Increasing spatial and temporal resolution in energy system optimization model for energy access – the case of Kenya

Energy has been identified as an enabler for several of the Sustainable Development Goals (SDGs). Globally, 759 million people (2019) still lack access to electricity. Energy planning is important to describe the pathway to achieve the nations goals, where energy systems models are important tools to explore scenarios and provide insight. Until recently, modelling energy access with low electrification rate was conducted either at low spatial (e.g. national) or temporal resolution (e.g. annual time slices).  The central grid is often modelled as a black box with approximate optimization methods. This is recognised as unsuitable for understanding integration of technological alternatives to a centralised grid, including distributed generation and mini-grids/renewables. However, methods to model national energy systems at very high spatial and temporal resolutions are data and computation intensive. At the same time increased transparency on the data and code behind these models and insight is important as energy infrastructure is both capital intensive and strategic for the nation.

In this paper we investigate the use of OSeMOSYS, an open-source energy systems model, and increase the spatial resolution while keeping a medium time resolution. OSeMOSYS is a linear programming model and conveniently finds the global optimum in contrast to approximate methods. The approach provides insights into the trade-offs across supply and demand. The model generation is available in an open-source repository where results can be reproduced.

For this paper we use Kenya as our case study where still 16 million people lack access to electricity (2019). We select the spatial resolution to 378 supply cells (40x40km square cells) which leads to 591 demand cells split between electrified and un-electrified. The modelled number of seasons are 12 and the day is split into 3 slices: day, evening, and night, leading to 36 time slices. Specific demand profiles for electrified and un-electrified are assessed in combination with location specific supply options (expansion from the grid, PV, wind, diesel gensets).

Our preliminary results show that the varying un-electrified demand profile, with a high evening peak and low night-time demand, hybrid solutions are preferred with more than one supply option to meet the demand. The expansion of the grid to cells located far away is not motivated due to the low expected consumption, therefore decentralized supply options are required to serve at a high service level.

The results highlight the need for further work to investigate the sensitivity of the spatial and temporal resolutions in combine in energy systems optimization models.