How generative AI is coming to the energy sector

How generative AI is coming to the energy sector
Matt Wytock, CEO, Gridmatic

Matt Wytock, founder and CEO of Silicon Valley-based power marketer startup Gridmatic, discusses how the same methodologies used to generate art can help to decarbonise the grid.

Generative AI – the type of AI able to generate texts, images, audio or data in response to prompts – has become the latest ‘buzz’ with the publicity and popular interest around OpenAI’s ChatGPT and its introduction by other companies such as Microsoft with Bing AI and coming up Google and Meta.

In practice, generative AI is not new but rather an evolution of AI as data processing capabilities – storage, power, speed – have advanced.

While for many the pervasiveness of AI may not be apparent, the energy sector has been an early adopter with its need for modelling and forecasting, while other applications have included cybersecurity and theft detection.

Have you read?
How to advance data analysis and AI in the grid – dena recommendations
‘We must demystify new tech like artificial intelligence’: Viviana Vitto

With the ever increasing importance of more sophisticated AI-based algorithms to support the complexities of decarbonising the grid and growing the uptake of renewable energies, Matt Wytock, a machine learning expert and former software engineer at Google, in an exclusive with Smart Energy International, gives some insights on generative AI as an emerging technology in the energy sector.

What is generative AI and what are its benefits compared with other types of AI?

Generative AI is a type of artificial intelligence that focuses on creating new content based on the patterns and structures it learns from existing data. It is thus often used for creative tasks whereas other AI types are often used for goals such as classification or decision-making.

That said, various forms of AI are often used together. For example, generative AI can create additional data for training other AI models and generating new images or text samples can help improve the performance of AI models in tasks like image recognition or natural language processing.

What types of energy sector use cases is it best suited for?

Generative AI is particularly well-suited for energy sector use cases that require complex data analysis, pattern recognition, forecasting and optimisation.

Some of these use cases include:

  1. Demand forecasting: Analysing historical data, weather patterns and socioeconomic factors to predict future electricity demand with high accuracy and thereby enable better resource management and planning.
  2. Renewable energy output forecasting: Predicting solar and wind output based on weather data, historical production and other factors, thereby helping to optimise grid integration and manage the variability of these resources.
  3. Grid management and optimisation: Helping to optimise power distribution and transmission, considering factors such as load balancing, congestion management and asset utilisation.
  4. Energy trading and pricing: Predicting energy market prices and volatility based on historical data and market trends to enable optimised trading strategies.
  5. Customer offerings: Analysing customer data to identify usage patterns, segment customers and develop targeted product offerings, energy efficiency programmes or demand response initiatives.
  6. Energy storage optimisation: Optimising the operation of energy storage systems to maximise results.

How is Gridmatic using generative AI?

Gridmatic was founded on the premise that with higher renewables penetration and more unpredictable weather, predicting supply and demand has become much more difficult and it’s no longer feasible to do forecasting via traditional methods.

AI is not just useful, but necessary. We use multiple forms of AI, but fundamentally we have built a model of the US electricity grid down to the nodal level. This foundation enables AI-powered forecasting that allows us to do all of the use cases mentioned above.

What are some of the project outcomes?

Among the outcomes are the use of our algorithms in energy trading to successfully trade on the day ahead/real time markets in all seven ISOs, which we started in 2017.

Another is storage optimisation and we proved via backtesting of all systems in the ERCOT market (and validated by DNV) that our AI could provide an average uplift of 28% in storage system revenue.

At the time actual storage revenue averaged just over half of the revenue that could be achieved with perfect foresight of electricity prices, which is impossible.

A third is carbon-free energy supply and per a white paper with data centre provider EdgeConneX, our AI has enabled 24/7 carbon-free energy for a Texas data centre by forecasting and matching supply to demand on an hourly basis.

Following the pilot, the project is now being extended through a customer agreement with EdgeConneX.

Most recently Gridmatic has launched a retailer, Gridmatic Retail, to optimise clean energy purchasing including providing time-matched 24/7 renewables for commercial and industrial customers, initially in Texas.

The initiative is being supported with a two-year PPA for 15MW from Avangrid’s Barton Chapel wind farm for supply to those customers.