Potential of data centers for fast frequency response services in synchronously isolated power systems

Grid frequency support is one of the most challenging issues in wind rich islanded power systems. This problem becomes critical with the displacement of synchronous generators and their associated services (i.e., inertia and primary operating reserve). The services that are lost can be replaced by other sources, such as demand response schemes to enhance the resiliency and security of power system operations. Demand response based on internet data centers is expected to become an increasingly important asset to make a significant contribution to frequency ancillary services. To exploit this resource, internet service companies are expected to combine the capabilities of a variety of data centers to participate as a single provider similar to a virtual power plant. In this context, this work develops a novel framework for cooperative participation of data centers delay-tolerant workloads and backup power supply units to provide effective fast frequency response service. This is achieved by employing the model predictive controller to initiate reference signals to data center resources while respecting device operating conditions and constraints. Various case studies are run on the modified linear model of the 39 Bus system via dynamic simulations for the projected 75 % system non-synchronous penetration. Simulation results demonstrate the potential of different data center configurations to stabilize grid frequency during signal delays and severe cascade failures. The analysis shows that the proposed framework is critical to the adoption of renewable energy and reduces the requirement for an expensive spinning reserve used in a typical power system.

Automated summary

This study proposes a framework for cooperative participation of data centers to provide fast frequency response services. By using a model predictive controller to generate reference signals for data center resources, the lost services can be compensated. Through dynamic simulations, it is demonstrated that different data center configurations have the potential to stabilize grid frequency during signal delays and failures.