No longer wishful thinking: AI in energy trading

Whether in voice assistants, navigation systems, facial recognition on our smartphones or recommendations on streaming services and social networks, AI is always working in the background, speeding up processes or providing us with information.

What does AI actually mean?

There is currently no universally accepted definition of AI that is used uniformly by all stakeholders. In general, it refers to the ability of machines to perform human-like cognitive functions, i.e. to perceive, learn, remember and solve problems, as well as to optimise themselves. AI systems are based on methods from mathematics and computer science. They are used when large amounts of data need to be analysed, patterns recognised and decisions made on that basis. The results are often available faster and more accurately than humans could ever achieve.

This speed is achieved by high-performance computers and data centres, which now have enormous power requirements. According to a study by the International Energy Agency IEA, , global electricity consumption by data centres in 2024 was around 415 TWh, which corresponds to about 1.5 per cent of global electricity consumption. According to the IEA, this demand will more than double to 945 TWh by 2030. By way of comparison, total net electricity generation in Germany in 2024 was approximately 470 TWh.

The EU AI Act

On 21 May 2024, the Council of the 27 EU Member States adopted the Artificial Intelligence Act (AI Act), a uniform framework for the use of AI in the European Union. Since it came into force on 1 August 2024, it has been the first set of rules of its kind in the world.

Among other things, the AI Act stipulates that AI applications may not be used, for example, to specifically influence and manipulate people's behaviour (‘social scoring’) or to monitor public spaces biometrically in real time. Furthermore, there are transparency and labelling requirements for artificially generated or edited content and special requirements for high-risk AI systems in the areas of critical infrastructure, healthcare and banking if they are to be approved for the EU market. In this context, Vattenfall Business Customer Sales has published an interesting article entitled AI skills as a success factor for SMEs.

How does Vattenfall use AI?

AI is used in many ways within the Vattenfall Group: at the Goldisthal pumped storage plant, we use AI to analyse drone images in order to detect cracks in dam walls. At the Juktan hydroelectric power plant in Sweden, a completely digital 3D version of the plant has been created, which will greatly simplify planning during a planned renovation. And in Aberdeen, AI analyses the flight behaviour of birds around our offshore wind farms there.

How AI helps in energy trading

The algorithms used in energy trading – also known as algo trading – are also becoming even more powerful with the help of AI. While traditional algo traders use fixed, rule-based ‘if A, then B’ strategies and are not adaptable on their own, AI-based trading strategies are self-learning, recognise patterns and adapt. The use of machine learning for pattern recognition, neural networks for forecasting and natural language processing (NLP) for evaluating publicly available news may sound like a thing of the future, but it is already being used successfully in energy trading. Of course, all such systems must be intensively tested and monitored – this is how we minimise potential risks and ensure that we comply with the applicable regulatory requirements (e.g. REMIT, MiFID, MAR).

AI in direct marketing – especially for weather forecasts

In the direct marketing of solar and wind energy plants, the electricity generated is sold either day-ahead for the following day or intraday on the same day, based on weather forecasts. For successful marketing, we constantly collate weather forecasts, electricity models and lots of other information – increasingly with the help of artificial intelligence. Our goal is to keep the costs for so-called balancing energy as low as possible. These costs arise when, depending on the weather, generation plants produce more or less electricity than previously forecast and traded. To keep the electricity grid stable, grid operators must activate reserve power plants. The resulting costs are passed on to those market participants who have deviations between the forecast and actual amount of electricity generated.

To achieve a high forecast quality, various weather and electricity models are combined to create an optimal mix. Increasing dependence on the weather means that the demand for data is constantly growing. Automation plays a central role here, as it is virtually impossible for humans to evaluate such large amounts of data manually. That is why we are also heavily automating Vattenfall's forecasting system and continuously developing it further.

Do you have questions about artificial intelligence in energy trading? Feel free to contact us!

<a class="arrow">renewables@vattenfall.de</a>

<a class="button">To the article overview</a>