Co-optimization of Coincident Peak Management and Ancillary Services for Flexible Data Centers

Flexible Data Centers (FDCs) have become an important demand-side resource in power systems as their scale and flexibility continue to grow. By rescheduling interruptible and delayable computing tasks, FDCs can rapidly adjust power consumption and participate in electricity market activities such as demand response and ancillary services (AS). Meanwhile, many electricity markets adopt a coincident peak (CP) mechanism, under which FDCs may incur substantial charges if they fail to reduce load during peak hours. As a result, FDC operations must balance computing performance, market revenue, and CP risk. Under these conditions, the major challenge is to determine operating strategies that can simultaneously satisfy computing needs, support market participation, and reduce peak-related penalties under uncertain CP events and multiple market objectives. This paper proposes a unified co-optimization methodology for FDC operations. A high-fidelity physical–economic model is first constructed to capture the relationship between input power, computational throughput, and economic returns. Then, a model-agnostic three-stage CP prediction strategy is developed to identify periods with high CP risk. Building on these components, a risk-driven and prediction-aware co-optimization methodology (RPCM) is developed to jointly optimize baseline workloads, AS bids, and peak-avoidance actions. Case studies in the ERCOT market in Texas show that the proposed method achieves significantly higher overall profit than traditional approaches that treat these decisions separately, as it can effectively avoid CP charges while increasing AS revenue. The results demonstrate that the proposed methodology provides a practical and robust decision tool for FDC operators under multi-market conditions.