Short-Term Load Forecasting for Baghdad Electricity Region

Abstract The problem of short-term load forecasting that is the prediction of the system load over an interval of time from 1 hour to 1 week is studied. An adaptive model to forecast hourly electric demand with a lead time of 1 to 24 hours based on records of hourly load and weather information is introduced. The predicted system load is divided into four components: daily, weekly, weather sensitive, and bias correction. The significant weather variables which influence the weather component are identified. Manual detection of bad data during load shedding and feeder disconnections are done. Correction for anomalous load behavior during holidays and special days in the data base is proposed. Results obtained by the application of this model to the historical load and weather data of Baghdad Electricity Region were very close to the actual loads.