Search

During severe nuclear accidents, radioactive materials are expected to be released into the atmosphere. Estimating the source term plays a significant role in assessing the consequences of an accident to assist in actioning a proper emergency response. However, it is difficult to obtain information on the source term directly through the instruments in the reactor because of the unpredictable conditions induced by the accident. In this study, a deep learning-based method to estimate the source term with field environmental monitoring data, which utilizes the bagging method to fuse models based on the temporal convolutional network (TCN) and two-dimensional convolutional neural network (2D-CNN), was developed. To reduce the complexity of the model, the particle swarm optimization algorithm was used to optimize the parameters in the fusion model. Seven typical radionuclides (Kr-88, I-131, Te-132, Xe-133, Cs-137, Ba-140, and Ce-144) were set as mixed source terms, and the International Radiological Assessment System was used to generate model training data. The results indicated that the average prediction error of the fusion model for the seven nuclides in the test set was less than 10%, which significantly improved the estimation accuracy compared with the results obtained by TCN or 2D-CNN. Noise analysis revealed the fusion model to be robust, having potential applicability toward more complex nuclear accident scenarios. [ABSTRACT FROM AUTHOR]

During a nuclear accident, estimating the source terms using environmental measurements is vital for emergency decision-making. In this study, we propose a forecasting model based on a temporal convolutional network to estimate the release rates of seven radionuclides (Kr-88, Te-132, I-131, Xe-133, Cs-137, Ba-140, and Ce-144) based on off-site sequential gamma dose rates and meteorological monitoring data. To determine the best structure of the neural network, Bayesian optimization and hyperband (BOHB) was used on the hyperparameters of the model to reduce the testing loss. Additionally, a gradient boosting regression model was used to predict missing gamma dose rates to ensure the model offers a relatively reliable estimate under certain circumstances. The international radiological assessment system (InterRAS) was used to generate datasets for model training and testing. The results showed that the optimal hyperparameters selected by BOHB can reduce the valid loss of the model to 0.0153, and the mean absolute percentage error of prediction for the seven radionuclides was below 12%, three of which (Kr-88, Te-132, Cs-137) reached 8% at 10 h. When the first and second time-steps of the data were missing, the mean absolute percentage error of the prediction for all radionuclides was less than 30% after using a gradient boosting regression. • Improving the estimation accuracy of multi-nuclide source term estimation method based on Temporal Convolutional Network. • Use Bayesian optimization and hyperband to search the optimal hyperparameters of the model. • Use a gradient boosting regression to complete the missing data to obtain a more accurate estimation. • Fast response and does not require priori information. • The mean absolute percentage error of prediction for the seven radionuclides below 12%. [ABSTRACT FROM AUTHOR]

When severe nuclear accidents at nuclear power plants release radioactive material into the atmosphere, the source term information is typically unknown. Estimating the emission rate of radionuclides is essential to assess the consequences of the accident before adequate decision-making can be performed. A recurrent neural network-based model, optimized with the Bayesian method, is proposed to estimate the emission rates of multi-nuclides using off-site sequential gamma dose rate monitoring data. Compared with the existing method that is based on least squares, this new model does not require a priori information and the complicated and time-consuming process of conducting atmospheric dispersion simulations following a nuclear accident, which is conducive to a faster response. Six typical radionuclides (Sr-91, La-140, Te-132, Xe-133, I-131, and Cs-137) were set as mixed source terms, combined with meteorological parameters, and input into the International Radiological Assessment System for simulation to generate data sets for model training. The results indicate that with the input of data describing the sequential gamma dose rate, the accuracy of the nuclide emission rates estimated by this new method is continuously improved, with a mean absolute percentage error for Te-132 of below 7% over 10 h. [Display omitted] • Source term estimation method based on recurrent neural network. • Estimate the emission rate of 6 radionuclides using gamma dose rates. • Use Bayesian optimization to search the optimal structure of the model. • Fast response and does not require priori information. • Mean absolute percentage error of estimated results for Te-132 below 7%. [ABSTRACT FROM AUTHOR]

• Build an estimation model based on Backpropagation Neural Networks for the release of I-131. • Dimensionality reduction of original data using Kernel Principal Component Analysis. • Optimize Neural Network using Particle Swarm Optimization. Rapid estimation of the release rate of source items after a nuclear accident is very important for nuclear emergency and decision making. A source term estimation method, based on the Backpropagation Neural Network (BPNN), was developed. Kernel Principal Component Analysis (KPCA) is used to reduce the dimension of the input parameters, which can accelerate the training of the neural network. Particle Swarm Optimization (PSO) is used to optimize weights and thresholds of BPNN, so that the neural network can better find the global optimal value, avoid falling into the local minimum. The large amount of data required for neural network training is generated using InterRAS software, the model constructed demonstrates the feasibility of this method. The proposed method can estimate the release rate of I-131 after half an hour of release, which is helpful to the emergency response, or provide an initial value or priori information for other methods. [ABSTRACT FROM AUTHOR]

• Construct a method to determine the release category using backpropagation neural network (BPNN) classifier. • Optimise Neural Network using Genetic algorithm (GA) and simulated annealing (SA). • Complete the missing monitoring data using Particle swarm optimisation (PSO). After a severe nuclear accident, the source term is typically unknown. Therefore, great importance is attached to obtaining source term information for subsequent emergency planning. The purpose of the study involved performing a rapid determination of the release category using the gamma dose rate monitoring data over a short period during an accident. The release categories PWR1–PWR9 in 'The Reactor Safety Study' of the United States were used as a reference. The International Radiological Assessment System (InterRAS) was used to construct a nuclear accident model and generate the required simulation data. After a series of experiments, appropriate parameters were selected to construct the backpropagation neural network (BPNN) classifier to estimate the release category. The genetic algorithm (GA) and simulated annealing (SA) algorithm were used to search for the weights and thresholds of the BPNN classifier in advance, and this avoided the problem wherein bad initial values can cause the classifier to fall into local minimums, decreased training time, and improved prediction accuracy from 98.36% to 99.10%. With respect to the possible absence of gamma dose rate monitoring data, particle swarm optimisation (PSO) was used to complete the missing data, thereby ensuring that the classifier can normally predict the release category. After testing, in the absence of 4 of the total 16 gamma dose rate data, the classifier can still maintain a prediction accuracy of more than 80%. [ABSTRACT FROM AUTHOR]