Comparing Short-Term Net Load Forecasting Methods for Solar Homes

As the prevalence of residential PV systems rises, operational decision making with relation to these resources becomes increasingly important. Forecasting is one of the most important methods of informing these decisions and while both short-term load and PV generation forecasting have been extensively studied, net load forecasting for solar homes features far less in the literature on the topic. This is despite the importance of aggregated load and PV behaviour for both homeowners and network operators. This paper provides a comparison of net load forecasting techniques across three categories: classical statistical techniques, artificial intelligence, and configurable forecasting tools, looking at both integrated and additive net load forecasting. It shows that the Prophet tool is a promising candidate for short-term net load forecasting of solar homes despite it being omitted from similar studies in literature to date, with a 20% performance improvement over daily average profiles in this study. It should be noted that forecasting accuracy of a particular method is highly dependent on the underlying dataset.