Improving solar power forecasting to reduce regulation frequency control ancillary services causer pay in the national electricity market

Abstract

Driven by many factors including sharply declining costs and the pressing need for climate change mitigation, in particular the Large-scale Renewable Energy Target (LRET) in Australia, the installation of renewable energy generation such as wind and solar is rapidly increasing all over the world. Since 2015, more and more utility-scale (>30 MWp) solar farms are being commissioned and connected to Australia's National Energy Market (NEM). However, the power generation from wind and solar resources relies on complex weather and climate processes and is inherently variable. Together with other uncertainties such as load forecasting errors, this could result in generation-load imbalance, which in turn causes fluctuations of AC frequencies in the grid. Since all electric equipment connected to the grid (e.g. household appliances and steam turbines) is designed to operate at or close to 50 Hz in Australia, frequency control is critically important to maintaining a secure and reliable power system. To address this issue, the Australian Energy Market Operator (AEMO) frequently corrects the generation-load imbalance via a 4-second market mechanism to restore system frequency back to the nominal 50 Hz through the procurement of regulating Frequency Control Ancillary Services (FCAS). Cost for this market are recovered under a mechanism known as 'causer pays', whereby a grid-connected solar farm is liable for the part of costs according to its estimated contribution to the need for regulating the grid frequency. However, although it is widely recognised that the quality of power forecasts of renewable generators relates to the need (hence the associated costs) for frequency regulation to some extent, a quantitative model of the relationship remains unavailable. In this study, we use high-temporal-resolution generation and forecast data published by the AEMO to reveal the functional relationship between causer-pay contribution factors which are proportional to actual payments assigned to individual solar farms and the quality of solar power forecasts issued by the Australian Solar Energy Forecasting System (ASEFS). We found that the contribution factors are largely due to the coincidence of a positive forecasting bias and a low system frequency. We then manage to effectively model the contribution factor of a solar farm using information only for that solar farm.