Prediction And Analysis Of Nuclear Radiation Dose Rate Based On Multi-Classification Fusion Model

In recent years,the country vigorously advocates the use of new energy sources,and nuclear energy,as one of the most promising new energy sources,is being exploited in large quantities,and the number of nuclear power plants is increasing day by day.HPIC dose rate is an important indicator to measure the gamma dose rate in the environment,which is also the core indicator of environmental nuclear radiation monitoring.If a more effective HPIC dose rate monitoring and early warning mechanism can be established,it will have a great impact on the development and construction of nuclear power plants.It is known from research that there are many factors affecting HPIC dose rate,including but not limited to atmospheric temperature,atmospheric humidity,atmospheric pressure,rainfall,and monitoring stations.The environmental nuclear radiation monitoring data used in this paper is a real dataset taken from 2015 to 2018 under the industrial internet of things environmental monitoring system of a nuclear power plant in Fujian Province,containing a total of14145873 observations.Based on this dataset,the ARMA time series model,random forest model,BP neural network model,and XGBoost algorithm model are applied to the prediction and analysis system of environmental nuclear radiation dose rate of nuclear power plant.In this paper,based on the random forest model,BP neural network model,and XGBoost algorithm model,a new multiclassification fusion model(MCFM)is proposed,which can effectively improve the classification accuracy by using the parallel fusion of multiclassification systems.The classification accuracy is used as the model judging index,and the four models are compared and analyzed.The experimental results show that the ARMA model,as a traditional statistical analysis model,performs better in fitting the HPIC dose rate development trend of environmental nuclear radiation monitoring site 1,and its average absolute error of prediction is 1.2517;compared with three single algorithm models,the multi-classification fusion model(MCFM)has better prediction effect on the HPIC dose rate level,and the final selection of the multi-classification fusion MCFM was selected for 10 predictions of HPIC dose rate level,and the classification accuracy was 90%.In this paper,an ARMA model-based monitoring and warning mechanism for univariate environmental nuclear radiation dose rate is established through an empirical study,and this mechanism can be used for trend prediction analysis of HPIC dose rate at a single site.A MCFM-based monitoring and warning mechanism for environmental nuclear radiation dose rate levels is also established,which is more effective for prediction analysis of multi-characteristic HPIC dose rate levels.The research in this paper provides an effective monitoring and early warning mechanism for the environmental nuclear radiation dose rate,which has certain practical guidance significance.