A Danish energy trader analyzes about 30 million data points each day in an effort to predict how cloud cover in Spain and rising temperatures in Germany could affect energy prices in France and elsewhere of Europe
In the region’s €1.3 trillion electricity market, MFT Energy A/S and more than a dozen other trading companies are increasingly embracing artificial intelligence and machine learning to make money , and in the process help set the prices that households and businesses ultimately pay.
Almost half of MFT’s electricity trades last year were done via algorithms, and the trend is upward. On Epex Spot SE in Paris, Europe’s largest exchange for short-term trading, automated trades accounted for up to 60% of trades last year, compared to 55% in 2021.
Algorithmic trading “is really going to be a requirement to maintain a competitive edge,” said Brad Blesie, chief investment officer at Trailstone Group, which uses AI to manage renewable assets, forecast weather and predict price levels. “It’s tough, but we really think there’s a lot of scope.”
The effort has gained more urgency as the continent increases its reliance on intermittent renewable energy. The supply can vary wildly if clouds settle in over the solar farms or gusts roll into the Atlantic. In Germany, electricity prices fell to a record negative 500 euros per megawatt-hour briefly in the afternoon, when a surge in solar power overwhelmed weekend demand earlier this month.
This kind of volatility and the massive amount of data needed to track supply and demand opens up lucrative opportunities for traders savvy enough to take advantage of satellite imagery, weather patterns and even social media posts to get ahead of price changes.
Energy traders are part of the complex interconnections that keep Europe’s electricity grid operational. They help keep the grid balanced by moving power to where it’s needed most, and they get a margin by getting ahead of the market. Volatility like that of last year’s energy crisis is a good thing for these companies, and MFT’s profits in 2022 rose more than eightfold to 576 million euros ($627 million), some 4.4 million euros per employee.
But automated systems aren’t perfect. During Germany’s solar rise, human traders had to take on a larger role in managing a situation that was too complex to be left entirely to software programs.
“It’s fair to say that fewer algorithms will be used on these extreme days, the first ever,” said Tim Kummerfeld, head of intraday trading at Danske Commodities A/S. But even in these scenarios, “algorithms will continue to help provide liquidity and reduce price volatility.”
The Danish firm made a profit of 2.25 billion euros last year, up from 265 million euros in 2021, as electricity and gas prices rose.
While automated trading takes place in all types of markets, from stocks and bonds to oil and metals, the amount of data about supply, demand and infrastructure is what sets the electricity market apart, especially in Europe, where interconnections between national networks create bottlenecks. Even in emergencies, when power outages are at stake, computers are likely to play an increasingly important role over time.
“The great thing about algorithms is that they also learn from training data,” said Danske’s Kummerfeld. “So the next time the market clears this low, there will likely be more algorithmic activity.”
More advanced traders are looking for unstructured data – information that normal systems cannot easily digest or understand. This includes text in multiple languages, images, financial reports from sites that do not use international accounting standards, and even anecdotes.
MFT traders obtain large amounts of information through interfaces on more than a dozen screens on their desktops. Signals come as graphics, numbers or plain text. It could be direct business ideas or hints of a beer supply problem somewhere in Europe.
In Germany, the market moves so fast that human traders in most cases cannot keep up with, for example, changes in cloud coverage. With the nation planning to increase solar capacity from about 68 gigawatts to more than 200 gigawatts by the end of the decade, spotting these changes will be even more vital.
In the past five years, the number of order submissions on Epex’s intraday markets has increased to more than 7 million a day from less than 1 million, according to exchange data.
At MFT, data models create over 1,300 trading signals daily. Trading is more than ever a game of understanding data and its impact on markets, said Jacob Guldberg, the firm’s vice president of algorithmic trading.
But there is no substitute for human knowledge and experience. Traders will still need to work and adjust the computer models.
“You still need a deep understanding of the market to use them fully,” Guldberg said.
