Concept of Intelligent Energy Grid Control Based on Upgraded Neural Network

The development of electric power systems requires an improvement in their operation quality, which, among other things, depends to a large extent on the possibility of predicting the electric load of the power system. Electrical load forecasting provides basic baseline information for decision making in power system control. Previously used single-factor time series forecasts based on regression methods are unable to take into account the totality of all factors affecting on electricity consumption, such as the weather forecast for the coming day; duration of light day; possibility of atmospheric precipitation or clouds, season, schedule of electric loads; condition of wear of power supply networks; previously recorded irregularities in the operation of the electric grid complex, the operation mode of the largest consumers; solvency of clients, etc. The need to use heterogeneous quantitative and qualitative parameters makes artificial intelligence use, neural network and expedient. The features of the approach proposed by the authors are the consideration in the neural network of the influence of a pair of neurons on the neuron of the intermediate layer and the dynamic change in the number of input parameters.