Short-range load forecasting for distribution system planning-an improved method for extrapolating feeder load growth

A new method of extrapolating feeder peak load histories to produce estimates of future feeder loads is described. The method, an improvement on past multiple regression curve fit methods, uses an assumed geometry based on substation locations and a classification by recent growth rates to group feeders into six classes, each extrapolated in a slightly different manner. The new method is simple enough to be applied in situations where computing resources are limited. A series of tests shows that the new method outperforms other distribution load extrapolation methods, and that for short range (less than five years ahead) forecasts, it matches the accuracy of simulation forecasting methods, which require considerably more data and computer resources