Computer forecasting of electrical loads by neural network methods

In the process of design and operation of urban electrical networks there is a problem of forecasting electrical loads for a certain time period, due to technological and economic reasons. Used in most cases to predict electrical loads method of expert assessments in real operating conditions does not provide the required accuracy of the forecast. The paper presents an approach to solving the problem of forecasting the electrical loads of the city power grid based on an artificial neural network. The possibilities of application of the methodology of neural networks in power engineering are considered, the analysis of existing problems of application of Neurocomputers in control systems of power systems is carried out. The task of forming a neural network to predict the loads of power grids. Possibility of application of systems of computer mathematics for realization of neural networks is investigated. The implementation of a neural network in the system of computer mathematics Matlab is developed. Configuration and training of the network on real source data was performed. Target and input data for creation of a neural network, results of its training are given. The test calculations of electrical loads and power consumption obtained with the help of a neural network are performed, the modeling error is calculated, the conclusion is made that the model built on the basis of a neural network is adequate.