How can AI and Analytics help the US Power Grid more resilient

Recently, at an innovation summit event hosted by Pacific Gas and Light (PG&E), Elon Musk predicted that in the next 3-5 years the US power grid will see more disruptions and outages due to increased electric vehicle (EV) sales as well as the increased use of Artificial Intelligence (AI) and Machine Learning (ML).

Large Language Models (LLMs) are characterized by their large size and are enabled by AI. LLMs are trained on large data sets and enable technologies such as ChatGPT, Google Bard, and Microsoft Bing AI. This training depends on large, distributed data sets that need to be accessed and processed, and which drive network traffic, computer resources, and storage – putting strain on the electric grid that powers all these information technology (IT) resources.

How can we avert a possible disruptive situation? In other words, how can we use AI and ML to improve the stability of the electric power grid? What are some ways AI and analytics can assist in helping to reduce outages and perhaps even to avoid outages?

AI and related technology have a great potential to assist the power grid.

Before looking at the power generation side of this equation, we should look at the consumer side. Many consumers already own smart and energy-efficient devices such as power outlets, dimmers, light-emitting diode (LED) lighting, and smart thermostats. However, consumers still need meaningful real-time information that can be acted upon that also explains the impact to the consumer of those devices and actions taken.

As an example, let’s say a consumer has a refrigerator that is in an insulated garage during winter when it is below freezing outside. We can assume that the power grid provider expects outages to occur due to freezing weather possibly causing ice build-up that snaps power lines. This provider may choose to send an actionable event to this consumer stating that they would like to cut the power to the refrigerator for four hours. The impact will be minimal (as the garage is insulated) and the temperature in the refrigerator will remain at the desired levels due to the temperatures outside. Therefore, the consumer impact in this case is minimal, but as an incentive the company may choose to grant the consumer five times the average usage of energy of the refrigerator for the complete four-hour window, even if they did not need to cut the power. This allows the consumer to make an informed decision based on personalized data.

Now if we look at the power generation and distribution side, we see these companies will need to provide customers with real-time data collection capabilities to enable AI and analytics insights to charge their devices, cool their homes, charge EVs, etc. Deploying AI technology allows these companies to provide real-time predictions of anticipated power grid usage and utilization ahead of any major event, whether it be weather or special events (such as a concert, sporting event, or other large-scale gathering). Additionally, these early predictive indicators can help drive preventative maintenance so that scheduled repairs can occur during low-usage time periods. This type of advanced modeling will provide the grid operators with data, insights, and actions to plan for and minimize power outages while also increasing overall power grid reliability.

This type of insight at the grid level can be translated into individual consumer actions as we saw in the example above; however, more than 70% of the US power grid is more than 25 years old, so in order to get this level of detail on the anticipated grid performance we need to augment the grid with intelligent devices, Industrial Internet of Things (IIoT), and other smart technology. The grid operators will need to be able to collect and process data on power needs in real time and use AI to gain more accurate and timely insights, actions, and recommendations regarding how much power to generate, how to improve efficiency, and how to minimize or prevent outages.

While AI, ML, and LLMs do drain the power grid, these same technologies can and should also be put to use to make the grid more resilient.

?

?

This is a joint article by René Aerdts and Hemang Davé. Please note that the views expressed here are ours only, and do not represent Kyndryl’s official positions. We would love to hear how you would approach the topic of the US power grid and how we can utilize AI, ML, and LLMs to help stabilize it. Please feel free to share this article with others in your network.?

?

?

?

?

?