Session: 07-10: Simulations and Predictions - II

Paper Number: 134726

134726 - Research on Long Term Trend Prediction of Nuclear Power Plants Based on Integrated Framework 

Abstract: 

In the current field of nuclear energy, the development of third-generation nuclear power plants is the mainstream, and their safety and economy are important directions for nuclear energy development. The safety of third-generation nuclear power plants needs to be greatly improved, and more advanced safety measures such as artificial intelligence safety technology need to be adopted to ensure that nuclear power plants can still maintain safety in extreme accident situations. Therefore, the normal, stable, and safe operation of nuclear reactor systems is particularly important. In order to improve the safety and stability of nuclear power systems, this study proposes an integrated framework based model for predicting the duration of nuclear power plants. The research objective is to improve the advanced perception ability of nuclear power plant operating conditions, ensure that the algorithm can predict the trend of operational data changes that nuclear power plants will present in the long future, provide more intuitive information for nuclear power plant operators, and thus improve the safety and stability of the nuclear power system. For this purpose, this study constructed an integrated framework based on models such as BP, LSTM, Informer, etc. The integrated framework can combine the advantages of all models, improve the robustness and accuracy of the output results, enhance the adaptability of the model, and ensure good prediction accuracy even under unknown operating trends, making the model practical, And use the operating data of the Qinshan simulation machine to verify the model. The results show that the integrated framework based long-term trend prediction model for nuclear power plants has lower MAE and MSE values compared to traditional BP and LSTM, which indicates that the integrated framework based long-term trend prediction model can have lower errors. And after comparing the predicted values and actual input values of the integrated framework with BP and LSTM, it can be found that the predicted data trend of the integrated framework has a small deviation from the actual value, which means that the trend prediction effect in the comparison is the best. In addition, the model also achieved the best long-term predictive performance and robustness under different accident conditions. The results indicate that the long-term prediction model for nuclear power plants based on the integrated framework can predict the future trend of data with minimal error when validated by the operating data of the Qinshan nuclear power plant simulator, providing strong support for the safety and stability of the nuclear power system.

Presenting Author: Sichao Tan Harbin Engineering University, Harbin 150001, China

Presenting Author Biography: Tan canyi is mainly engaged in research on artificial intelligence based early fault diagnosis technology at Harbin Engineering University.

Authors:

Canyi Tan Harbin Engineering University, Harbin 150001, China
Biao Liang Harbin Engineering University, Harbin 150001, China
Bo Wang Fundamental Science on Nuclear Safety and Simulation Technology Laboratory, Harbin Engineering University, Harbin 150001, China
Jiangkuan Li Harbin Engineering University, Harbin 150001, China
Rui Han Harbin Engineering University, Harbin 150001, China
Sichao Tan Harbin Engineering University, Harbin 150001, China
Ruifeng Tian Harbin Engineering University, Harbin 150001, China