Artificial Neural Network Short Term Load Forecaster

هدف المشروع إلى تطوير وبناء نظام تجريبي للتنبؤ بالحمل الكهربائي على المدى القصير بواسطة الشبكات العصبية الاصطناعية لمركز التحكم بوزارة الكهرباء والماء، حيث أن واحدة من السلبيات في النظام المستخدم الحالي هي عدم القدرة على تقدير بعض المؤثرات الغير قابلة للإحصاء والتي لها علاقة مباشرة مع الحمل الكهربائي.

The quality of short term load forecasts with lead times ranging from one hour to several days ahead has a significant impact on the efficiency of the operation of the Ministry of Electricity and Water (MEW) production and transmission of electric power. The goals of this project are to develop, implement and evaluate an off-line Artificial Neural Network (ANN) based short-term forecaster that supports the MEW load forecasting system by enhancing its prediction accuracy rate. One of the main limitations of the MEW load forecasting system is the incapability to account for some of the variables that affect the electric load volume. In a recent research activity at KISR, a self-organizing map (SOM) ANN to estimate the effects of these non-measurable input variables was developed. Moreover, sets of feed-forward multi-layer perceptrons (MLPs) ANNs were developed to predict the day-ahead electric load. SOM and MLP networks are trained and tested using the data from 1998 and 1999. Testing results show that ANN based forecaster is able to reduce the absolute average forecasting error from 6.27% to 1.32%. These results will be used as a base for the development of the proposed forecaster.