Energy consumption modelling using neural networks of direct distribution on example of Russia united power system

The article describes a model to estimate an electrical energy consumption on the basis of neural network of direct distribution. The model is tested on actual hourly data of both United energy system of Wholesale electricity market and power of Russia. An algorithm to train a neural network with different numbers of neurons in the hidden layer is described. We tested the obtained model and find that a forecast error is 2.13 % for a network with 72 neurons in the hidden layer. The designed scientific instrument is recommended in operating activities of electric power subjects, when main parameters of energy market are forecasted in order to reduce the penalties by improving the accuracy of forecasts.