Application of a rule-based technique to weekly load forecast

The application of rule-based techniques to a weekly (168 h) load forecast is discussed. A rule-based algorithm is developed that is capable of issuing a 168 h lead-time load forecast. The algorithm is written using the FORTRAN language. The database requirement for this algorithm does not extend beyond more than four weeks of hourly load and few weather variables. Results of this algorithm in its current form for the summer season using actual weather information is encouraging compared to those that could be obtained using statistical techniques. For example, the 168 h load forecast, as applied to the load data of a southeastern US utility for the full month of August 1983, has resulted in absolute average forecast error of 2.5% with a standard deviation of 2.3%.<
>