For the transition to fully renewable energy the residual load – the portion of the load that is not yet generated from renewables – has to be replaced with renewable electricity. Photovoltaics (PV) is of particular interest given possible low electricity costs. However, firm electricity from PV is still expensive due to intermittency, in particular seasonally. We describe and apply a two-part non-linear optimization method. First, optimal percentages of solar generation capacity at different sites are determined to closely approximate a given load pattern. Results are best when sites on both hemispheres and in many time zones are combined. The second method utilizes the combination of sites determined during the first phase to find a cost-optimal pairing of PV and storage that delivers firm electricity for the given load. Costs of firm electricity for common load patterns, e. g. the European Union or a linear load, could be less than USD 20/MWh, without transmission, if global generation sites are utilized; transmission would add between 34. 8 and 48. 9 for, respectively, high and low learning rates by year 2034. Long submarine power cables are being planned and built globally, enabling enhanced technological learning and consequently declining costs. We discuss several examples of combinations of solar generation sites with electricity costs including transmission depending on expected learning rates. This approach could help identify stable configurations for affordable and firm electricity from renewables and inform plans for necessary long-distance power transmission infrastructure. We give an example of an intercontinental power cable that could be built along the known route of an existing submarine telecom cable. • Non-linear optimization enables firm renewable electricity at low cost • Two-part non-linear optimization approach allows overcoming computing limitations • Spatial optimization of renewable energy generation supports energy planning • Countries active in multi-lateral transmission planning may gain an advantage
Grossmann et al. (Sun,) studied this question.