Artificial-Intelligence-Based Methodology for Load Disaggregation at Bulk Supply Point

Abstract

Real-time load composition knowledge will dramatically benefit demand-side management (DSM). Previous works disaggregate the load via either intrusive or nonintrusive load monitoring. However, due to the difficulty in accessing all houses via smart meters at all times and the unavailability of frequently measured high-resolution load signatures at bulk supply points, neither is suitable for frequent or widespread application. This paper employs the artificial intelligence (AI) tool to develop a load disaggregation approach for bulk supply points based on the substation rms measurement without relying on smart meter data, customer surveys, or high-resolution load signatures. Monte Carlo simulation is used to generate the training and validation data. Load compositions obtained by the AI tool are compared with the validation data and used for load characteristics estimation and validation. Probabilistic distributions and confidence levels of different confidence intervals for errors of load compositions and load characteristics are also derived.