The computation of regionalized short-run marginal costs of electricity: The ARCOMALT model

Abstract

Generation operation models, giving expected values of marginal fuel and failure costs, have been developed at EDF since about fifteen years. In order to study the regionalization of the future tariffs, we have recently developed a model for computing short-run marginal costs including not only generation but also the interconnection network. We shall first deal below with the necessary reduction of the general problem of optimization of the interconnection network operation before explaining how we solve the reduced problem. The French 400 kV interconnection network enables more than hundred plants to deliver electricity to consumers through more than hundred main nodes. The optimization of the operation has been examined for network planning and has led to a separation between shortage cases (for which additional outage due to the network is minimized) and normal cases (for which economic dispatching, taking into losses account, is established). In order to deal with these two cases with ARCOMALT, and to be able to introduce the randomness of all the variables concerned, we have reduced the number of nodes to about a dozen, which will be justified by both technical and economic reasons. It will also be shown that a good way to represent the interconnection network, from an economic point of view, is to let the various regions exchange electricity at marginal cost (either fuel or failure cost) with a transaction cost (a polynome, the degree of shich will be discussed) and bounds on exchange capacities. The reduced problem is neither linear (both criteria and constraints are not) nor of a small size, so that usual methods cannot be applied. Graph theory cannot be used in this case because it is a problem with variable gains related to the links, and not a problem with fixed gains related to the nodes (which could be solved). It will be proved that, using the additive properties of the criteria, the main problem can be decomposed into smaller sub-problems and the optimal size of these sub-problems being the 3 nodes-problem. Finally, the relaxation method, which proceeds in two steps, is described: a rough solution is first obtained through a primal approach before using a dual approach to complete the computation of the marginal costs.