Short Term Load Forecast In Jimma City By Using Artificial Neural Network

Abstract

For optimal power system operation, electrical generation must follow electrical load demand. The generation, transmission, and distribution utilities require forecasting the electrical load so they can utilize their electrical infrastructure efficiently, securely, and economically. The short-term load forecast (STLF) represents the electric load forecast for a time interval of a few hours to a few days. This thesis is a study of short-term electric power forecasting in the jimma power system using artificial neural network model. The model is created in the form of a simulation program written with MATLAB tool. The model, a feed forward neural network, for radial basis neural network and recurrent current artificial neural network trained with error, was made to study the pre-historical load pattern of a typical jimma power system in a supervised training manner. After presenting the model with a reasonable number of training samples, the model could forecast correctly electric power supply in the jimma power system 24 hours in advance. An absolute mean error was obtained and compares three neural networks feed forward neural network 0.5180 to 6.3868, for radial basis neural network 0.0861 to 2.8703 and recurrent current 0.2811 to 13.8851 from this choose the least absolute mean error radial basis neural network 0.0861 to 2.8703. The trained neural network model was tested on one week, daily hourly load data of a typical jimma power station. This result demonstrates that ANN is a powerful tool for load forecasting. One week (winter Monday 22/9/07 – Sunday 28/9/07) , One week (Summer Monday 25/12/07 – Sunday 1/13/07) and One day (Holiday Wednesday 1/1/08) of electrical load. Load data was recorded for JIMMA CITY, so there are 15 days of data collected.