End-to-end learning of representative PV capacity factors from aggregated PV feed-ins

Energy system models rely on accurate weather information to capture spatio-temporal characteristics of renewable energy generation. Whereas energy system models are often solved with high abstraction of the actual energy system, meteorological data from reanalysis or satellites provides rich gridded information of the weather. The mapping from meteorological data to renewable energy generation usually relies on major assumptions as for solar photovoltaic energy the photovoltaic module parameters. In this study, we show that these assumptions lead to large deviations between reported and estimated energy as shown in case of photovoltaic energy feed-ins in Germany. To decrease these deviations, we propose a novel gradient-based end-to-end framework which is able to learn local representative photovoltaic capacity factors from aggregated reported transmission system operator feed-ins. As part of the end-to-end framework, we compare physical and neural network model formulations to obtain a functional mapping from meteorological data to photovoltaic capacity factors. We show that all developed methods have better performance than commonly used reference methods. The neural network shows remarkable success to predict the aggregated Transmission System Operator photovoltaic energy feed-ins leading to an accurate, unbiased prediction model. However, choosing the neural network model is not always the strictly preferred choice as it depends on the use case: Operational use cases may decide for the neural network implementation due to its higher accuracy whereas academic settings prefer the physical model due to its high interpretability and transferability. In this talk, we discuss the development of the end-to-end framework which we believe is highly relevant due to its energy-meteorological and methodological contributions.