Comparison of regression and neural network approaches to forecast daily power consumption

Forecasting of power consumption and planning of the balances of electric power may be said to be the main objective of management of EPS. The amount of energy consumption defines the structure of generating equipment, electric networks configuration, the production of electric and thermal energy, use of energy resources, reliability of power supply, the quality of electric power, and also plays an important role in pricing. Nowadays there is a set of methods and models of forecasting of power consumption including the following time ranges of management: quick (daily range), short-term (monthly range) and long-term (annual range). The planning accuracy is an actual task and it depends on calculation methods. This work shows a comparative analysis of regressive and neural network models for the solution of a problem of forecasting of daily power consumption. The power consumption of Siberia UPS over 5 years was calculated during the experiment. The problems of the models accuracy and adequacy of using them in forecasting are shown in the work.