Tracking progress towards the green energy transition: Nationwide mapping of roof-top photovoltaic installations

Kurzfassung

The decarbonization of the global energy system through the transition to renewable energy sources is one of the main pillars of mitigation measures addressing climate change. A key determinant in the realization of a successful energy transition is the rapid installation of renewable energy infrastructure, which, for the most part, is of very heterogenous nature and spatially decentralized [1]. This is particularly true for roof-top photovoltaic systems (PVs), which are often small-scale, privately owned, differ widely in their capacity and are non-randomly distributed in space, with even the occasional emergence of patterns of socio-economic and political boundary conditions. In order to foster the rapid expansion of PV installations, which is needed for reaching aims and commitments with respect to the energy transition, an automatically updatable monitoring of the evolution of such PV installations over time can be a vital asset for political decision makers, both on the communal and national level, supporting the efficient allocation of resources. Here, we demonstrate a system for nationwide mapping of rooftop PV installations based on multiple timesteps of aerial imagery for all of Germany. We demonstrate our system with an exemplary wall-to-wall analysis, where we exploited publicly available registry data as well as manually labelled samples for training data collection. Based on this curated training data set of around 350.000 samples, we trained ensembles of supervised, state-of-the-art deep neural networks (ResNets, ResNests, EfficientNets, ConvNexts and VisionTransformers) for predicting the presence of PV installations for each building in Germany. Following a rigorous validation exercise based on over 20.000 samples, we report a very high predictive performance of 0.96% F1-score overall, with a regional variability of +/- 0.03. Going beyond single date classifications, this system enables us to track the growth of PV installations over time throughout Germany. Add-on analyses of installed PVs in combination with spatially explicit solar potential models [2] allow us now to identify and suggest priority areas for high-return-on-investment policy stimulation for fostering the much-needed growth of decentralized PV installations. References: [1] M. Victoria, K. Zhu, T. Brown, G. B. Andresen, and M. Greiner, 'Early decarbonisation of the european energy system pays off' Nature Communications, vol. 11, no. 1, 2020. [2] S. Joshi, S. Mittal, P. Holloway, P. R. Shukla, B. Ó. Gallachóir, and J. Glynn, 'High resolution global spatiotemporal assessment of rooftop solar photovoltaics potential for renewable electricity generation' Nature Communications, vol. 12, no. 1, 2021.