Data-driven Assessment of Aggregate Distributed Photovoltaic Generation and its Impacts on Electricity Network Planning

Abstract

Photovoltaics (PV) has emerged as a globally significant source of electricity generation over the past decade. Distributed PV systems in particular have significant implications for electricity network planning and operation. Assessing the aggregate impact of distributed PV systems is challenging due to their inherently variable and somewhat unpredictable operational characteristics, and the wide range of installation conditions found on residential, commercial or industrial rooftops. In this thesis, a set of techniques is proposed to improve the quality of data and metadata from these PV systems, and to estimate their aggregate generation. These techniques are applied to the historical output data of some five thousand distributed PV systems, and used to assess the impact of distributed PV in Australia’s transmission and distribution networks. A new method to estimate or correct missing or invalid metadata (location and orientation) of the PV systems is introduced. The operational performance of distributed PV systems is then evaluated and compared to the publicly available estimates for PV generation in Australia, highlighting the value of data-driven approaches for better understanding the real operation and hence impact of distributed PV systems. A model for upscaling the output of a limited number of monitored PV systems to estimate the output of all PV systems in a region is also proposed. This method is applied to assess the impact of current and potential higher penetrations of PV systems in different zone substations of a large distribution network, and zone substation load profiles are clustered based on the availability of the PV in their peak times as a means to characterise the inter-annual and spatial variability of this impact in different distribution feeders. Finally, the impact of current and higher penetrations of distributed PV on peak demand and the time of the peak in the different regions of Australia’s National Energy Market (NEM) is evaluated. This thesis therefore contributes a framework for better managing and using data from distributed PV systems to estimate the output of distributed PV systems for network planning and operation. Many of these techniques will also have broader applications for integrating distributed PV, for instance into generation investment and system operation decisions.