COMBINATION OF HUMAN AND MACHINE-BASED DEMAND FORECASTS

Computer generated forecasts, manual predictions, forecasts of princip al turning points of the load curve and component day forecasts are ma jor basic load forecast elements available to system control operators . An important requirement of the operational facility is that its hum an-machine interface should provide a clear structure of information a nd be linked to an algorithm which combines various on- and off-line p redictions to generate a final best estimate demand forecast. In parti cular, through this association of rule and knowledge based human inte rference with the data driven on-line methods, the difficulties in for ecasting special events demands can be overcome, where the problems of sudden load pattern change and insufficient data often lead to unsati sfactory performance of time series based data extrapolations. This pa per presents a method which combines manually entered (off-line) with automatically generated (on-line) forecasts; its ability to forecast d emand during special periods is demonstrated. The method has been impl emented in the real time environment and links directly to other progr ams which calculate unit dispatch schedules.