Neuro-Fuzzy Approach Based Short Term Electric Load Forecastig

Abstract

The basic problem for electric utilities and system operators is to maximize the short and long term operations system performance given the operating and economic characteristics generating units, the transmission line constraints and the limited amounts of the capital available for new units and equipments. Load forecasting has been a central and integral process in the planning and operation of electric utilities. Load forecasting helps the system operators to schedule spinning reserves allocation efficiently and also in for power system security. If applied to system security assessment problem, valuable information can be obtained in order to detect vulnerable situation in advance. The short term load forecast is required unit commitment, energy transfer scheduling and load dispatch. With emergency of load management strategies, the short term load forecast is playing a broader role in utility operations. The development of an accurate, fast and robust short term load forecasting methodology is of importance to both electric utility and its customers, thus introducing higher accuracy requirements This paper makes use of two approaches, the first being artificial neural network using back propagation algorithm (BP), second neuro-fuzzy hybrid system. The results obtained from two approaches show the superiority of the neuro-fuzzy hybrid system over the other one. This case study has been performed on the load and weather data pertaining to the Neo pool region (New England) for the year 2003