Integrated energy and ancillary services optimized management and risk analysis within a pay-as-bid market

In liberalized electricity markets, trading energy between generators and consumers occurs primarily on the Day-Ahead Market (DAM) one day in advance. However, the scheduled programs may not comply with grid requirements or real-time conditions. To ensure grid stability and sufficient reserves, system operators procure resources on the Ancillary Services Market (ASM). With the increasing share of renewable energy sources, many programmable generators are shifting their business model, from generating energy at base load to providing grid services. In this context, a DAM-based traditional approach to dispatch scheduling, widely adopted by existing techno-economics analysis, may result significantly suboptimal. This paper presents a novel model for dispatch optimization maximizing profits simultaneously on both the DAM and ASM, utilizing a mixed integer linear programming (MILP) formulation and a machine learning algorithm considering a pay-as-bid pricing system and predicting the probability of offer acceptance based on historical data. The proposed framework is modular and flexible, allowing for separate use of the MILP dispatch optimizer and the machine learning offer acceptance prediction model. A risk propensity factor is defined and the impact on the optimal bidding strategy, the expected profits, and their variability, is studied. A Montecarlo approach is used to evaluate the profits probability density function. The performance obtained (i.e. 20 min to optimize one week of operation of a Combined Cycle Gas Turbine) allows in applying the proposed methodologies for both long term energy system planning and daily production offer scheduling.