Spatial pattern recognition of urban sprawl using a geographically weighted regression for spatial electric load forecasting

Distribution utilities must perform forecasts in spatial manner to determine the locations that could increase their electric demand. In general, these forecasts are made in the urban area, without regard to the preferences of the inhabitants to develop its activities outside the city boundary. This may lead to errors in decision making of the distribution network expansion planning. In order to identify such preferences, this paper presents a geographically weighted regression that explore spatial patterns to determines the probability of rural regions become urban zones, as part of the urban sprawl. The proposed method is applied in a Brazilian midsize city, showing that the use of the calculated probabilities decreases the global error of spatial load forecasting in 6.5% of the load growth.