Physical System-Constrained Transactive Bidding Strategy for Data Centers With Workload Flexibility in Electricity Markets

Abstract

The workload flexibility in data centers remains underutilized in electricity markets, limiting their market-driven adaptability and overlooking dynamic market conditions. To address this challenge, we propose a physical system-constrained transactive bidding method for data centers to leverage market-driven flexibility of workload. First, we construct the energy model of data centers considering virtual machine multiplexing and support infrastructure constraints. Second, the market-driven flexibility of computational jobs is analyzed based on market requirements for classification and clustering. To reduce the computational burden of spectral clustering, we replace the intensive eigen-decomposition step with a lightweight orthogonal iteration method, enhancing scalability for large-scale datasets in data centers. Finally, the physical system models are integrated into the modified multi-agent deep deterministic policy gradient (MADDPG) method to ensure the inclusion of operation constraints and the optimal transactive bidding strategy. Numerical tests validate that the proposed method enables data centers to achieve higher profits across various ancillary markets while significantly improving computational speed.