Load migration would boost renewables, save money and reduce greenhouse gas emissions, study finds
image credit: Photo 168075784 © Media Whalestock | Dreamstime.com
- Sep 15, 2020 9:13 pm GMTSep 15, 2020 9:04 pm GMT
- 660 views
Moving processing work from data centers on PJM Interconnection’s grid to data centers on the California Independent System Operator’s grid when renewable generation is peaking in California would boost renewable generation, reduce greenhouse gas emissions and save money for data center operators, according to a paper published last month.
The paper was published in the journal Joule and written by Sangwon Suh, an industrial ecology professor at the University of California, Santa Barbara; Jiajia Zheng, a UCSB doctoral student; and Andrew Chien, a computer science professor at the University of Chicago.
In an article about the paper, Harrison Tasoff wrote that the researchers looked at the effect of moving the performance of computing tasks from data centers in PJM’s territory to data centers in CAISO’s territory at times when renewable generation on CAISO’s grid was being curtailed or producing power that was being sold at a negative price.
Using historical curtailment data from CAISO, Suh, Zheng and Chien found that data center load migration, or “moving bits not watts,” could have absorbed up to 62 percent of CAISO’s curtailed electricity capacity, which was nearly 600,000 megawatt hours, last year. That, they said, would have resulted in a reduction of the equivalent of up to 240,000 metric tons of carbon dioxide emissions. Additionally, because the data center operators would have been using less expensive power, each of those tons would have saved them $240.
PJM’s territory was a logical choice to be the territory from which load would be migrated because it includes northern Virginia, where many data centers are located. CAISO's territory was a logical choice to be the territory to which load would be migrated because on sunny days CAISO often has to curtail solar generation or pay other states to take the power that solar generation produces. Migrating load from PJM to CAISO is possible, Zheng said, because the average server usage rate at data centers typically is below 50 percent, meaning data centers in CAISO’s territory would be able to accept the work that the data centers in PJM’s territory are transferring.
The researchers said the study’s results were within the range they expected. What they found surprising, they said, was that data center operators weren’t already migrating load to save money. Suh speculated that one reason they aren’t is because some operators of data centers in PJM’s territory don’t have data centers in CAISO’s territory and therefore can’t migrate load without giving other data center operators work they’ve been contracted to do. He had to speculate because he and his fellow researchers had a hard time getting information, including the power usage characteristics of data centers and how much energy data centers consume, from data center operators.
Public estimates of how much energy data centers use vary pretty widely, but all peg it as a lot. David Mytton, a British tech entrepreneur and angel investor who recently earned a Master of Science degree in environmental technology, devoted a blog post to the subject in February in which he cited estimates of annual power consumption by data centers ranging from 200 to 500 terawatt hours, meaning data centers consume at least 1 percent of all power produced globally. As a result, widespread load migration designed to take advantage of renewable power surpluses when they exist could generate some big savings for data center operators and produce a significant reduction in greenhouse gas emissions. Load migration also could help grid operators reduce peak demand during heat waves.
For the load migration modeled in the paper to happen, Suh said, data center operators would have to cooperate with one another. The governments of the states where their data centers are located or the federal government could encourage that by offering them incentives, imposing a tax on carbon emissions or through other policies. Virginia, where many of the data centers in PJM’s territory are located, will join the Regional Greenhouse Gas Initiative at the start of next year, which could eventually inspire some data center operators there to consider load migration at least within their own operations if not outside them. Additionally, some big technology companies that operate multiple data centers also could take a look at using it within their operations to help satisfy their climate pledges.
For example, Alphabet CEO Sundar Pichai told Reuters earlier this week that his company has put in place the “stretch goal” of having its Google subsidiary completely power its data centers and offices with carbon-free electricity by 2030. In an article on Pichai’s announcement, Paresh Dave wrote that the goal will force Google to move beyond offsetting the carbon emissions produced by its power consumption and require technological and political breakthroughs to achieve. In that context, load migration would appear to be low-hanging fruit. Google already has designed and deployed a system that enables its largest data centers to shift the timing of many of the computing tasks they perform to when low-carbon power sources, such as solar and wind generation, are most plentiful. That, of course, is a type of load shifting, so load migration would seem to be the next logical step for the company to explore.
Microsoft also is doing some work that could be a precursor to load migration. A post on the PJM Inside Lines blog said that as part of PJM’s Advanced Technology Pilot Program, Microsoft is testing the ability of data centers with battery-based storage systems to act as flexible load that can both absorb power and send it back to the grid.
Large-scale load migration no doubt would require some additional data transmission infrastructure to be put in place. But the amount of work needed likely would pale compared to the amount of work needed to build the transmission infrastructure required to link the country’s Eastern and Western interconnections and the Electric Reliability Council of Texas to enable large-scale flows of solar power from California to the Midwest and East during the day and wind power from Texas and the Great Plains to California during net peak hours there.
By itself, load migration won't produce the benefits that could be reaped from expanding the links between the interconnections and constructing additional transmission infrastructure to move renewable energy around the country. But it could be a good way to take advantage of California's bountiful solar power either before or regardless of whether those links and infrastructure get built,