The Smarter Grid: Expanding Impacts of AI on Energy Distribution
How exactly can AI help distribution utilities with proactive grid management?
AI is all the rage these days - and with good reason. ChatGPT put a UI on an LLM. This “simple” act opened the flood gates to millions of people, allowing human-to-computer conversations that put Siri and Alexa to shame.
When it comes to clean energy, conversations with AI don’t mean much. Sure, customer service and energy savings programs are huge opportunities, but these are not primary drivers of the clean energy transition.
So, what are the primary drivers, and where does AI fit in to this? Let’s take a look.
Getting to Basics - Grid Digitalization
The foundation of AI is data. In order for AI to act on data, there needs to be two-way communication. Smart meters, powerline sensors, intelligent streetlights, and other key grid components need to be networked for communication and data sharing so AI has the opportunity for automated decision making.
Thus, grid digitization is the first step to implementing AI. According to Bloomberg’s New Energy Finance:
Digitalization, which helps improve and extend the utilization of the grid infrastructure, represents 24%, or $5.1 trillion, of total investment to 2050. Most of this goes toward implementing automation and control of the power system or to increasing monitoring and situation awareness.
The key driver of grid digitization is the meteoric rise of renewables. Managing large amounts of variable, weather dependent load requires new tools - AI is the key to making this work and preventing grid destabilization.
Build More Transmission Lines!
Interconnecting existing grids through transmission and distribution lines is one tool to send energy where it is needed any given moment, but permitting in the United States is burdensome and often takes 4-5 years before projects are even approved for construction. Permitting reform is a hot topic and beyond the scope of this article, and I expect things to change soon, but grid digitalization will run in parallel to this effort and work within existing distribution systems until regulatory changes are in place.
Distribution Management
The majority of grid digitalization efforts fall under Distribution Management and is often referred to as ADMS - Automated Distribution Management System. Energy folks love our acronyms (there’s more to come)! An ADMS is a suite of tools designed to efficiently balance supply and demand, ensuring grid stability while minimizing the need for human intervention.
Let’s look at some of the components of an ADMS and how AI can play a role in strengthening each component.
Outage Management System - OMS
Managing a functioning grid is challenging enough, but how do you respond when the system breaks down leading to power outages? This is where an OMS comes into play.
An OMS detects and locates system faults to manage power outages efficiently. In today’s world, most utilities know very quickly when there’s a large-scale outage (or even a single premise outage) without a customer having to make a phone call. This is accomplished via Smart Meters changing status in the head-end system and sending a “last gasp” message indicating the power has been lost.
This information is then fed into the utility’s OMS, which analyzes and triages field response efforts to address the outage. The response from the utility depends on the situation - in some cases a tree may fall on a line and cause an outage to dozens of customers. In other cases, extreme weather conditions can lead to entire system instability.
In the wake of Hurricane Irma, Florida Power & Light was able to use an OMS to restore power to millions of customers within a week thanks to their grid modernization efforts. The data provided by the OMS was crucial in developing a plan to address the outages quickly and efficiently.
Enhancing an OMS with AI
The best way to manage large-scale outages is to “predict the future” and obtain as much information as possible about the potential impacts of extreme weather events before they occur.
DeepMind has developed an AI system that can forecast the weather by swapping the physics equations normally used for the task with a type of machine learning algorithm known as a generative adversarial network.
The result? From VentureBeat:
In a paper published in the journal Nature, meteorologists gave an AI model for predicting short-term weather events top rank in terms of accuracy and usefulness in 88% of cases. It marks the first time professional forecasters have expressed a preference for a machine learning-based model over conventional methods…paving the way to new weather forecasting approaches that leverage AI.
Weather forecasting is a common theme with grid digitization, not just in managing outages but predicting the output of renewables. I'll touch more on weather below.
SCADA
Supervisory Control and Data Acquisition (SCADA) is like a remote controller for all kinds of devices in a power grid. It continuously collects information about what's happening - in real time.
Imagine a control room where operators can see live information from all parts of the system on their screens, sort of like how you might check the battery level on your phone or the amount of fuel in your car. If something is wrong or needs adjusting, they can make changes from their computers without having to go out to the physical device. That's what SCADA enables them to do.
Enhancing a SCADA System with AI
One of the key advantages of AI is its ability to process huge amounts of information that would take humans months or years to understand. This is the intervention point in a SCADA system: the ability to analyze massive datasets and turn that data into actionable intelligence quickly.
For instance, if a voltage drop in a particular section of the grid occurs repeatedly around the same time, AI can identify this pattern. If transformers typically fail after a certain amount of voltage fluctuation, AI can identify when a transformer is showing these patterns and suggest maintenance before a problem occurs. This approach can prevent blackouts and save costs by preventing failures and reducing unplanned maintenance work.
This is particularly relevant when it comes to system optimization. AI can analyze the operational data and identify ways to improve the reliability of the system. This might involve improving energy distribution during peak times or recommending changes to operational processes. For example, AI might analyze patterns of power use and suggest reconfiguring the grid to better distribute power during high-demand periods.
DERMS
Within the grid modernization toolkit is the Distributed Energy Resource Management System (DERMS). A DERMS is a software platform, meant to embrace and manage the increasingly decentralized nature of power generation. (Distributed Energy Resources = DERs).
As renewable sources like rooftop solar panels and wind turbines become commonplace, they change the traditional power distribution model. The energy no longer flows in just one direction. Instead, "prosumers" generate their power and feed the surplus back to the grid.
A DERMS oversees this two-way street of energy flow, optimizing the use of these energy resources. For example, on a sunny afternoon when a multitude of solar panels are producing more power than is needed, DERMS can direct this excess energy to storage units or to areas of the grid experiencing high demand. Many utilities are already using DERMS to manage these resources. Austin Energy, for example, has managed to integrate various distributed energy sources which has been critical in ensuring peak demand is met during heat waves.
Enhancing DERMS with AI
Harnessing the power of AI can significantly boost the performance of DERMS. Take the challenge of managing energy flow from various distributed sources, considering variables like weather patterns affecting renewable energy generation, varying consumer demands throughout the day, and the health of grid equipment.
Machine Learning algorithms can thoroughly analyze these data points and make near real-time decisions to dispatch energy where it's most needed.
Moreover, AI can predict moments of peak demand and assess in advance whether it might be too cloudy or windless for renewable energy sources to meet this demand. Such predictions can help the grid to prepare, ensuring a reliable supply of power and optimizing the use of renewable energy sources.
Furthermore, AI can also contribute towards predictive maintenance and anomaly detection, similar to its use with the OMS and SCADA systems. For instance, if data patterns suggest a solar panel's performance is declining, the AI could raise a flag for maintenance — ensuring the smooth operation of not only the panel but the overall grid.
Virtual Power Plants
DERMS provides the software platform and data analytics needed to manage and control DERs. On the other hand, Virtual Power Plants (VPPs) aggregate power generation and storage solutions from multiple sources spread across geographies by connecting them virtually, making them act like one large-scale generator.
Renewable generation's reliance on weather conditions makes predicting exact supply schedules challenging, leading to potential penalties for deviations from committed schedules.
VPPs offer a solution by aggregating supply from multiple plants and harnessing AI tools for more accurate weather and demand-supply predictions. As a result, VPPs can effectively minimize deviation penalties for individual power plants, ensuring better overall grid stability.
Weather Forecasting
Weather forecasting is foundational for balancing a grid full of renewables. By 2050, 50% of the world’s electricity is expected to be generated by wind and solar.
The rate of renewable growth is outpacing the growth in technologies to integrate these variable sources into the grid. The ability to accurately forecast supply from these sources is crucial to a stable grid.
IBM’s Environmental Intelligence Suite
IBM offers a platform that uses AI for weather forecasting, air quality, and water management. Their platform combines historical forecasts, observational data, and machine learning to increase the accuracy of forecasting. Here are some of the results:
15% to 30% improvement in wind and solar forecasting accuracy over publicly available weather models.
92% accuracy for day-ahead wind and solar forecasts
DeepMind - Wind Energy Forecasting
DeepMind has developed an AI system that can predict wind power output 36 hours ahead of actual generation with a reasonable degree of accuracy. They claim machine learning is able to boost the value of wind energy output by ~20%, compared with the baseline scenario.
The system uses a neural network trained on widely available weather forecasts and historical turbine data. Based on these predictions, the model recommends how to make optimal hourly delivery commitments to the power grid a full day in advance. This is important because energy sources that can be scheduled are often more valuable to the grid.
Distribution is Key - Catalyzed by AI
The convergence of AI and distribution management are crucial for the clean energy transition. By facilitating grid digitalization, empowering outage management, optimizing DERMS, aggregating DERs into virtual power plants, and bolstering weather forecasting, AI stands as a critical driver in achieving a sustainable and efficient power grid.
Chris Prato, creator of WattMind, combines 16 years of energy industry experience with a passion for clean energy and AI. With degrees in Mathematics, Philosophy, and an MBA, Chris offers unique insights into the world of AI-driven energy tech. Follow Chris on LinkedIn, Twitter, or Threads for more energy hot takes.