|
Variable renewable electricity (VRE), generated for instance by wind or solar power plants, is characterised by negligible variable costs and an availability that varies over time and space. Locating VRE capacity at sites with the highest average availability maximises the potential output. However, potential output must be curtailed, if system constraints prevent a local use or export. Such system constraints arise from the features defining the system, which I denote as system topology. Therefore, site choices that are unfavourable from a potential output perspective may still be optimal from a total system cost perspective. Previous research has shown that first-best investments require nodal prices that take account of the system constraints. Market designs that do not reflect nodal prices, such as uniform pricing, typically fail to achieve optimal site choices. However, a profound theoretical understanding of the economic trade-offs involved in the optimal spatial allocation of VRE is lacking. My paper contributes to filling this research gap. To do so, I develop a highly stylised model in which producers, taking into account the system topology, allocate VRE capacity in a one-shot game. Using the model, I analytically show that the optimal spatial allocation can be grouped into three spatial allocation ranges. Which of these ranges applies, I find to be highly dependent on the system topology parameters. In the first range, valid for relatively low VRE penetration levels, it is optimal to allocate all capacity to the node with the higher average availability. In the second and third range, it is optimal to allocate marginal capacity either fully or partially to the node with the lower average availability, i.e., the less favourable site from a potential output perspective. For uniform pricing, I show that producers allocate capacity inefficiently when VRE penetration exceeds a certain threshold. The resulting welfare losses I find to be especially high when transmission capacity is low, the difference in average VRE availability is large, and demand is concentrated at the node with the lower availability.
|
