AI for utilities: use cases and examples

Entrepreneurs in the Energy & Utilities sector, operating in complex and technology-intensive fields, are driven to adopt innovations quickly and extensively. With its potential to boost human productivity, artificial intelligence (AI) is attracting significant attention, with 92% of Energy & Utilities companies already invested in or planning to invest in AI within the next two years to gain a competitive edge. According to PWC, 67% of energy executives have experienced AI benefits in improving customer experiences, while more than half have realized advantages in decision-making (55%) and product and service innovation (53%). As AI is expected to play a crucial role in the future of Energy & Utilities, let's explore the most profitable AI use cases.

1. Detecting defects and implementing predictive maintenance.

Energy and utilities companies encounter the challenge of identifying faults in vulnerable processes such as pipes, wiring, and machinery. Such faults result in substantial financial and reputational losses for these companies, making defect detection a top priority.

According to PWC, 35% of energy executives are recognizing the advantages of AI in conducting predictive maintenance and automating routine tasks (33%). By leveraging AI image recognition and computer vision systems, facilities can efficiently process photos and videos to identify any potential faults and alert employees accordingly, thus mitigating the risk of damages.

Drones are being utilized to aid power grid operators in power line inspections. By creating a specialized inspection platform that employs continuous machine learning to analyze a large amount of visual data obtained from drones, a birds-eye view of the grid and its critical components can be generated. This comprehensive report is then reviewed by work crews, with suspicious images flagged for human inspection. The inclusion of artificial intelligence has expedited the processing time by up to 15% (ML), enabling hundreds of kilometers of power line data to be analyzed in mere minutes and resulting in the detection of more defects per line kilometer compared to the traditional process without AI.

In addition, pattern recognition algorithms, video stream processing, or AI image recognition can be implemented for safety purposes, detecting violations of proper dress code during hazardous operations.

Predictive analytics can provide early warnings on equipment health, allowing operators to take proactive measures to prevent health, safety, and environmental damage. Predictive and AI-powered maintenance is also being utilized by water suppliers. AI and machine learning analyze and learn from data patterns that indicate potential breakdowns, resulting in increasingly accurate predictions with each iteration.

Sensor-rich AI technologies are aiding utility businesses in leak prevention. Advanced devices track liquid flow and usage patterns, establishing a baseline. If an issue arises, the system will either alert a human to the exact location of the leak or shut off the supply, or both. When Energy & Utilities staff are connected to such a system via mobile devices, they can promptly address the issue, minimizing the risk of significant financial and reputational losses and enhancing customer satisfaction.

An additional instance is an AI-driven application that enables operators to anticipate failures in wind turbines' turbine blades, generators, and gearboxes while also optimizing energy production. Offshore operators can utilize cloud-based platforms that offer access to advanced analytics software incorporating AI algorithms to scrutinize incoming data for anomalies that could indicate potential issues with the monitored equipment.

A favorable outcome of this AI application is a reduction in equipment downtime.

Another responsibility of utility maintenance teams is to prevent unscheduled downtimes, which can cost millions of dollars in a single day. For example, a report by the World Economic Forum for the oil and gas industry suggests that 92% of shutdowns resulted from unscheduled maintenance, costing suppliers an average of $42 million to $88 million annually in the most severe cases. By identifying defects and leaks in the early stages, utility companies can minimize the need for maintenance.

2. Revolutionizing the customer experience

With the aid of AI, utility suppliers can forecast water/heating/energy usage, allowing for dynamic pricing with reduced rates during periods of excess capacity.

AI facilitates a personalized view of household appliance interactions through advanced customer segmentation and disaggregation. This empowers customers to regulate and adjust their utility consumption in a more rational manner, resulting in reduced bill sizes.

For example, AI can inform a utility that Household A charges their electric vehicle between midnight and 6 a.m. each night. As a result, utility suppliers can advise these customers to charge their electric vehicles earlier when electricity is less expensive.

AI presents new opportunities for the marketing departments of utility companies to target customized and personalized campaigns and loyalty programs based on valuable insights into customer behavior. For example, a utility supplier can use AI to analyze the charging behavior of household A's electric vehicle and offer them special discounts or new electric devices and appliances as part of a tailored marketing campaign.

3. Forecasting the water and energy consumption

Efficiently distributing utility production is crucial for the sustainable growth of utility companies. The USEPA estimates that energy consumption accounts for 25-30% of total operation and maintenance (O&M) costs. Artificial intelligence can optimize the runtime of equipment to ensure it's only used when necessary, resulting in quick cost reductions for companies using AI.

The graph below illustrates an example where supply and consumption must match in real-time, excess power should be transmitted to regions with higher demand, and energy transfers between regions are limited by network capacity. AI can create cost-effective strategies for scheduling power consumption, potentially saving consumers money and accelerating decarbonization efforts.

Utility companies use data analytics tools such as Power BI to forecast and prepare for future customer demand. According to Microsoft, its product can shift companies from making reactive decisions to proactive strategies with better management of critical equipment and resources in energy production and distribution channels.

Microsoft has also created a solution to enhance the predictive analytics capabilities of power companies through Power BI. This suite of business analytics tools can improve and revolutionize the way sustainable energy is managed by generating value throughout the energy production, supply, distribution, and consumption process with data-driven insights.

Another aspect of predicting consumption is forecasting the load that will be placed on grids or pipelines.

Recent research has highlighted the importance of reliable and efficient forecasting systems for residential energy consumption, which accounts for 20% of overall energy demand. A report by Natural Resources Canada in 2017 found that domestic hot water (DHW) alone contributes to 17% of energy consumption, emphasizing the need for accurate predictions of hot water usage to conserve energy. There are algorithms and hybrid artificial intelligence systems that can forecast energy and water consumption, thereby predicting the network's load. However, accurately forecasting energy or water demand during abnormal weather conditions or public holidays remains a challenge and could be a potential competitive advantage if addressed.

4. Ecologic concerns

A digital twin is an AI-based model that serves as a virtual replica of physical equipment or systems. It supports energy optimization management by providing a simulation environment for training and testing AI systems, which can then be used to monitor and distribute energy and provide accurate forecasts. Multiple data sources are processed by algorithms, and the results are visualized for better analysis.

Despite the economic challenges, many companies are committed to decarbonizing their operations and value chains. According to BCG studies, the potential impact of AI on corporate sustainability is significant, with an estimated $1.3 trillion to $2.6 trillion in value generated through additional revenues and cost savings by 2030.

In addition to energy, AI can also be used to effectively maintain clean water. Meeting effluent compliance standards is crucial for many organizations, both public and private. AI models can learn from unique characteristics to ensure compliance and avoid costly fees.

5. Optimizing energy production and scheduling

An increasing number of energy executives, 31% to be exact, are recognizing the benefits of using AI to analyze production scheduling scenarios through simulation modeling.

The current trend in energy efficiency advocates for the use of decentralized microgrids that combine various renewable and non-renewable sources of power. In contrast, traditional energy supply chains are centralized and rely on power plants that provide electricity to end-users.

AI technologies can predict the most efficient times to produce energy, such as when the sun is shining or the wind is blowing. Excess energy can be stored in home batteries when energy is cheap, as selling it to the supplier may not be profitable.

Additionally, AI can forecast periods of high energy usage and sell the accumulated energy when prices are higher. The main goal of these AI enhancements is to maximize profits for micro-producers and reduce costs for end-users.

6. Optimizing logistics network

Logistics is a crucial aspect of certain sectors in the energy industry, such as oil and gas. With complex supply chains involving various stakeholders and decision-makers, such as producers, distributors, and environmental recycling companies, artificial intelligence can aid in coordinating operations teams and warehouses to ensure the availability of critical products like refill tanks. In the realm of transmission and distribution, AI-based simulation techniques, coupled with IoT technology, are widely used to improve processes.

For instance, AI can efficiently process vast amounts of data from sources such as customer feedback, traffic data, and GPS locations of trucks to optimize fleet management, plan routes, and make better route selections. Additionally, AI can assist in predicting and refining market prices for gas and electricity.

Machine learning can aid energy companies in various ways, such as optimizing shipping operations, replacing assets, mitigating risks, and enhancing delivery times while reducing overall costs.

7. Enhancing inventory management

Efficient network and infrastructure planning is crucial for companies to avoid losses resulting from inventory lagging behind demand. AI technology helps companies in this regard by enabling more proactive planning. For example, as New York plans to transition to clean energy by 2040, AI can be used to determine the optimal number of solar panels, wind turbines, and energy storage capacity needed for the transition based on available data.

In the case of renewable energy sources, such as offshore wind farms, challenges arise due to high costs, weather conditions, resource limitations, legal and government restrictions, and ecological and biological considerations. To overcome these challenges, AI models can be developed that incorporate all of these factors to effectively plan and schedule the construction and operation of offshore wind farms.

Energy and utility companies, including those that operate with renewable energy sources, can improve their inventory planning by gaining visibility into demand patterns. For example, they can adjust the number of charging points for electric vehicles based on new driving habits, or offer customers the option to use nearby unoccupied charging points. This not only results in happier customers but also reduces operational costs.

8. AI-Powered increasing the efficiency of Energy Storage

EPRI reports that current energy storage systems are typically designed to last for a duration of 4 hours or less, which is suitable for peaking capacity and ancillary services needs. However, in the near future, there may be a need for longer duration storage as it is deployed to replace higher capacity factor conventional generation, absorb longer periods of renewable overgeneration, and support resilience during severe weather events.

Traditionally, material research, prototyping, testing, validation, and lifecycle assessment required a significant amount of time in the labs. However, with the help of virtual laboratories, these processes can be accelerated, leading to faster research and development of new materials. With the increasing demand for high charge and discharge rates, as well as low cost, due to the rapid storage and release of energy from renewable sources, AI and ML are being used to predict new battery material discovery and establish a better understanding of material behavior.

Conclusion

It's evident that the energy and information industries, each worth trillions of dollars, are now intersecting in a way they never have before. Machine learning technologies have the potential to revolutionize the utility industry and lead us to a better world by improving safety, optimizing water and energy usage, and promoting environmental and user-friendly practices.

However, implementing AI is not a straightforward process. To simplify it, one can imagine it as several steps in an execution plan for an AI project: defining the problem to be solved with AI and setting business goals, forming an AI project team, exploring data, developing and testing the AI model, deploying it, and continually improving the AI system.

Fortunately, utility companies often already have the most critical component of AI: data. The next step to enhancing processes is finding the right use case, cleaning the data, and using a good AI model to achieve profitability.

Although off-the-shelf AI solutions and platforms are widely available in the market, they may not always be optimized for specific needs, or may only address a portion of a problem, resulting in dealing with symptoms rather than the root cause.

Therefore, it is crucial to select the appropriate platform that enables the custom building of an AI solution tailored to solve specific business problems with the highest accuracy. This task requires a team of high-profile AI and ML specialists to ensure the success of the project.