Atypical VRE Variability in Power System Planning and Performance: Illustrated by the case studies of Guinea-Bissau and Turkey

Abstract: This study aims to identify and quantify (1) how Typical Meteorological Year (TMY)-based power systems perform when exposed to atypical variability and (2) how TMY-centric power system design differs from full variability design. A simplified least cost power system planning model, including a novel performance analysis formulation, is proposed and tested for Guinea-Bissau and Turkey, covering both main Variable Renewable Energy (VRE) technologies, solar PV and wind turbines. TMY is compared against 24-year timeseries datasets containing hourly resolution solar PV and wind capacity factor (CF) data. This study finds that TMY-based power system designs underperform when exposed to atypical variability. Over the entire dataset, foregone VRE generation and additional expenses approximate 36 GWh and 10 million USD for Guinea-Bissau, and 92 GWh and 232,000 million TL for Turkey. Moreover, Turkey faces unmet demand, amounting up to 50 TWh. Likewise, power system design significantly differs when including atypical variability, illustrating how TMY-centric design underestimates non-VRE capacity and overestimates VRE capacity.  In the case of Guinea-Bissau, solar contingency was tested in combination with atypical variability, and harvested the complete exclusion of solar PV from the system. This illustrates the effect of factors such as atypical variability and contingency measures on power system design, especially considering the high solar availability in the country combined with low solar PV costs. Moreover, including batteries as spinning reserve providers, is found to reduce overall system costs and under certain circumstances increase the selection of solar PV. Besides addressing these two main questions and the additional stability measure study, this document proposes two new data reduction methods to account for both typical and atypical variability, without adding significant computational costs. The first, labelled the ‘incremental year analysis’ was found to be a proxy to estimate the amount of years required to reach an optimal design. The second, ‘power system performance analysis’, in this study reduces the data size at least by half. Moreover, this approach allows power system planners to further reduce the dataset size when including low underperformance tolerance levels and enables the identification of extreme underperforming years. These approaches do not solve the issue and are time intensive. However, they can complement the existing TMY approach by including atypical variability whilst minimizing computational costs.   Thereby, this study concludes that atypical variability impacts power system design and performance, and hence should be included in long-term planning methodologies. Concurrently, this research acknowledges the need for solutions that can circumvent the need for additional, often costly computational resources.