Power/Performance Bits: March 27

Equalizing batteries; deep spiking neural nets; cobalt concerns.


Equalizing batteries
Engineers at the University of Toledo propose a bilevel equalizer technology to improve the life span of batteries by combining the high performance of an active equalizer with the low cost of a passive equalizer.

“Whenever we are talking about batteries, we are talking about cells connected in a series. Over time, the battery is not balanced and limited by the weakest cell in the battery,” said Ngalula Mubenga, assistant professor of electrical engineering technology at UT.

Typically, battery cell voltages in a large battery pack are balanced using either a passive circuit, which loses more energy, or an active circuit, which is 10 times more expensive.

“In spite of their significant losses, passive equalizers are used in most applications because they are relatively simple and low cost,” Mubenga said.

Dr. Ngalula Mubenga holding a battery cell next to the bilevel equalizer. (Source: Dan Miller, The University of Toledo)

In the new technology, the cells are grouped into sections. Each cell within the section is balanced by a passive equalizer, while the entire section is balanced by an active equalizer.

“If there are 120 cells in a battery, divide the cells into 10 groups of 12,” Mubenga said. “Then you only need nine active equalizer units and 120 passive equalizer units using the bilevel equalizer. With current active equalizers, manufacturers would have to use 120 active equalizers. For manufacturers that can’t afford to use only active equalizers, the bilevel equalizer is the solution to the problem.”

Experiments have shown that the bilevel equalizer increases the discharge capacity of lithium ion batteries by about 30%, and the pack lasts longer because the cells are balanced.

The team is licensing the hybrid equalizer and retrofit kit to manufacturers.

Deep spiking neural nets
Researchers at Oak Ridge National Laboratory developed a new method to make deep neural networks more energy efficient by converting deep learning neural networks (DNNs) to “deep spiking” neural networks (DSNNs).

DSNNs imitate neurons in the human brain via pulses or “spikes” in the place of actual signals, with the individual spikes indicating where to perform the computations. This process minimizes the necessary calculations and maximizes the network’s energy efficiency. However, energy efficiency comes at the cost of task performance, which the team hoped to overcome with a stochastic method for implementing DSNNs.

The team’s approach achieved nearly the same accuracy as the original DNN and performed better than a state-of-the-art spiking neural network. The team’s stochastic-based DSNN, which distributes spikes uniformly over time, consumed 38 times less energy than the original DNN and almost 2 times less energy than a conventional DSNN while delivering better task performance.

The researchers trained their network on clinical text data encompassing cancer statistics such as incidence, prevalence, and mortality across populations. The team applied the newly trained networks to clinical pathology reports, which contain vast amounts of unstructured text. The goal is to develop intelligent language understanding systems to extract the most relevant clinical concepts in the sea of text.

The clinical reports represent a “sparse” dataset, according to the team, which typically pose unique challenges to spiking networks. Most DSNN techniques have focused on computer vision tasks like handwriting recognition with “dense” datasets, where all variables in the dataset are populated with values.

“Spiking the network lowers energy consumption because we disregard the unnecessary computations and we look only for the relevant nodes of the network,” said Hong-Jun Yoon of ORNL, “and this is one way we get energy efficiency improvements while identifying important clinical information with high accuracy.”

The spiking networks were optimized on GPUs. The team says the methodology can be extended for training spiking networks, further increasing their energy efficiency.

Cobalt concerns
Researchers at the Helmholtz Institute Ulm (HIU) of the Karlsruhe Institute of Technology (KIT) are raising concerns about the future of cobalt supplies for lithium-ion batteries as demand continues expanding.

Besides lithium as charge carrier, cobalt is a fundamental component of the cathode in present lithium-ion batteries, determining the high energy and power density as well as the long lifetime. However, cobalt suffers from both scarcity and toxicity issues.

The researchers conducted a scenario-based analysis until 2050 for various applications of batteries, which shows that a shortage and price increase of cobalt are likely to occur since the cobalt demand by batteries might be twice as high as the today’s identified reserves. Lithium reserves are expected to be much less strained, but the production will have to be scaled up potentially more than ten times to match future demand.

However, both elements are strongly geographically concentrated, often in countries which are reported to be less politically stable. According to the researchers, this gives rise to concerns about a possible shortage and associated price increase of lithium-ion batteries in the near future.

Regions with highly concentrated reserves: the “lithium triangle” in South America and, for cobalt, the Copperbelt in Central Africa. (Source: Nature Reviews Materials ¬©Macmillan Publishers Limited)

“The future availability of cobalt for the mass production of LIBs has to be classified as very critical, which is also evident from the price increase of cobalt higher than 120% within one year (2016-2017),” according to Marcel Weil, HIU system analyst. Greater adoption of battery recycling would also be necessary to decrease the pressure on critical materials.

The researchers say it’s necessary to focus more efforts on developing cobalt-free battery technologies such as sodium-ion, magnesium-ion and other batteries based on abundant materials.

Leave a Reply