Photo credit: The Washington Post via Getty Images
The world’s top minigrid developer has pledged to install 2,500 projects in Africa by 2028 — and said using artificial intelligence is the only way to achieve that goal.
Colorado-based Husk Power Systems, which unveiled the plan in a September 2023 Africa Sunshot announcement, is using the tech to cut the manpower required for minigrid operations in remote African communities.
“AI is the only way to automate operations and get to the scale needed to eradicate energy poverty in the shortest time possible,” said Nikhil Murarka, the company’s vice president of engineering and innovation, adding that it would require thousands of people to manage the minigrids it plans to install in the coming years.
“AI can help reduce that number by more than 75%,” he told Latitude Media. “Automating monitoring and management is critical.”
According to Andrew Allee, a manager in think tank RMI’s Africa energy program who has worked with Husk in Nigeria, minigrids are key for communities that have power needs beyond what could be met with tools such as solar lamps, but are too far from a grid to warrant a network extension.
In February 2023, the World Bank said solar minigrids are the most cost-effective way to bring power to the around 595 million people in sub-Saharan Africa without access to electricity. But the scale of the buildout required is immense. The World Bank estimates that getting electricity to 380 million sub-Saharan citizens by 2030 would require 160,000 minigrids, costing $91 billion. Developers have so far struggled to deliver even a fraction of that number.
The World Bank reported that there were 3,000 installations in sub-Saharan Africa as of February 2023, with 9,000 more planned “over the next few years.”
Husk's installations could represent a major chunk of these if the developer makes good on its plans. The company was founded in 2008 and has since grown into a market leader in terms of installations.
Even so, to meet its 2,500-minigrid goal, Husk would have to accelerate its work dramatically. So far, the company has built 220 minigrids, 200 of which are in India. Going forward, the company is focused largely on projects in Nigeria, having recently abandoned plans for expansion into Tanzania.
Getting it right in Africa will not be easy, however, given several quirks of the market.
For one, many developing economies in Africa lack regulation governing small power plants, so minigrid developers may have to fill out as much paperwork as major utilities, a major hurdle.
Furthermore, the available public funding often needs to be matched by private investors that may be wary or ignorant of the business models supporting minigrid revenue streams. A 2020 report found only 13% of donor support for minigrids was actually getting through to projects.
Husk attempts to circumvent this challenge by selling electricity on a pay-as-you-go basis. This allows it to recoup the cost of development over the lifetime of its plants and avoids the need for its customers to pay for projects upfront. To work, the approach requires a detailed knowledge of likely usage patterns and involves cutting operating costs to the bone. But it is not cost effective to rely on humans for these tasks. That’s where AI comes in.
“Husk uses predictive AI to forecast supply and demand,” said Muraka, “and then deploys AI-powered algorithms to deliver electricity to its customers at the lowest cost at any given time.”
The company started piloting in-house-developed AI systems in 2020 and has found they could cut the use of backup diesel gensets by 40% — while increasing solar utilization 20% and extending battery lifespans by up to six months.
“Husk is ultimately able to pass on some of those savings to its customers, while also reducing its own operating costs,” Murarka said.
After turning cash-positive in January 2023, the company claimed to be the first profitable minigrid company in Africa and Asia. It is currently able to offer what it says is the industry’s lowest levelized cost of electricity, at $0.27 per kilowatt-hour, while developing new projects at a rate of 16 per month in India and four per month in Nigeria.
Husk is now planning to expand the use of AI from the supply side to the demand side of its operations.
“Demand-side management will help us to shape the demand curve and introduce time-of-use pricing and other mechanisms for lowering the cost of energy production and reducing prices,” Murarka said. The company has developed an e-commerce platform specifically to sell and finance electrical appliances and electronics, and plans to integrate AI into customer recommendations.
Allee at RMI said technologies such as AI have major potential in developing economies.
“If we were creative, we could use AI to take a look at the weather forecast, see how much battery reserve I’m going to need tomorrow and make sure that I run my diesel generator enough that my batteries are topped up optimally,” he said.
But while such applications could play a role in helping minigrids scale to the level required for full rural electrification, Allee said the potential for human support should not be ignored.
“The real game-changer is if their tech works, and it doesn’t always in these rural environments,” he said. “My experience in this sector has made me increasingly convinced that people are really important problem solvers.”