Privacy-Preserving Behind-the-Meter PV-Load Disaggregation via Prior Knowledge Constrained Blind Source Separation

Accurate disaggregation of behind-the-meter (BTM) photovoltaic (PV) generation and load demand (PV-load) is essential for utility operators to make informed decisions (e.g., demand response). Existing methods for BTM PV-load disaggregation commonly rely on the availability of detailed information, such as observations of neighboring end users’ PV generation or load demand and geographical location, physical parameters of the PV panel and weather conditions, or labeled historical disaggregated observations. However, such information is often unavailable in practice due to privacy concerns or high economic costs for monitoring devices. Instead, this paper proposes a privacy-preserving BTM PV-load disaggregation algorithm that only requires the observed net load data based on prior knowledge constrained blind source separation. Specifically, ensemble empirical mode decomposition is first used to enhance the signal dimension of the observed net load, and then prior knowledge constrained independent component analysis is applied to disaggregate PV generation and load demand from the enhanced mixed signal. Furthermore, the demand response potential of the BTM PV-load unit is evaluated based on the disaggregation results. Numerical experiments on a real-world dataset demonstrate the effectiveness and efficiency of the proposed algorithm.