With the rise of AI-driven solutions for optimisation of trading using battery energy storage system (BESS) assets, Prudence Heck and Andrew Young of Spearmint Energy consider strategies and risks.
Recent advancements in generative AI have raised significant questions around its new potential applications, practical and theoretical limits, and government regulation.
Enjoy 12 months of exclusive analysis
- Regular insight and analysis of the industry’s biggest developments
- In-depth interviews with the industry’s leading figures
- Annual digital subscription to the PV Tech Power journal
- Discounts on Solar Media’s portfolio of events, in-person and virtual
Or continue reading this article for free
We believe that continued AI advancements will result in an increase in AI utilisation across industries, including one of our company’s key sectors of focus, battery energy storage trading in Texas’ ERCOT market.
However, the widespread adoption of AI tools to optimise battery energy storage trading poses a notable risk: Will AI become the future of battery trading even though many vendor-supplied AI solutions have little-to-no transparency around the way their algorithms work?
The short answer is yes – we see the utilisation of vendor-supplied black-boxes for battery energy storage optimisation as a trend that is here to stay. However, it should be noted that AI solutions need not be black-boxes.
Before elaborating on how Spearmint approaches battery energy storage optimisation, it’s important to note that black-boxes are not inherently bad. In fact, many are integral to our day-to-day lives. For example, Americans use personal vehicles to make 957 million trips a day, yet most will never need to know the specifics of how cars function to get from point A to point B.
Similarly, you may not know which neurons in your brain were involved in any decision, movement, or word you have spoken in your life, yet you have certainly made good decisions, moved, and spoken as you intended.
In the current state of the battery energy storage industry, a multitude of technology vendors provide AI-driven optimisation solutions consisting of visually appealing software built around one or more proprietary algorithms. These algorithms are tightly guarded, making it difficult to hedge against their weaknesses.
It is hard to blame vendors for protecting their intellectual property, but the secrecy still leaves clients like us exposed to risk. All technology fails at some point, and being prepared for those failures is critical. So, how can a company feel comfortable using AI technology without knowing its weaknesses?
We are not truly “in the dark” about the mechanics of AI optimisation tools built in-house. It is true that classic statistics presents a dichotomy between model interpretability and model flexibility. It is also understandable how deep-learning models may seem hard to understand. However, there are entire fields and subfields dedicated to model explainability.
These ideas are proven, coded, and optimised in open-source code libraries and utilised heavily across numerous industries and by Fortune 500 companies. The formerly prevalent tension between model interpretability and model flexibility has been addressed to the point that it can be considered a false choice: you can have both.
‘No single path to success’
Companies can manage the risks of vendor-supplied black-box optimisation through diversification. In aviation, there is a concept called multi-engine redundancy: if one engine were to fail during flight, the second engine can compensate for the loss of power, allowing the plane to continue flying and land safely. This redundancy greatly reduces the risk associated with engine failure.
Similarly, we pursue a multi-faceted approach to battery energy storage trading. We employ philosophies utilising ensembles of advanced, high-dimensional machine learning algorithms, coupled with the hard-won eyes of our experienced team of energy market professionals and third-party forecasting and pricing tools.
Indeed, we believe there is no single path to success, and therefore use multiple paths simultaneously to avoid a single point of failure. The result is a more robust approach to battery energy storage optimisation.
We are enthusiastic about the opportunities presented to our company and the battery energy storage industry at large driven by the accelerating deployment of AI tools in the market. However, like AI vendors who guard their IP, we too will remain guarded in our deployment of AI.
We value and strive for transparency and maintain a robust human backstop to guard against AI’s potential missteps. We believe such an approach will create significant upside for our investors and partners, while mitigating downside risk.
About the Authors
Prudence Heck is head of research and analytics at Spearmint Energy, with previous roles at Deutsche Bank, Noble Americas Gas & Power and Elustria Capital Partners.
Andrew Young is a data engineer at Spearmint Energy, responsible for building scalable, cost-efficient data pipelines using data from a variety of platforms.
Spearmint Energy is a US-headquartered green merchant trading company developing, owning, operating, and trading around battery energy storage, solar, and wind to reduce grid volatility, increase system resiliency, and help to reduce carbon emissions in a responsible and efficient way.