A New Way to Store Sustainable Energy: ‘Information Batteries’

A future powered by sustainable energy sources can save the globe from drastic climate change and reduce energy expenses. But the renewable resource has an intermittency issue– the sun offers no power during the night, while winds can stop unexpectedly.

Additionally, power grids must keep supply and demand in balance, or threat rises and blackouts. As a result, renewable energy is discarded during times of excess production, while at other times, power plants burn fossil fuels to satisfy grid shortages.

” The way things are heading, in 5 years, the amount of renewable energy wasted away in California every year will amount the amount of power L.A. uses annually,” claimed Barath Raghavan, an assistant professor in computer science at the USC Viterbi School of Engineering.

Much better battery storage space– a holy grail for scientists worldwide– is considered essential to addressing the intermittency issue by storing energy when the wind and sun are strong. However, existing storage solutions, including lithium-ion batteries and pumped hydro, are costly and difficult to scale.

Suppose surplus renewable energy could be kept as computation instead? That’s the thinking behind “information batteries,” a new system proposed by Raghavan and Jennifer Switzer, a Ph.D. student from U.C. San Diego, recently published on the ACM Energy Informatics Review.

Predicting potential computations

The basic idea for information batteries is simple: When the renewable energy is readily available in excess, it is utilized to perform computations in big, energy-intensive data centers speculatively. These data centers– from Google and Facebook to Hollywood movie rendering– spend 10 to 50 times the energy of an average commercial building, according to the Office of Energy Efficiency and Renewable Energy. The stored computed results can be utilized later on when green energy is much less numerous.

“We had the observation that if we can anticipate possible computations that may happen in the future, we can do those computations currently, while there is available energy, and store the results, which now have embodied energy,” stated Raghavan, whose research focuses on systems as well as sustainability.

For example, each day, YouTube information facilities transcode more than 700,000 hours of videos to different resolutions. Most of these computations are predictable and can be executed when there is excess renewable energy. Now, the data is stored on servers for later usage, when there is much less renewable resource offered on the grid– relocating electrical energy intake from one time period to one more.

Optimization of sustainable energy storage

Then, how does this work like a battery? In the scientific sense, said Raghavan, batteries are stores of potential energy to do valuable work, electric or otherwise. A lot of storage space of energy into batteries transforms one sort of energy right into one more some kind of prospective energy, for example, electric right into gravitational. In this instance, information gives energy in the same way as a battery because electrical energy is developed into what might be considered “informative potential energy.”

Along with capitalizing on task predictability, the system is also versatile: the computations that are completed in advance do not require to match precisely with the analyses conducted later.

“We support pre-computing many pieces of computation and after that later can pick and choose small pieces of computation done in the past, like puzzle pieces, and construct them to quickly compute a completely new computational task,” said Raghavan.

For specific sorts of workloads, said Raghavan, the information battery system offers better efficiency than lithium-ion batteries. The particular efficiency depends on multiple variables, such as the sorts of computation performed and the predictability of power. However, unlike lithium-ion batteries, keeping data is cost-efficient in terms of both money and energy. This could assist in reducing dependence on fossil fuels, which are in charge of three-quarters of worldwide greenhouse gas discharges.

An encouraging future option

While the idea itself is reasonably simple, the challenge, said the researchers, is establishing what computation to do, where and when, and also exactly how these computations ought to be done to successfully recover the results later on.

In the paper, entitled “Information Batteries: Storing Opportunity Power with Speculative Execution,” Raghavan and Switzer offer a design and proof of principle implementation of the zero-carbon system, which includes recurring neural networks for anticipating the future schedule of renewable energy as well as upcoming tasks in data centers.

It additionally includes a cache where functions are saved and also a modified compiler to automatically change code to store and receive results. The infrastructure would be geographically dispersed, consisting of lots of small, dispersed data centers, each situated in a region of the nation where wind or solar manufacturing is known to be high.

“With this system, companies would utilizing power that would have been unloaded, and everybody else profits since the grid operator doesn’t need to spin up gas power at night hours to make up for demand,” claimed Raghavan.

There are specific restrictions, which the researchers explore in the study: for example, it is only possible in some workloads and in some contexts. But Raghavan thinks with the improved prediction as well as integration into large systems; the technology points in the direction of a promising future option for keeping eco-friendly energy. “In the civilization-scale obstacle of sustainability,” claimed Raghavan, “we need every tool we can get.”

Read the original article on Science Daily.

Related “New Tree-Derived Material Could Pave the Way for Better and Safer Batteries”

Reference: Jennifer Switzer, Barath Raghavan. Information batteries: storing opportunity power with speculative execution. ACM SIGEnergy Energy Informatics Review, 2021; 1 (1): 1 DOI: 10.1145/3508467.3508468