Penalising Forecast Accuracy? Data Valuation Challenges In Power Distribution Systems

Operation of modern power systems depends on high-quality data and forecasts. As the provision of valuable forecasts shifts to independent providers, such as renewable and distributed generators, there is a growing need for forecast-sharing markets. However, existing forecast valuation mechanisms, often inspired by game theory, overlook physical network constraints and the dispersed locations of forecast providers in realistic power systems. This work shows that widely used valuation methods, such as the Shapley value, can yield counterintuitive results when applied to load forecasting in distribution networks. Specifically, due to the location of providers and voltage constraints, less accurate forecasts may appear more valuable than precise ones, while some providers may even be penalised. Moreover, valueoriented mechanisms lack incentive compatibility, as certain providers may systematically misreport load forecasts to receive higher payments. These findings expose the limitations of existing forecast valuation approaches and suggest that incorporating accuracy-based scoring may enhance value-oriented mechanisms.