State Prediction Of Nuclear Power Equipment Based On LSTM Recurrent Neural Network

In recent years,with the rapid development of the global economy,human demand for power resources is increasing.Nuclear power as an important part of power resources,coupled with its catastrophic consequences caused by accidents,has led people to pay more attention to the safety of nuclear power equipment.Therefore,more effective measures are needed to observe the operating status of the nuclear power system and to avoid risks to ensure the normal operation of nuclear power equipment.How to more accurately predict the operating status of nuclear power equipment has become an urgent problem.This paper,through consulting domestic and foreign related literature and data,to understand the current common data denoising algorithm,combined with the characteristics of nuclear power data,effective denoising of the original data.On this basis,the neural network model is trained using the denoised experimental data,and finally the model is applied to the prediction of the operating state of the nuclear power equipment.The details are as follows:Traditional data denoising algorithms are not suitable for non-stationary and nonlinear data.Both wavelet transform and Empirical Mode Decomposition(EMD algorithm)can effectively analyze and process complex signal data.In this paper,two algorithms are used to denoise the nuclear power data.The experimental results show that the effect of EMD algorithm is better than wavelet transform denoising.Based on the Recurrent Neural Network(RNN),the Long Short-Term Memoryneural network(LSTM)adds memory cells to the hidden layer neurons to control the memory information in the time series.In this way,the shortcomings of the disappearance of the RNN gradient and the lack of long-term memory are effectively improved,and the long-distance time series data can be effectively utilized,which has been widely used in the field of device state prediction.Considering the characteristics of nuclear power data,large amount of data and many noise data,this paper proposes a state prediction method for nuclear power equipment based on LSTM neural network.Combining the EMD algorithm with the LSTM model,the EMD algorithm is first used to denoise the original data,and then the denoised data is used to train the LSTM prediction model to improve the accuracy of the prediction model.Finally,this paper applies the state prediction method of nuclear power equipment based on LSTM cyclic neural network to the state prediction of nuclear power equipment,and proves that the model has higher prediction accuracy through comparison experiments.In this paper,experiments are carried out in the simulation environment under the Keras framework.The experimental results show that the proposed method can effectively predict the operating state of nuclear power equipment.