An unsupervised data-driven approach for behind-the-meter photovoltaic power generation disaggregation

Abstract

An increasing number of behind-the-meter (BtM) rooftop photovoltaic (PV) panels is being installed and maintained by site owners. However, invisible PV power generation (PVPG) will lead to the difficulty for system operators in power dispatch and affect the safety and stability of the power system. To better quantify BtM PVPG, a novel unsupervised data-driven disaggregation method freedom from PV system physical model assumption for BtM PVPG is proposed. After clustering the prosumers’ net load curves, a PVPG sensitivity estimation model is firstly built, based on the net load with approximate energy consumption (EC) and the corresponding irradiation data obtained from the pairing date. Then, an EC sensitivity model is developed according to the net load and temperature of the date with similar irradiation. Finally, a new net load disaggregation model is constructed by the PVPG sensitivity model with EC compensation. Case study based on Ausgrid data shows that the proposed method provides a better quality BtM PVPG disaggregation. The disaggregation accuracy improves by 5.06–5.87% as compared to the state-of-the-art methods.