Valuation of Load Forecasts in an Ensemble Model

Abstract

Load forecasting is one of the bases of power system economic scheduling. High accurate load forecasts help the power system operator make better resource allocation and thus reduce the operational cost. The system operator can buy load forecasts in the data market and then combine them in an ensemble model to enhance the quality of final forecasts. Consequently, the operator should share forecast providers (agents) with the operational profit (or reduced cost) fairly. However, data from different agents affect the ensemble forecast jointly, making it hard to quantify the contribution of each individual forecast. There are few works regarding load forecast valuation in an ensemble model, especially in the electricity market. To fill this gap, this paper investigates valuation approaches. Four profit-sharing schemes with different computational complexity and synergy considerations are proposed and compared. Case studies on a real-world dataset illustrate how forecasts can be evaluated in an ensemble model.