Spatio-temporal load shifting for truly clean computing

TL;DR

This work tackles the challenge of achieving hourly 24/7 Carbon-Free Energy (CFE) matching for datacenter fleets by introducing a co-optimization framework that couples procurement, dispatch, and spatio-temporal load shifting within a continental European electricity system. It identifies three key signals—$S_1$ varying local renewable quality, $S_2$ low wind correlation over long distances, and $S_3$ solar time-lags due to Earth's rotation—and demonstrates how informed load shaping using these signals can significantly cut 24/7 CFE costs, with a marginal value of $1.29 \pm 0.07$ €/MWh per percentage point of flexible load. The results, drawn from 2025 parameterizations and 56 datacenter location combinations, show robust cost reductions across locations and seasons and reveal diminishing returns as flexibility grows. Open-source code and guidelines accompany the study, providing a practical path for broader adoption of truly clean computing across industries.

Abstract

Companies with datacenters are procuring significant amounts of renewable energy to reduce their carbon footprint. There is increasing interest in achieving 24/7 Carbon-Free Energy (CFE) matching in electricity usage, aiming to eliminate all carbon footprints associated with electricity consumption on an hourly basis. However, the variability of renewable energy resources poses significant challenges for achieving this goal. We explore the impact of shifting computing jobs and associated power loads both in time and between datacenter locations. We develop an optimization model to simulate a network of geographically distributed datacenters managed by a company leveraging spatio-temporal load flexibility to achieve 24/7 CFE matching. We isolate three signals relevant for informed use of load flexiblity: varying average quality of renewable energy resources, low correlation between wind power generation over long distances due to different weather conditions, and lags in solar radiation peak due to Earth's rotation. We illustrate that the location of datacenters and the time of year affect which signal drives an effective load-shaping strategy. The energy procurement and load-shifting decisions based on informed use of these signals facilitate the resource-efficiency and cost-effectiveness of clean computing -- the costs of 24/7 CFE are reduced by 1.29$\pm$0.07 EUR/MWh for every additional percentage of flexible load. We provide practical guidelines on how companies with datacenters can leverage spatio-temporal load flexibility for truly clean computing. Our results and the open-source optimization model can also be useful for a broader variety of companies with flexible loads and an interest in eliminating their carbon footprint.

Figures (6)

• Figure 1: Illustration of spatio-temporal load-shifting optimisation problem faced by a datacenter operator.
• Figure 2: Illustration of load flexibility concept: the absolute value of deviations between dispatched $\widetilde{d}_t$ [MW] and requested $d_t$ [MW] loads must fall within a certain flexibility range $f$ [%] for each hour $t$.
• Figure 3: Illustration of the signal 1: quality of local renewable resources. a: Assumed datacenter locations: Denmark, Germany, Portugal. b,c: Annual average capacity factor of onshore wind and solar photovoltaic. Data is simulated using the ERA5 reanalysis dataset for weather year 2013 and aggregated to 256 regions in Europe. d: Cost-optimal portfolio of renewable resources and battery storage sufficient for 24/7 matching. Steps on x-axis represent increasing share of flexible load. e: Cost breakdown of 24/7 matching strategy. f: Total annual costs of 24/7 matching strategy as a function load flexibility. Relative axis is normalized to the costs of inflexible load. h,j,l: Hourly spatial load shifts for the three datacenter locations. Color mapping represents the quantity of load "received" from other locations (positive values) or "sent" away (negative values). i,k,m: Hourly temporal load shifts for the three datacenter locations. Color mapping represents the quantity of load shifted to a given hour from other times, or shifted from a given hour to another time.
• Figure 4: Illustration of the signal 2: low correlation of wind power generation over long distances. a: Assumed datacenter locations: pairwise connections across regions with similar quality of renewable resources. Here: Ireland with Northern Ireland, or England, or the Netherlands, or Denmark (-west zone). b,c: Peason correlation of hourly capacity factor for onshore wind generation, if Denmark (-west zone, panel b) or selected region of Ireland (panel c) are taken as basis. As a result of different weather conditions, wind feed-in has a noticeable correlation falloff over distances of 300-400 km. Data is simulated using the ERA5 reanalysis dataset for weather year 2013 and aggregated to 256 regions in Europe. d: Hourly spatial load shifts for the selected scenario and datacenter; here datacenter is located Denmark (and another one is located Ireland). Color mapping represents the quantity of load "received" from other locations (positive values) or "sent" away (negative values). e: Cost savings of 24/7 matching with increasing distance between datacenters. Costs are normalized to the cost level of inflexible load.
• Figure 5: Illustration of the signal 3: time lag in solar radiation peaks due to Earth's rotation. a: Assumed datacenter locations: Denmark, Portugal, Greece. b,c: Peason correlation of hourly capacity factor for solar photovoltaic generation, if selected region of Portugal (panel b) or region of Greece (panel c) is taken as basis. Solar generation remains highly correlated over long distances, in contrast to wind generation. d: Difference in solar photovoltaic hourly capacity factors between two selected locations: Greece and Portugal. The two locations are approx. 2700 km apart, which results in a noticeable lag in solar generation peaks due to Earth's rotation. e: Difference in wind generation hourly capacity factors between two selected locations: Greece and Portugal. As expected, low correlation of wind feed-in over long distance results in stochastic pattern. f: Cost-optimal portfolio of renewable resources and battery storage sufficient for 24/7 matching. Steps on x-axis represent increasing share of flexible load. g: Cost breakdown of 24/7 matching strategy. h: Total annual costs of 24/7 matching strategy as a function load flexibility. Relative axis is normalized to the costs of inflexible load. i,j: Hourly spatial load shifts for the selected datacenter locations: Greece (panel i) and Portugal (panel j). Color mapping represents the quantity of load "received" from other locations (positive values) or "sent" away (negative values).