to accurately evaluate the health state of the unit and predict the deterioration trend of the unit, a prediction model of the deterioration trend of the hydropower unit based on improved support vector machine (SVM), variational mode decomposition (VMD) and gate recurrent neural network is proposed. The model is based on improved support vector machine algorithm and field test data to establish the unit health state model. Secondly, the trend sequence of unit deterioration degree is calculated according to the health state model. Thirdly, the deterioration degree trend sequence is input into the variational mode decomposition algorithm for decomposition, and the gate-cycle neural network is used to predict the trend of the decomposition modes. Finally, the forecast sequence of unit deterioration trend is obtained by integrating the result of trend prediction. The results of example analysis show that the method can fit the health state of the unit well and make reasonable and accurate prediction of deterioration trend.