Privacy-Preserving Bi-Level Optimization of Internet Data Centers for Electricity-Carbon Collaborative Demand Response

The escalating electrical demands of large-scale computational models in Internet data centers (IDCs) coupled with their significant carbon footprint underscore the potential synergy with demand response (DR) for promoting sustainable power system operations. Despite its potential, this intersection has been insufficiently investigated in existing studies. To fill the gap, an electricity-carbon collaborative DR (ECCDR) framework is developed and a privacy-preserving bi-level optimization model is proposed to fulfill this goal. First, the ECCDR framework is designed by combining dynamic carbon emissions from power systems with traditional DR, aiming to concurrently maximize the economic and emission reduction benefits. Second, a privacy-preserving bi-level optimization model is proposed to orchestrate computational task distribution within IDCs, facilitating load shifting in power systems. It is done by exchanging nonsensitive information between power systems and IDCs, ensuring privacy yet paving the way for ECCDR's pragmatic deployment. Third, distributed photovoltaic (PV) and battery energy storage systems (BESSs) are integrated into IDC operations, further amplifying ECCDR's potential. Simulation results reveal that the bi-level optimization model results in cost-efficient operations for both the power system and IDCs without invading privacy, while the ECCDR paradigm demonstrates superior advantages compared to the conventional DR.