Overview
The energy sector has become increasingly more challenging over the past decade with the sustained growth in demand for electricity, management of centralized power generation, grid complexity, rising expectations for customer service, and most importantly global de-carbonization efforts. These complexities have resulted in a renewed demand for technology adoption in the sector to fulfill modern operational and technical requirements. Areas ripe for AI enabled disruption in the energy and utilities sector include smart grids, electricity trading, demand and supply forecasting and predictive maintenance.
However, these use cases require supporting conditions such as a high level of digitalization and a correspondingly large data volume that can be analyzed and evaluated in multiple ways. The energy sector has systematically been taking steps in this direction to increasingly harness AI to ultimately derive maximum efficiency and profitability from its application.
Figure 8: Energy & Utilities: AI market spending in North America in US$ billion
Source: AgileIntel
Trends
Managing Smart Grids and Optimizing Energy Distribution
Unpredictable extreme weather incidents and the increasing number of grid-connected devices, both consumer and commercial, have led to a surge in AI investments in power grids. Modern power grids need to support multi-directional flows of electricity between distributed sub-stations, users and the grid itself, all the while managing a complex demand and supply mechanism and retaining resilience against outages and shutdowns. AI-aided power grids, known as smart grids, tackle this complexity by allowing devices and assets within the grid to communicate with each other thereby facilitating enhanced control and self-regulation.
Predictive maintenance algorithms used in smart grids analyze data from sources such as usage stats, weather, current mechanical efficiencies and historical maintenance records, to anticipate potential breakdowns before they occur. By using sensors and other specialized tools, these algorithms also perform simple troubleshooting and repairs, notifying technicians only when necessary. The operational readiness enabled by these predictions ensures minimized downtimes, a reduction in repair costs and bolstered reliability of grid infrastructure.
During sudden surges in demand, AI tools aid in improving the distribution of electricity by ensuring that power flow is directed where it’s required most, thereby minimizing the risk of a major outage. This is also true for renewable sources like wind and solar, which are often subject to drastic variabilities. AI algorithms assess climatic data, historical output generation data, and real-time conditions to enable grid operators and providers to reasonably forecast the amount of energy that can be available for consumption over a period, thus in turn allowing informed decisions and a better balance in supply and demand. AI-powered supply optimization also allows surplus energy generated during peak times to be stored and used when regular output is not achievable.
On August 6, 2024, the U.S. Department of Energy (DOE) announced a US$2.2 billion investment in the nation’s smart grid infrastructure across 18 states to protect against weather incidents and add grid capacity to meet load increases. This announcement is part of a larger, US$20 billion investment plan under the DOE’s Building a Better Grid Initiative launched in January 2022, with a focus on developing and integrating advanced AI capabilities. Such high levels of capital expenditure are expected to lead to a steep rise in the demand for data collection devices and IoT tools used in smart grid operations. Examples of major energy and utility companies utilizing machine learning and smart sensors to achieve major operational benefits include:
- Duke Energy has partnered with AiDash and Amazon Web Services (AWS) to build new smart grid software and services to expand its suite of custom-built applications that help anticipate future energy demand and power grid problems. Known as the “Duke Energy SmartGen Program” it includes online sensors, a data management infrastructure, and equipment health and performance monitoring. Over 30,000 sensors have been integrated across 50 power plants thus far
- NextEra is a renewable energy company that uses predictive analytics based on machine learning models to boost the functionality of wind turbines and reduce maintenance costs. The software, developed by Space Time Insight, helps with performance optimization, real-time diagnostics and troubleshooting, and maintenance crew scheduling. An example of benefits derived from these applications is the company’s West County Energy Center where analysts using machine learning models discovered unusual patterns in the functioning of the combustion turbines in the plant. The information was shared with the inspection team which discovered that the turbine had suffered internal damage, that could have been aggravated if not for the early detection.
- In 2019, Enel, Italy began installing sensors on power lines to monitor vibration levels. Machine learning algorithms then enabled Enel to identify potential issues from the resultant data and determine their causes. Subsequently, Enel has witnessed a significant reduction in the sheer number of power outages by more than 15%.
Other examples include Estonian technology startup Hepta Airborne, which uses a machine learning platform with aerial video of their transmission lines to identify transmission issues, and State Grid Corporation of China which uses AI extensively to analyze data from smart meters to identify technical issues.
Energy Trading
AI is also being used to analyze volatile market dynamics by processing ground data on pricing, demand, and supply trends, enabling energy companies to make informed and profitable trading decisions. Algorithms used in trading can execute a significantly large number of trades in microseconds, saving massive amounts of human tasking and time. These trading algorithms are becoming an increasingly standard practice in the energy sector where output is dependent on a lot of natural factors that vary significantly over time. More so in case of renewable energy sources where power production does not always adhere perfectly to pre-planned metrics and as such requires abrupt modifications to be made at short notice.
An example is GE’s Alpha Trader which leverages advanced AI/ML to produce generation and price predictions, then combines them with a risk management approach matched to the customer’s risk profile. Using the Alpha Trade GE registered US$2.5 million in additional annual revenue for a 250-MW wind farm in the U.S. over 9 months.