The Energy Transition Needs AI

The COP26 Climate Change Conference is over and The Glasgow Climate Pact is live — it’s the first ever UN climate deal to explicitly mention the need to phase out coal power and move away from fossil fuel subsidies. Though the language may be watered-down, the message is clear: The transition of greener forms of energy must be rapidly prioritized, especially in more developed countries.

But inciting a large-scale energy transition is not as simple as flipping a switch. Energy infrastructures and supply chains are vastly complex, presenting an urgent need for coordination. And in order for a green energy transition to be truly sustainable, it must also be able to provide stability and cost-savings.

Artificial intelligence (AI) technology can play a critical role in delivering the energy sector’s climate goals in a quicker and more sustainable way. AI technology can support the efficient integration of diverse sources of renewable energy into power networks, ensure better grid stability, and increase savings for companies, governments, and the general public.

The Use Case for AI in the Energy Transition

Over the last handful of years, we’ve witnessed AI prove its adaptability as a critical, enabling technology across a diversity of use cases — from helping researchers solve hard physics problems to providing solutions in the healthcare space, and everything in between. The energy sector too is a scene rife with opportunity for AI to work its magic.

In a whitepaper released earlier this year, Harnessing Artificial Intelligence to Accelerate the Energy Transition, the World Economic Forum (WEF) proposes that “[AI] can act as an intelligent layer across many applications to identify patterns, improve system performance, and predict outcomes of complex situations.”

The whitepaper identifies the backdrop upon which this transition is occurring as one marked increasing by decentralization, digitalization, and decarbonization. It highlights the impact of AI, not just as a standalone technology, but as a tool that can coordinate and work in tandem with other fourth industrial revolution (4IR) technologies like IoT, sensors, blockchain, and quantum computers. These technologies will converge in different ways to constitute the energy infrastructures of our future, but they require a connective tissue that AI can provide.

Integrating Renewables

Energy networks are inherently complex. The intricate logistics networks and supply chain operations that currently define fossil fuel-based energy systems will only become more voluminous and intertwined as new technologies and their accompanying infrastructures are added to power grids. As a recent article in Bloomberg puts it:

“The future will not just involve a move from power systems with hundreds or thousands of large generators to systems with millions of small solar projects and wind turbines. It will also involve hundreds of millions of networked electric vehicles and also, potentially billions of networked sources of energy demand — things like lighting systems, boilers, and heat pumps.”

It’s not just a case of adding more renewable sources of energy to the grid. The green transition will also require the rapid ability to aggregate data, coordinate functions, understand demand, monitor output, scan for disruptions, and predict the needs and outcomes of ever-more aggregated systems. Our future energy networks, operating at a scale of billions of interconnected technologies and smart devices, will simply be too big and tangled to be run by humans alone. They will need AI.

While AI is not a suitable substitute for many tasks that require human judgment, AI algorithms are very good at carrying out complex tasks at a rapid speed. This function suits the demand of energy grids — complex systems that require split-second decisions to be executed in real-time.

Ensuring Grid Stability

The expansion and entanglement of future, renewable-dominant power networks speak to the WEF’s point of decentralization. With the increasing decentralization of energy sources, it becomes even more important that the diverse pieces of the energy puzzle can “speak” to each other. Coordinating a decentralized power network, predicting capacity levels, monitoring and predicting the weather, and integrating data will all become paramount to securing a stable and efficient grid with fully-integrated renewable sources.

Today’s wind turbines, for example, are equipped with sensor technology. The sensors provide an enormous wealth of real-time data, allowing for the constant monitoring of a diversity of variables like wind speed, maintenance requirements, and power output. AI systems can help to make sense of the data and make predictions about how to operate distinct parts of a whole system in order to help prevent blackouts and optimize efficiency.

Of course, incorporating more renewable sources of energy comes with caveats that fossil fuels avoid. Wind and sun are intermittent resources. Making efficient use of these forces of nature will require far more accurate weather forecasting capabilities — a task in which AI is beginning to demonstrate a clear advantage.

In 2015, IBM’s SunShot Initiative used a machine learning model that synthesized historical weather data, real-time measurements from local weather stations, sensor networks, satellites, and sky images to improve the accuracy of solar forecasting by 30%. More recently, DeepMind made headlines creating an AI-based weather prediction model that proved to be better at making short-term predictions for the location, extent, movement, and intensity of rain than existing models.

The increasingly large amounts of data now captured and made available means that AI-modeled predictions will become more accurate, advanced, and nuanced. Algorithms can even be trained to predict variables like additional power used during a festive holiday like Christmas or a large-scale international event like the Olympics. These predictions can help to coordinate energy sources and networks in order to avoid blackouts and dispatch resources more efficiently where and when they are required.

Promoting Cost Savings and Efficiency

In order for renewables to gain a competitive advantage over traditional forms of energy production, they must be cheaper and more efficient. These two requisites are inherently linked. The WEF whitepaper acknowledges the ability of AI to identify patterns and produce insights to increase efficiency and savings as the number one most important use-case for AI in the energy transition. It states:

“According to BNEF’s net-zero scenarios, fully decarbonizing the global energy system will require between $92 trillion and $173 trillion of investments in energy infrastructure between 2020 and 2050. Even single-digit percentage gains in flexibility, efficiency, or capacity in clean energy and low-carbon infrastructure systems can therefore lead to trillions of dollars in value and savings.”

Indeed, efficiency in and of itself has been dubbed as “the first fuel of a sustainable global energy system” by the International Energy Agency (IEA). According to the IEA’s Sustainable Development Scenario, using resources efficiently can deliver more than 40% of the reduction in energy-related greenhouse gas emissions over the next 20 years. AI can help move this goal forward in a number of ways.

Unexpected disruptions, for example, are a major cash drain in the energy sector. According to EY, system disturbances and mechanical failures across the industry can cost between 3%-8% of capacity, and amount to $10 billion lost production cost per year. AI algorithms can be trained to automatically detect and locate malfunctions, making split-second decisions that can stop a domino effect of disruptions from occurring, resulting in improved grid reliability and increased efficiency in the power system.

AI can also have a positive impact on consumers’ wallets. By predicting energy demand, home solar and battery systems can be optimized to produce heating and cooling when needed. These predictions are made by analyzing both historical consumer behavior data and data collected from smart device-equipped energy infrastructures. AI systems can be programmed to execute decisions about load dispatch autonomously, without any input from consumers required. By producing and coordinating energy supplies efficiently, demand forecasting can play a critical role in reducing energy bills and inducing cost savings.

The Linchpin of the Energy Transition

There is no doubt that there’s a full spectrum of important use cases for AI in the energy transition. Just one of a host of 4IR technologies, AI could be the linchpin in a more efficient, cohesive, and sustainable energy transition — effectively integrating diverse energy sources, monitoring and predicting variables to improve grid stability, and enabling the pieces of a renewable energy puzzle to “speak” to each other in order to promote cost-saving efficiency.

As governments and industries race to confront the challenge of transitioning to decarbonized energy systems, critical technologies like AI will play a defining role. A collaborative and cohesive global approach to implementing AI and other 4IR technologies in the energy sector will be critical to creating solutions that are both scalable and sustainable.