Using neural networks to forecast the electric power consumption
AbstractIn this article, the state of the electrical grid in Ukraine and abroad is briefly described. The urgency of creating a new generation networks is explained. The formulation of the Smart Grid control problem is presented as well as other subtasks which occur during the problem solving process. The paper provides a solution to the problem of forecasting electric power using neural networks. The neural network structure and selection of the input parameters as well as an algorithm for network training are described to produce daily forecasts of electricity consumption. The input data, output data, and feature selection algorithm are described. Authors studied the dependence between the accuracy of prediction and the choice of input features. An interesting fact was revealed that when holidays data were added as separate features indicators to the network input, the quality of the forecast results could be improved. Also, the effect of weather conditions on the accuracy of the forecasts is shown.
Fan Z., Kulkarni P., Gormus S. Smart Grid Communications: Overview of Research Challenges, Solutions, and Standardization Activities. — http://arxiv.org/abs/1112.3516.
Galli S., Scaglione A., Wang Z. For the Grid and Through the Grid: The Role of Power Line Communications in the Smart Grid. — http://arxiv.org/abs/1010.1973.
Sayt DP "Enerhorynok". — http://er.gov.ua.
Lin J. Strategen, Advanced Energy Storage, i4Energy seminar: i4Energy Center Initiative. — http://www.youtube.com/watch?v=h2iyQcXQUSw.
A hierarchical load forecasting problem: backcasting and forecasting hourly loads (in kW) for a US utility with 20 zones. — http://www.kaggle.com/c/globalenergy-forecasting-competition-2012-load-forecasting.