Machine Learning Makes More Money For Renewable, Fleet, & Microgrid Operators

Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!

Electricity markets are becoming more and more sophisticated. Ancillary markets are emerging in multiple jurisdiction, continent-scale electricity trading is the norm in many places, and legacy assets are looking to optimize their dumb-grid approaches to work better with modern smart grids. Into this evolving and complex mix comes the potential of machine learning approaches to maximize fiscal benefits from generation and demand assets, with three firms taking different tacks. ENGIE, AMS, and GridBeyond each have a different approach to the opportunity, but all are leaning on machine learning to achieve their goals.

Smart grid graphic
Image courtesy US National Energy Technology Laboratory

ENGIE is a French multinational energy company, one with historical roots back to the Suez Canal Company of the mid-19th Century. But it has a very modern stance to innovation, with an ongoing program of open project competitions for key advances that the company sees as necessary in its core business areas. It owns many energy assets and constantly seeks ways to maximize its profits through greater revenue or decreased costs. One initiative that it has funded is to use state-of-the-art machine learning techniques to provide a competitive advantage for its traders by predicting the future market price evolution of the German Intraday Power market. Gaining an edge requires integrating weather, market, regulatory, grid congestion, production facility operation, and fuel price data, a constantly changing and challenging set to internalize.

“Instead of building a single model to analyze the market, we build thousands of such models simultaneously and dynamically assign to each of them a specific influence in the prediction, depending on its ability to extract knowledge from current observable market conditions.”

As Thinking, Fast and Slow points out, studies show that non-institutional traders are terrible decisions makers. More and more sophisticated computer technology and models are required to be more accurate, more of the time and to be less blind to key events which may be hidden in the data. This mixture is a good target for the ability to both discern signal from high volumes of noise, the trainability against historic results and the unblinking eye of neural nets. And unlike Skynet, it’s beneficial to us as efficient and profitable electricity markets are key to drive rapid growth of renewables.

Advanced Microgrid Systems (AMS) from California takes this a step further. Established in 2013, it has several solutions, with a core focus on battery storage management for most effective participation in both demand management programs and in time-arbitraged power provision. It has 360 MWh of battery storage under management, the most of any battery system management vendor per its claims, and that’s a growth market. But it also manages generation assets as well as storage assets, integrating both to the market. The company asserts 90% increases in revenue on battery storage and up to 10% increases in revenue for generation assets. A key focus is on virtualized power plants, an emerging trend in multiple jurisdictions around the world.

AMS uses machine learning in its market prediction space, seeing patterns in the vast quantities of data there. Other analytical and algorithmic techniques are used to make decisions, but the insights come from the neural nets it maintains. A key goal of theirs is to accelerate the transition to a renewable electric grid, and its part is distributed renewable production and storage assets.

In Australia, they are creating market-compliant bids every five minutes into the National Energy Market. In California, they deal with the complexity of the states’s ISO wholesale markets for storage. As more and more renewable assets emerge in commercial and microgrid deployments, and storage technology infuses the distribution grid, AMS’ value proposition will increase.

Heading back across the Atlantic, we find GridBeyond, a UK-based organization. Its focus is on intelligent demand-side management and storage. It works with companies across the UK to identify high-consumption assets such as refrigeration that can be remotely managed to provide demand reductions when the price signals are right. It ties them to its network of aggregated demand reduction and storage assets to provide a unified integration to UK utilities.

“CentrePoint is the cloud-based platform that sits at the heart of our technology. It’s the real brains of the operation. CentrePoint collates data from a number of sources, including site assets, and uses machine-learning algorithms to automatically place flexibility in the programme likely to generate the best returns.”

The fifth step in its five-step process involves its clients both spending less on energy, but also receiving direct payments from its demand management involvement. The team has experience across roughly 50 industrial sub-sectors including commercial electricity management.

Something I find interesting is that the recent major whitepaper from machine learning experts from North America and Europe, Tackling Climate Change with Machine Learning, doesn’t address this direct revenue optimization option. The paper talks about designing new markets and optimizing carbon markets, but doesn’t talk about storage and renewable generation assets maximizing profits with existing markets. This seems to be an oversight on the team’s part.

Machine learning initiatives are emerging in multiple markets where there are less irrational forces at play than are seen in the stock market. Commodity and energy markets have clear drivers for price, and emerging ancillary markets such as demand management are providing new and often complex ways to gain a profitable edge. The vast amounts of data that must be assessed in day-ahead and short-term markets to make efficient bids is beyond most humans, and in many cases very difficult to deal with through traditional computing models. Expect a lot more in this space.

Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.

CleanTechnica Holiday Wish Book

Holiday Wish Book Cover

Click to download.

Our Latest EVObsession Video

I don't like paywalls. You don't like paywalls. Who likes paywalls? Here at CleanTechnica, we implemented a limited paywall for a while, but it always felt wrong — and it was always tough to decide what we should put behind there. In theory, your most exclusive and best content goes behind a paywall. But then fewer people read it!! So, we've decided to completely nix paywalls here at CleanTechnica. But...
Like other media companies, we need reader support! If you support us, please chip in a bit monthly to help our team write, edit, and publish 15 cleantech stories a day!
Thank you!

CleanTechnica uses affiliate links. See our policy here.

Michael Barnard

is a climate futurist, strategist and author. He spends his time projecting scenarios for decarbonization 40-80 years into the future. He assists multi-billion dollar investment funds and firms, executives, Boards and startups to pick wisely today. He is founder and Chief Strategist of TFIE Strategy Inc and a member of the Advisory Board of electric aviation startup FLIMAX. He hosts the Redefining Energy - Tech podcast ( , a part of the award-winning Redefining Energy team.

Michael Barnard has 651 posts and counting. See all posts by Michael Barnard