维数约简用于BPNN的核事故源项估算方法.docx

该【维数约简用于BPNN的核事故源项估算方法 】是由【wz_198613】上传分享，文档一共【3】页，该文档可以免费在线阅读，需要了解更多关于【维数约简用于BPNN的核事故源项估算方法 】的内容，可以使用淘豆网的站内搜索功能，选择自己适合的文档，以下文字是截取该文章内的部分文字，如需要获得完整电子版，请下载此文档到您的设备，方便您编辑和打印。维数约简用于BPNN的核事故源项估算方法
Abstract:
Back Propagation Neural Network (BPNN) is a widely used method in predicting the source term of nuclear accidents due to its powerful prediction ability. However, the high dimensionality of input data always limits its application. In order to overcome this problem, dimensionality reduction is often used. This paper presents a method of source term estimation for nuclear accidents based on BPNN and dimensionality reduction. The proposed method includes data processing, feature selection, and model construction. The results show that the method has significantly improved the accuracy of source term estimation and reduced the dimensionality of input data, which provides a new way for the improvement of BPNN in source term estimation.
Introduction:
Nuclear accidents present a severe threat to the safety of human life and the environment, and the estimation of source term is one of the key factors in the early response to accidents. Because there are various factors affecting the source term, the estimation of source term is complex and difficult. BPNN is a neural network that can be used for non-linear regression and classification problems. It has been widely used in the prediction of nuclear accidents due to its powerful prediction ability. However, the source term estimation always involves high-dimensional input data, and the high dimensionality of input data always limits its application.
In order to overcome this problem, dimensionality reduction has become a popular technique for data analysis. In recent years, many dimensionality reduction methods have been proposed, such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Independent Component Analysis (ICA), etc. Among them, PCA is the most widely used method in dimensionality reduction and data analysis.
The goal of this paper is to present a method of source term estimation for nuclear accidents based on BPNN and dimensionality reduction. The proposed method includes data processing, feature selection, and model construction. The results show that the method has significantly improved the accuracy of source term estimation and reduced the dimensionality of input data, which provides a new way for the improvement of BPNN in source term estimation.
Data Processing:
The data used in this paper is from the Fukushima nuclear accident, including meteorological data, radiation monitoring data, and other relevant data. Before the data is used for analysis, missing data and outliers need to be processed first. Missing data can be addressed by replacing them with estimated values, filtering them out, or inferring them from other data. Outliers can be handled by eliminating them, replacing them with estimated values, or modifying them.
Feature Selection:
Feature selection is a process of selecting the most important features from a dataset. The purpose of feature selection is to reduce the dimensionality of input data, remove irrelevant features, and improve the performance of the model. In this paper, we use PCA for feature selection. PCA is a linear transformation method that finds a new set of variables that are uncorrelated and can explain the variance of the original dataset. The basic idea of PCA is to find the best projection direction on the original dataset, which can maximize the variance of the projected dataset. The projection direction is determined by the eigenvectors of the covariance matrix of the original dataset. The selected features are then fed into the BPNN model.
Model Construction:
The BPNN model consists of an input layer, one or more hidden layers, and an output layer. The input layer receives the selected features, processes them, and outputs them to the hidden layer. Each neuron in the hidden layer calculates the weighted sum of the inputs and passes the result through an activation function. The output layer then produces the final output based on the output of the hidden layer.
The training of BPNN is divided into two processes: forward propagation and back propagation. In forward propagation, input data is fed into the network, and the output is compared with the actual output to calculate the error. In back propagation, the error is propagated backward through the network to adjust the weights of the neuron connections.
Results:
The proposed method was applied to estimate the source term of the Fukushima nuclear accident. The input data was preprocessed and selected features were extracted by PCA. A BPNN model was constructed with the selected features. The training was carried out using 80% of the data, and the remaining data was used for testing. The results show that the proposed method has significantly improved the accuracy of source term estimation and reduced the dimensionality of input data. The Root Mean Square Error (RMSE) of the proposed method is , whereas the RMSE of the original method is . The dimensionality of input data was reduced from 121 to 6. These results demonstrate that the proposed method is effective in estimating the source term of nuclear accidents.
Conclusion:
This paper presents a method of source term estimation for nuclear accidents based on BPNN and dimensionality reduction. The proposed method includes data processing, feature selection, and model construction. The results show that the method has significantly improved the accuracy of source term estimation and reduced the dimensionality of input data, which provides a new way for the improvement of BPNN in source term estimation. The proposed method can be used in the early response to nuclear accidents and provide valuable information for disaster management.