A Pattern Recognition Algorithm to Estimate the Bidding Strategy in the Ancillary Service Market

In recent years, renewable energy sources are changing the way in which power systems are operated. Within this context, Ancillary Service Markets (ASMs) play a fundamental role, since they may contribute to support system operators in maintaining the balance between demand and supply and managing unexpected contingencies. Typically, energy is committed in advance for ASM, thus microgrids and plants should compute the dispatching program and the bidding strategy before real-time requests by the system operator are acquired. Hence, Energy Management Systems aimed at computing the optimal dispatch and bidding program should rely on estimates of ASM requests, whose reliability influences the outcomes. In this framework, this paper presents a model based on pattern recognition techniques to estimate the system operator requirements and the related possible income referred to energy exchange proposals in the ASM. A multiclass classification approach is adopted, i.e., the proposed approach estimates the revenue corresponding to acceptance of exchange proposals in the ASM, as discretized within a finite set of classes. Operatively, the model is tested on real data of the Italian ASM. Outcomes demonstrate that rejection is correctly identified in 80% of cases and the profitability range is properly estimated for 65% of proposals.