Explainable DC optimal power flow decisions

This paper introduces a framework for explaining the solutions to instances of the DC Optimal Power Flow problem by deriving counterfactual explanations (CEs). We aim to answer questions such as “Why was this generator not dispatched?” by identifying the minimum change in input parameters, specifically the nodal demands. This allows the user to identify the most important difference between the real and expected market outcomes and observe which constraints have led to the solution. The framework uses a bilevel optimisation problem to find the counterfactual scenario. State-of-the-art methods are compared with data-driven heuristics on the basis of computational feasibility and CE accuracy. We find that, by simply querying a dataset of previous solutions to obtain candidates for possible sets of active constraints, we can derive CEs for problems that would be intractable using conventional methods. The research addresses the need for more transparent decision-making in electricity markets, particularly regarding economic dispatch decisions, which are becoming increasingly difficult to comprehend due to the complexities of modern power system optimisation problems.