Disaggregation of Behind-the-Meter Solar Generation in Presence of Energy Storage Resources

Residential rooftop photovoltaics are commonly installed with energy storage in the form of batteries for the purpose of storing excess energy generated for use when extra energy is needed. Recently, there has been a lot of interest in improving the observability of behind-the-meter solar for better grid optimization strategies. However, existing algorithms only consider standalone photovoltaic systems not connected to a battery. Presence of battery changes the load pattern recorded from the households, thus these algorithms are not able to perform disaggregation accurately when batteries are present. We propose an improved disaggregation method for behind-the-meter solar that takes into account the charging and discharging activity of batteries by including the hidden-battery model. We compare our methods with a state-of-the-art temperature based load model for disaggregation and evaluate the methodologies on simulated data based on inputs from a real life dataset based in Austin, Texas. Our results show that hidden-battery model reduces mean absolute error by 28% on an existing method on aggregated customer data.