Data centres, with their growing energy demands, present both a challenge and an opportunity for power systems, especially those with high levels of variable renewable energy. Researchers Mehmet Turker Takci, James Day, and Meysam Qadrdan from the University of Edinburgh have been exploring how data centres can transition from being passive loads to active participants in the grid, providing much-needed flexibility. Their work, published in the journal Applied Energy, offers a novel approach to quantify and utilize this flexibility.
The team developed an integrated, whole facility optimization model to investigate the most cost-effective operating schedule for data centres. This model considers IT workload shifting, uninterruptible power supply (UPS) energy storage, and cooling system dynamics. The goal is to establish a cost-optimal baseline operation for data centres, which serves as a starting point for assessing their flexibility.
The researchers then introduced a duration-aware flexibility assessment method. This approach calculates the maximum feasible duration for data centres to deviate from their baseline power consumption, given any start time and power deviation. Importantly, this method respects all operational, thermal, and recovery constraints, providing a realistic assessment of data centre flexibility.
The study revealed a clear temporal structure and asymmetry in data centre flexibility. Upward flexibility, which involves reducing electricity load, is primarily driven by deferring IT workload. This deferral also leads to a secondary reduction in cooling power. On the other hand, downward flexibility, which involves increasing electricity load, relies on boosting the cooling system’s power consumption, supported by thermal energy storage (TES) buffers, and charging the UPS.
The practical applications of this research are significant for the energy sector. By translating abstract flexibility potential into quantified flexibility magnitude and duration, system operators can better integrate data centres into grid services such as reserve, frequency response, and price-responsive demand. This could help alleviate the increasing strain on power systems while leveraging the untapped flexibility potential of data centres.
In essence, this research provides a robust framework for quantifying and utilizing data centre flexibility, offering a promising avenue for enhancing power system stability and efficiency in the era of variable renewable energy.
This article is based on research available at arXiv.