Photovoltaic Power Disaggregation using a Non-Intrusive Load Monitoring Regression Model

Non-intrusive load monitoring (NILM) techniques represent an opportunity to increase the flexibility and resilience of the electrical system. However, there is limited research and practical work focused on implementing these methods for distributed energy resources on distribution networks. Non-intrusive load monitoring can contribute to demand-side management strategies required for real-time operation of smart grids as well as reducing the negative impacts of distributed generation and low carbon technologies in low voltage networks. In this paper, a unique supervised machine learning algorithm to disaggregate electric power generation of photovoltaic systems from the main power feeder of a residential dwelling is proposed. The algorithm uses low complexity sliding windows and conventional machine learning techniques applied to a real residential households' data obtained from the dataset of the Smart* project. The results exhibit an accurate performance of the proposed NILM in fast computation time with a mean average error below 5.2%.