Software design and evaluation of a microcomputer-based automated load forecasting system

A discussion is presented of the software design and evaluation of a microcomputer-based automated load forecasting system. This system is based on an IBM RT/PC microcomputer, and Extended Fortran, Ratfor, and C languages. The online load and weather data collection, filtering, archiving, and load forecasting are totally automated. The forecast is based on the relationship between weather and electric utility load (MW) as extracted from historical data. 'If. . .then' types of rules and other statistical relationships form the rule-base of the algorithm. This rule-base has been applied to generate the hourly system load forecasts for 1983 and 1986 in the Virginia Power service area. A 24 hour load forecast takes 20 seconds on the IBM-RT/PC. The absolute annual averages of the hourly load forecasts errors are found to be 1.437% and 1.298% for 1983 and 1986, respectively. Out of these two years, there were only nine occasions when the software failed to predict the hour of the daily peak. Moreover, these peaks were missed by only one hour.<
>