Using AI and ML to Improve Energy Efficiency

AI and ML can be used to address energy efficiency challenges and reduce global carbon emissions.

Energy efficiency and decarbonization are crucial strategies to reduce global carbon emissions and combat climate change. The International Energy Agency (IEA) highlights energy efficiency as the most significant measure in achieving Net Zero Emissions by 2050. Increasing efficiency measures on the generation and transmission of electricity (supply) as well as the use (demand) can reduce costs, help meet emissions targets, and promote public health.

AI solutions have the potential to play a pivotal role in the power sector, given the vast amount of data involved in making informed decisions. However, the adoption of AI solutions remains limited due to their adolescence and undetermined impact. Some success stories, like Vista Corp., have emerged by forming partnerships with academic institutions and professional service firms. This collaborative approach allowed them to advance their technology while sharing costs and minimizing internal organizational changes.

Incorporating AI into the energy sector offers several advantages. It enables predictive maintenance for utility companies and consumers, allowing them to proactively address potential equipment failures, reducing downtime and minimizing asset waste. AI algorithms can also analyze extensive sensor data to recommend upgrades for equipment with excessive consumption. Additionally, AI is valuable for optimizing the electrical grid by analyzing weather, historical demand patterns, current events, and consumer behavior, enhancing load balancing for more efficient electricity generation and transmission. As the US power grid becomes smarter and more interconnected, the risk of cyberattacks rises. AI can help by detecting anomalies in real-time, potentially preventing cyberattacks and enabling a quicker response to security threats.

Vistra Corp. is a “leading Fortune 500 integrated retail electricity and power generation company.” Based in Texas, Vistra supplies both electricity and gas to over 4 million consumers in the West and Southwest United States. In addition to being the largest competitive power generator in the US4, they have committed to “harness AI to improve efficiency, reduce greenhouse gases, and supply more reliable and predictable power” as well as “reduc[e] emissions by 60 percent by 2030 (against a 2010 baseline) and achiev[e] net-zero emissions by 2050.”

Vistra Logo,

Vistra embarked on its journey to fulfill its commitments with two key projects. The first project took place at the Martin Lake Power Plant in Tatum, Texas, where Vistra collaborated with McKinsey to create an AI-based heat rate optimizer (HRO). The heat rate measures a plant’s thermal efficiency in converting fuel to usable electricity. AI was utilized to monitor and adjust hundreds of variables continuously, which was previously impractical for manual management. The algorithm that was co-developed learned from two years of historical plant data, resulting in a 2% efficiency improvement. While seemingly modest, this translates to annual savings of $4.5 million and a reduction of 340,000 tons of carbon emissions, underscoring the significance of AI in enhancing energy efficiency.

The second AI project was done in partnership with the University of Texas’s Center for Applied AI and Machine Learning (CAIML) in Dallas. The goal was to “help the company project [electricity] pricing for its Moss Landing Energy Storage Facility in Monterey County, California.” Electricity prices are highly volatile due to the complex balance between supply and demand, especially with the increasing integration of renewable energy sources like wind and solar, which are inherently variable. As Dr Feng Chen puts it, “AI can help a company like Vistra forecast future generation and demand on load, wind and solar energy, and optimize bidding, scheduling and deployment of energy to improve profitability and market participation.” This project reflects AI’s capacity to navigate the challenges of an evolving energy landscape.

Vistra’s strategy of partnering with external organizations to develop AI products has yielded swift and tangible benefits. While not without investment, this approach enabled Vistra to implement solutions rapidly, leading to significant cost savings. Moreover, it allowed them to test and validate the technology’s advantages before committing to substantial internal investments. Said one operations supervisor, “We were all pretty skeptical of the tool at first, but when we got to play with the heat rate optimizer and see how it worked… we understood how it could help.” As a result, Vistra has successfully deployed over 400 AI models in various locations and is now shifting focus towards building its internal AI capabilities through hiring, training, and new projects.

Vistra has the potential to expand its revenue and share its AI solutions with other utilities and consumers, despite concerns about giving away competitive advantages. The power sector’s AI market is projected to reach $5.3 billion in 2024, offering a significant opportunity. For Vistra, winning even a third of this market could result in a substantial 13% increase in top-line growth, given their 2022 revenue of approximately $13.7 billion. Furthermore, utility companies can be incentivized to share best practices for emissions-reduction, aligning with national and global climate goals. In their 2022 Energy Efficiency Report, the IEA describe energy efficiency measures as the “the unambiguous first and best response to simultaneously meet affordability, supply security and climate goals [of governments].” Partnering with local or federal public institutions would allow Vistra to share project costs and risks while assisting these institutions in fulfilling their commitments.

Vistra faces several challenges in the quest to harness AI for energy efficiency. Competition is a significant factor, as other utility companies like NextEra Energy, Inc. and Duke Energy are also embracing AI and machine learning in their operations. Additionally, there’s the concern of diminishing marginal returns in efficiency investments, meaning that Vistra must continuously find new applications for their AI technology. Finally, another hurdle is the growing awareness of AI and data privacy. As consumers become more conscious of data usage and privacy concerns, it could become harder and more expensive for Vistra to access specific consumer-level insights needed for their algorithms. While Vistra already owns a significant amount of its data as a utility company, ensuring access to all required information in an environment of heightened data privacy concerns may pose a challenge. Despite these obstacles, the potential for AI-driven energy efficiency remains promising for Vistra.


TikTok – Transform Entertainment with AI


AlwaysAI: a Practical Computer Vision Solution to Optimize Operations

Student comments on Using AI and ML to Improve Energy Efficiency

  1. Really interesting post, Eleanor! Main question I have for you is what the overall impact of this tech can be for the energy industry. Is this going to move the needle, particular in terms of affecting carbon emissions, or is it a band aid over a gunshot wound?

  2. Thank you so much for this informing post Eleanor! I have always been thinking that there must be some AI implementations in the energy industry and there it is! It is amazing to see how much reduction in carbon AI could provide. However, I am concerning about the cases when AI made wrong judgements that post the company with significant risks and how could they mitigate that risks specifically.

Leave a comment