Short term load forecasting using support vector machines for different consumer types

Abstract

Deregulated electric energy market pushes utilities to control and maintain production, transmission and distribution system on more economical way. Economical control focuses accurate electric energy forecast. This paper presents short term load forecasting method using support vector machines. Short term load forecasting predicts load in range from one to seven days. Support vector machines is new artificial intelligence methods from family of supervised learning methods and it is successfully used for pattern recognition, classification, regression, forecast and more. Proposed method uses historical data of consumer behavior, historical, actual and forecasted weather data and day type for inputs. Results are shown for three consumer types, residential with distant heating, residential with heating on electric energy and industrial type. Proposed methodology obtained good results in the area of electric power consumption prediction at distribution level. Also, this methodology is useful for prediction of generation and consumption of other types of energy from different sources.

Keywords

• short term load forecasting
• support vector machines