It is perhaps unsurprising that Artificial Intelligence (AI) and Machine Learning (ML) have vital roles to play in energy and sustainability. This is due to their adaptability and efficiency at “learning” from complex data and generating timely predictions as well as insights. The rapid advance in capabilities and applications of AI & ML in this sector has given rise to use cases, such as energy forecasting, grid management and expanding innovations in sustainable technologies.
Forecasting the supply of renewable energy is challenging due to the variable nature of this energy source. Optimising the energy delivery, selling strategy and thus the commercial outcome of this energy source further compounds this challenge. Notwithstanding these complications, Google and its AI subsidiary DeepMind developed a neural network to increase the forecast accuracy for its Renewable Energy fleet. Using historical data, the network developed a model to predict future output 36 hours in advance. This greater visibility enabled Google to optimise its forward selling strategy and hence, increased the financial value of its wind power assets.
AI & ML can also assist with predictive maintenance of the power grid to enhance reliability and security. ML algorithm, such as that developed by utility company E.ON, help predict when medium voltage cables in the grid need to be replaced by leveraging data to identify patterns in electricity generation and to flag inconsistencies. More recently, ML was used to accelerate progress in perovskite photovoltaics*. Researchers at RMIT have combined a bespoke ML protocol together with a reproducible fabrication technique to predict the performance of potential chemical compositions for new perovskite cells. Insights from the ML models resulted in shortened discovery time and improved productivity.
Whilst some of the aforesaid benefits are significant, increasing use of automated and self-learning software would raise concerns around responsibilities, cost-effectiveness and scalability. A prime example of this is when establishing accountability of decision making by AI/ML algorithms, especially in sensitive sectors such as public infrastructure, energy markets or government functions. Furthermore, there are potential challenges in the interpretability and transparency of ML models. These challenges in turn can erode trust, hinder collaboration and scalability in larger deployments. Another less appreciated fact is the voracious demand on energy of this form of computing. A holistic approach is thus needed to maximise the benefits of AI & ML, while at the same time, minimising risks inevitable with any rapidly maturing technology.
* Perovskite solar cells are promising candidates for next-generation, inexpensive solar panels due to their commercially competitive cost and high power conversion efficiencies
References:
https://www.iea.org/commentaries/why-ai-and-energy-are-the-new-power-couple
https://www.linkedin.com/advice/3/what-main-challenges-benefits-using-machine-learning