Stochastic Characteristics of Load Profiling in Distribution Systems based on AMR measurements

In order to enable the key applications provided by smart grid, a dynamic distribution system state estimator has to function properly giving the real time knowledge of a distribution network load profiles. However, due to unavailability of enough measurements and lagging communication of the measured data, the current system cannot readily access measurements in real time. Therefore a practical load modeling technique has to be applied to provide real time distribution system load profile estimates with the respective confidence intervals. Nevertheless, modeling individual customer loads is far statistically complicated and computationally prohibitive for real time application. The customer loads are not always normally distributed but rather a combination of normal distribution, lognormal distribution, generalized pareto and many others. In this study, models of primary heating type load classes are used to represent individual customers. A statistical formulation based on clustered customer groups’ hourly consumption model is used to calculate aggregated substation loads and day ahead load forecasts with the respective confidence intervals. The stochasticity of customer loads when aggregated at distribution substation is investigated for four customer type load classes.