FORECASTING SHORT-TIME LOAD DEMAND BY NON-LINEAR AUTOREGRESSIVE WITH EXTERNAL INPUTS

Abstract

Accurate estimation of the load demand in each period could be a useful approach to implement necessary policies for better management in the load supply-consumption chain. Hourly load demand is a function of external parameters (such as calendar data) and internal parameters (load demand). Nonlinear Autoregressive with External Inputs (NARX) was selected because of the ability to use external and internal inputs. The model was trained and evaluated using the hourly load demand data for Shahrekord city in a four-year period. The novelties of this study include (i) identifying essential factors that have a significant effect on load demand and (ii) Reduce model inputs (parsimonious model). By analysis of variance (ANOVA), external factors were selected to increase the accuracy of the model in addition to simplicity. External inputs of the developed model included calendar and holiday data. Unlike other existing models, the number of inputs were reduced but the days classification was increased from two groups (working day and holiday) to three groups (working day, semi-holiday and holiday). These changes caused, in addition to increasing the accuracy of the model, the number of inputs to be reduced. The mean absolute percentage error average (MAPE) is 0.983%, which shows higher accuracy than similar models.