Exploring spatial and temporal resolution in energy modelling for developing economies’electricity systems

Abstract

Agenda 2030, with its 17 Sustainable Development Goals (SDGs), has set the direction on where development should be focussed. There are still around 675 million people who lack access to electricity (SDG7.1 – electricity access), mainly in Sub-Saharan Africa. 

Focusing on SDG7.1 and electricity access in developing economies, the transition has a short timeline, and the current reach of the electricity network is low in many of the countries with low electrification rates. Other supply options, solar PV and wind, have also had a dramatic decrease in cost over the last decade and do not need to be connected to a central grid, operating as a stand-alone or mini-grid. SDG7.1 and the transition of the electricity system pose new challenges for energy modelling, with a need to increase the spatial resolution as the location of the unelectrified population is a key parameter to understand. At the same time, understanding the overall electricity systems transition with increasing demands with economic growth, climate change and CO2 mitigation can increase tensions in the system while reaching SDG7.1.

Modelling electricity access increases the number of technologies and details needed in the system which in turn increases the complexity of the models, particularly spatial, temporal and mathematical. However, more detail, both parametrical and structural, can introduce more potential errors and uncertainty into the model. Therefore, energy models should be as simple as possible and as complex as necessary.

This thesis aims to give qualitative and quantitative insights into the mathematical, spatial, and temporal aspects of electricity systems modelling for electricity access in developing economies. The thesis analyses trade-offs between heuristic and linear programming methods when modelling electricity access, and the global sensitivity and relative importance of parametrical and structural parameters in ESOMs.

The method for achieving the aim of the thesis uses a four-step approach and is developed over the four papers appended to the thesis. First, the geospatial electrification problem is explored using two modelling methods:  a linear programming method, using the linear programming model generator GEOSeMOSYS specifically derived as part of this thesis; and a heuristic method, soft-linking the open-source tools OnSSET and OSeMOSYS. Second, these two models are compared with respect to computational effort,  insights derived from results, and detail – this all with regards to modelling electricity access in a developing economy. Third, using  GEOSeMOSYS and the method of Morris for global sensitivity analysis, the relative importance of spatial and temporal resolution, compared to other parameters (e.g., demand, discount rate, and capital cost) was examined. Finally, the global sensitivity analysis method of factor mapping, using scenario discovery was used for further analysing the relative important parameters that determine cost and low-carbon dioxide futures in the regional multi-country energy systems optimisation model ‘South America Model Base’ (SAMBA).

The results show that, for the modelled pathways, the two methods for electricity access show similar trends when the demand is changed, with low demand predominantly resulting in PV panels and batteries to serve the formerly unelectrified population, while higher demand results in more grid-connected households. The targeted demand level and profile for to-be-electrified households affect the optimal technology choices, and one such example is highlighted in the supply option of PV with battery. The cost competitiveness of PV panels with batteries decreases significantly when the demand profile increases during the night, meaning that when more continuously operated appliances are added, the PV and battery will not be cost-optimal in most cases – connecting to the grid is then the least cost choice. The two presented methods have different solution times with the linear programming method having a much longer solution time. The mathematical approaches to solving the transmission network are different, and both methods have trade-offs in their methods. These trade-offs are in the mathematical approach where OnSSET uses an exclusive technology selection optimisation leading to a suboptimal overall network, and GEOSeMOSYS rely on the assumption of linearity, which leads to very small incremental installations of transmission lines which in practice is not realistic.

The global sensitivity analysis of GEOSeMOSYS for electricity access showed that structural parameters: spatial and temporal resolution largely influence the result parameters and cannot be simplified without changing the results. The temporal resolution had a slightly greater relative influence on the assessed results parameters than the spatial resolution. This means that the results change for both the renewable electricity production share and the extent to which the unelectrified population gets connected to the grid/mini-grid as opposed to stand-alone. The scenario discovery analysis for understanding low-cost, low-carbon dioxide pathways pointed to the relative importance of demand. Together with low/medium discount rates, the low/medium demand was relatively the most important parameter for the South American case.

This thesis has therefore shown that, even though models should be as simple as possible, the spatial and temporal resolution cannot be simplified to a one-node analysis or low temporal resolution without this affecting the results when modelling electricity access. The mathematical choice for selecting the method of electricity access was analysed and trade-offs were highlighted. The main trade-off was in the network expansion where both methods use approximations that can lead to potential over/underestimating the investment need. 

The consistent results of PV and battery adds questions if the soft-linked method is necessary when modelling at very low demands, as the interactions between them two was low at low demands for the unelectrified population. 

The soft-linked method is, however, a good option on a higher level to explore electricity access and the implications on the whole electricity system. If the question is more complex (e.g., adding transportation, heating and cooling), then GEOSeMOSYS provides more readily available options for expanding the analysis, but at a coarse spatial resolution. Demand is a highly influential parameter, relative to other parameters such as discount rate, and learning rates for renewable energy technologies. Demand determines both the cost and the potential for achieving low-carbon dioxide futures. This thesis has explored the relative importance of parameters for electricity access and decarbonisation pathways on a national and continental level. The relative importance can differ depending on the energy system that is studied, calling for more research on different types of energy systems.

urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-352222