Load forecasting at distribution transformer using IoT based smart meter data from 6000 Irish homes

Energy Consumption in a neighborhood depends upon its socioeconomic parameters. Demographical diversities in a neighborhood in India warrants load prediction at distribution transformer (DT) rather than at utility level. In this paper two interesting techniques of load forecasting have been proposed which have not be explored till date. In both these technique a unique pattern of consumption has been deciphered for a day using parametric estimation and subsequently regression, neural network and support vector regression have been used to find the total consumption of the day which is subsequently redistributed according to pattern of the day to deduce final load pattern. In the first technique a unique model has been created for each day of the week. Though the results have been very encouraging with average error of 12% however it is not sufficient for many applications. In the second approach a set of model is created for the entire year and depending upon the previous pattern. A particular model having correlation more than 95% and similar total consumption is selected out of these models. In this case mean error has been reported as approximately 7%. Neural network considers all factors affecting the consumption and hence its corresponding predictions have been found more accurate.