Quantifying the Risk When Using Single Year PV Data in South African Electricity System Models

Numerous studies have been conducted where a single year of photovoltaic (PV) electricity production data is used to study the impact that PV electricity production has on electricity systems. In this paper the potential margin of error when using a single year of PV electricity production data to approximate long-term PV electricity production is analyzed in the South African context, both for individual PV plant electricity production locations and for PV production aggregated across multiple locations regionally or country-wide. The ability to accurately represent mean annual production, mean seasonal production and the minimum and maximum production for 1-,7- and thirty-day periods is investigated. It is found that while a single year of data performs poorly in general, it does give reasonable estimates of annual electricity production in areas with very high annual average Global Horizontal Irradiation. In terms of spatially dispersed PV electricity production, it is found that combining PV production from various randomly selected locations, whether dispersed across the country or more geographically constrained, does not significantly increase accuracy from what is achieved using individual locations when using a single year to represent long-term electricity production. Additionally, it is shown that using a single year of data is likely to overestimate short term production from low-irradiation periods and underestimate short term production from high irradiation periods. Thus, in cases where studies using a single year of PV production data have identified phenomena related to periods of high or low production, such as curtailment or unserved energy, these phenomena are likely become more prevalent when multi-year datasets are considered. These findings are relevant to researchers scoping future PV related research projects and to the analysis of results from studies that have used a single year of PV production data as input.