Research On Load Rate Prediction Of Nuclear Safety Level Main Controller Based On Machine Learning

Nuclear power plants have strict requirements for the load rate of the main controller of the digital safety system.In order to realize the normal operation and dispatching control of the load rate of the main controller in the safe operation range of the nuclear power plant,it is necessary to predict the load rate of the main controller in order to facilitate the safe and reliable operation of the instrument control system of the nuclear power plant.Therefore,scholars have carried out active research on the prediction of the load rate of the main controller of nuclear power plant.Through analyzing the factors that affect the load rate of the main controller,they find out the appropriate scheme to improve the calculation accuracy of the load rate.With the rapid development of artificial intelligence and the successful application of digital controller load rate prediction cases in thermal power plants,the limitations of traditional load rate calculation models in nuclear power plants are more and more obvious.In view of this situation,this paper applies the machine learning method to the load rate prediction of the main controller of the nuclear power plant,and through a nuclear powerproject example,the simulation experiment of the relevant methods is carried out to verify the effectiveness of this method.The main research work and innovations are as follows:1.Completed the analysis and Research on the influencing factors of nuclear safety level load rate.Based on the characteristic that the main controller of the domestic safety level digital control system generally adopts the fixed period processing mechanism,and the same design is adopted for the safety level digital control system platform of the nuclear power plant,so the load rate prediction problem of the main controller can be transformed into the study of the data processing time.Through the analysis and analysis of data processing mechanism,it can be seen that the factors that affect the data processing time of the main controller mainly include the characteristics of the program module itself,the hardware configuration and the natural environment of operation.On this basis,the experiments of feature extraction for the influencing factors of load rate under different schemes are completed to avoid the influence of redundant variables on the accuracy of load rate prediction and training time.2.According to the requirements of load rate safety assessment of the main controller of the safety level I &C system of nuclear power plant,a BP neural network model for load rate calculation is established for the first time for the main controller of a safety level digital I &C platform.Based on the data of a nuclear power project,the neural network model is used to train the prediction network.In view of the disadvantages of the model easily falling into the local optimization problem,a load rate prediction model of the main controller based on genetic algorithm optimization neural network(GA_BP)is proposed,which can solve the limitations of BP neural network well and improve the prediction accuracy at the same time.3.Aiming at the problem that the program quantity set of the safety level control system of a nuclear power project is insufficient and the characteristic quantity is large,and the conventional multiple linear regression method can not get the ideal result,this paper further establishes the support vector regression(SVR)model as the main controller load rate prediction model,and completes the prediction model training with the designated data set samples.4.Based on the above three models,the load rate prediction of the main controller of the reactor protection system of a nuclear power project is simulated.The results show that the SVR model is more suitable for the prediction accuracy of the load rate of the safety level main controller of nuclear power plant.