Development of real-time core monitoring system models with accuracy-enhanced neural networks

Core monitoring models using neural networks have been developed for prediction of the core parameters for pressurized water reactors. The neural network model has been shown to be successful for the conservative and accurate prediction of the departure from nucleate boiling ratio (DNBR). Several variations of the neural network technique have been proposed and compared based on numerical experiments. The neural network can be augmented by use of a functional link to improve the performance of the network model. Use of twofold weight sets or weighted system error backpropagation was very effective for improving the network model accuracy further. An uncertainty factor that is a function of output DNBR is used to obtain a conservative DNBR for actual applications. The predictions by the network model need to be supported by extensive training of network and statistical treatment of the data. Studies for further improvements are suggested for future applications