Prediction of Critical Processes in Nuclear Power Plant Using Genetically Trained Neural Networks

Neural network is one of many models used in power engineering process prediction. In most cases, the accuracy of prediction models is critical in operational safety or is used to support human irreversible decisions. We use neural network, when the real model of process is unknown, or it is too difficult to identify them in domain specific environment. Neural networks training algorithms are different. Typically, measured data are divided into 2 sets called train and test set. On the train set, algorithms set neural network parameters so that network simulate process on train interval. On the test set is network tested if it can generalize the process from train set. We take a look on special genetic training and compare it with algorithm used today to generalize the process.