Power sector explores ways in which to leverage artificial intelligence, machine learning

The electric power sector is increasingly looking for ways in which to leverage artificial intelligence (AI) and machine learning for its systems as the industry undergoes a digital transformation.

On March 18, the Electric Power Research Institute (EPRI) convened a roundtable that looked at AI and the power sector.

EPRI is leading a collaborative effort (AI.EPRI) between the electric power and AI industries. To kick off a 2021 community-building event series, EPRI hosted the virtual roundtable with executives in both industries.

Arshad Mansoor, President and CEO of EPRI, said “We’re starting a decade where the full digital transformation of how we generate, how we distribute, how we use electricity is going to happen.”

AI “will play a huge role in that digital transformation,” Mansoor said during the roundtable. “The reason this is the decade to seize on is the technology advancement that has happened just in the last five years.”

The use cases for AI are huge, he said. “There are a lot of entities – data companies, universities, technology providers – who are working with our members to shape this future.”

When asked by Heather Feldman Director for R&D at EPRI, to discuss AI use cases in the power sector, Bhavani Amirthalingam, Senior Vice President and Chief Digital Information Officer at Missouri-based investor-owned utility Ameren, mentioned asset health.

“Being able to do predictive, preventative maintenance, condition-based maintenance on key assets I think is a fantastic use case,” Amirthalingam said. “It’s great for customers from a reliability standpoint. It’s great for customers from an affordability standpoint.”

Another roundtable participant, Arun Majumdar, a professor in the Department of Mechanical Engineering at Stanford University, said that “it’s important to point out that the people who are focused on data do not always understand the issues related to the electricity industry and vice versa.”

Majumdar also said that “there are security issues with this, there are privacy issues with this. There needs to be anonymized data sets that could be used and this needs to be at scale and the scalability of organizing data and making it useful is something that the information industry understands very, very well. They have techniques that the electricity industry does not have.”

“I think the topic of data is really very central to how you make AI useful,” said Jatinder Singh, Vice-President, Digital and Data Transformation at Texas public power utility CPS Energy.

“Even if I have data, even if all of us have data, do we have the right tools and talent to do something with that data?” If not, this is one way that data sharing can help, he said.

In a Q&A with the American Public Power Association’s Public Power Current newsletter, Singh noted that he is responsible for developing strategy and multi-year roadmaps for digital capabilities that will elevate customer and employee experiences.

“The roadmap will transform our technological capabilities and the way we work – from technology projects to digital products – from waterfall to agile. In parallel to the digital transformation is the transformation of our data eco-system and governance that will allow us to capitalize on the promising newer emerging technologies,” he said.

Singh was also asked whether there are any projects currently underway at CPS Energy that involve the application of artificial intelligence to the utility’s power system.

“There are several examples of where our teams are currently leveraging machine learning to improve management insights into operations,” he noted.

“The teams are leveraging several data sources for outage prediction for seasonal weather patterns to reduce potential outage duration for customers and engage in preventative maintenance.”

For San Antonio’s wide load forecasting, the team is leveraging machine learning for better demand management. And for vegetation management, the team is at the early stage of leveraging LiDAR and GIS data combined with machine learning to be more effective and efficient in addressing potential vegetation related outages. The team is also leveraging machine learning on AMI data to solve several asset management, maintenance, and demand management use cases, he said.

Meanwhile, Singh elaborated on what he sees as the long-term benefits that the power industry could derive from the use of AI.

“There are at least four areas where I see opportunities for artificial intelligence (AI) to potentially provide meaningful benefits for the energy industry,” he said.

 One is asset management. “By leveraging image recognition and data such as GIS and LiDAR, AI will dramatically improve the industry’s efficiency to manage sprawling transmission and distribution assets, which are the highways through which our services are made available to our consumers,” Singh said.

Another benefit is in customer service, he said. Through natural language processing and voice recognition, the industry “can provide personalized, relevant, and convenient access to information and services to new and existing customers in their language and at the time of their preference,” he said.

There could be opportunities to explore how AI could help improve the efficiencies with renewable energy by combining weather data and mechanical devices that will maximize renewable energy generation from solar panels and wind turbines, he said.

Singh also mentioned new solutions. The industry “could generate new revenue sources through services and products, powered by data and AI, that will allow customers to reduce their carbon footprint and power their increasingly electrified lives.”

CPS Energy “is taking a test and learn approach towards several use cases in the above categories. However, there are some potential use cases that we haven’t yet envisioned that are becoming more apparent as the technology and its usage gain maturity within the energy and other industries. We will continue to explore those new opportunities to leverage AI as they present themselves.”

SRP

In 2019, Arizona public power utility Salt River Project (SRP) signed a deal to use AI to improve its information technology (IT) operations. The Phoenix-based public power utility adopted ScienceLogic’s SL1 platform to monitor its IT operations and applications.

In a Q&A with Public Power Current, Joe Kosmal, manager of data center operations at SRP, said that SRP completed phase one of the ScienceLogic implementation in July 2020, which was a value-added replacement of SRP’s legacy IT operations platform. 

“We are currently well into our phase two project which includes the advanced capabilities to assess the health, availability, and risk to applications and systems that serve critical business processes in the enterprise,” he said.

Kosmal was asked to discuss how ScienceLogic’s platform has improved SRP’s IT operations since its implementation.

“Since implementation, ScienceLogic has helped us regularly identify issues before they caused impact to the reliability and resiliency of our IT systems,” he said. “This has helped mature our IT operations from reactive monitoring and escalations to proactive monitoring that fosters a stronger partnership with application and system support teams in the organization.”

When asked whether SRP is considering deploying AI in any other parts of its operations beyond IT, he said SRP is exploring use cases for AI in different areas of the utility and evaluating how those capabilities can improve operations and better serve SRP customers.

NYPA

Meanwhile, the New York Power Authority (NYPA) has been awarded two $125,000 grants from the American Public Power Association’s Demonstration of Energy & Efficiency Developments (DEED) program to fund demonstration projects that will analyze the impact of ice on a hydropower plant and test an advanced technology that evaluates the health of high voltage assets in a substation.

The DEED program funds research, pilot projects and educational programs to improve the operations and services of public power utilities. 

NYPA will undertake the following projects:

Analyze the Impact of Ice on Hydro Power Resources with Machine Learning: NYPA has had significant power generation losses due to ice blockages near intake valves in the Niagara River and has worked to address the issue with industry groups and other utilities. During the winter, water can become supercooled all the way to the bottom of the river, leading to the formation of frazil ice crystals, anchor ice, or both. Anchor and frazil ice affects water availability estimation by Niagara River Control Center and frazil ice can affect hydropower plan operation since it’s “sticky” and can result in ice formation on the plant’s water turbines. These studies will include using 3-D sonar to quantify the impact of frazil ice on the efficiency of the hydropower units and forecasting the formation of anchor and frazil ice with a look ahead of a few days to a few weeks.

Smart Insulation Condition Monitoring System (CMS) for Substation Assets: A state-of-the-art Condition Monitoring System will be developed to constantly monitor the insulation condition of various high voltage assets (transformers, GIS, switchgears and cables) in a substation. The CMS consists of smart sensing, advanced noise mitigation and artificial intelligence for data interpretation. The system will use an advanced diagnostic technology that recognizes and evaluates defects and provides guidance for maintenance planning. The system will improve the power grid reliability, reduce customer outage costs, and help asset management optimize maintenance and maximize asset life.

NYPA deploys AI-based application

In 2018, NYPA selected C3 IoT to provide a software platform to help NYPA and the state implement and meet its energy efficiency targets.

Under a multi-year, software-as-a-service agreement, NYPA agreed to deploy C3 Energy Management, an AI-based application, as part of New York Energy Manager, NYPA’s advanced, secure energy management center, headquartered in Albany, N.Y. It provides public and private facility operators across New York State with timely data on energy use.

The C3 Energy Management application enables the New York Energy Manager program to aggregate enormous volumes of data, including real-time data from smart meters, building management systems, end-use equipment controls, sensors, weather data, occupancy and daylight data, solar data, and utility bills.

C3 IoT said the application would allow the New York Energy Manager program to employ machine learning at scale, generate insights about individual customers’ energy usage, and deliver personalized recommendations to help customers reduce their energy use.

The company said the software would also allow NYPA to offer its customers services such as building energy load forecasting, fault detection and diagnostics, continuous optimization of energy use, dynamic demand response, solar and energy storage monitoring, and aggregation and dispatch of buildings as distributed energy resources.

New York Energy Manager “is utilizing C3’s Energy Management application to help our customers reduce their energy costs, improve their building operations, and track and report their progress towards energy efficiency and sustainability targets,” said Paul DeMichele, Manager, Media Relations, at NYPA. The application gives New York Energy Manager advisors “visibility into customers’ energy use and expenditure, helps them identify and prioritize actions to reduce operational and energy-related costs, and reduce their carbon footprint.”

APPA receives patent tied to machine learning techniques

APPA recently received a patent related to protecting the ability of public power utilities to use machine learning techniques for advanced analytics and benchmarking to improve safety.

This is the third patent APPA has received in its work to help ensure that public power utilities have long-term access to advanced analytical technologies for business-related decision making.

While it is likely that utilities will increasingly use specialized machine learning techniques to predict and prevent outages and equipment failure, this application is focused on increasing the likelihood that maintenance actions are safe.

“Using machines to help us see patterns that aren’t obvious is a great role for technology and can help keep us safe,” said Alex Hofmann, Vice President, Technical and Operations Services, APPA.

“Through the system we have designed, our systems and workers will be able to take actions that are safer for a given situation. How many times has the weather drastically changed and line workers keep working without adjusting to the new risk, leading to injury?”

Though participation in APPA’s eSafety Tracker service, APPA is helping public power utilities work together to build and train machine learning models to predict the safety-related outcomes of planned future maintenance actions. 

Renewable energy and AI

Machine learning and AI are also being looked at for renewable energy projects.

Case Western Reserve University computer scientists and energy technology experts are partnering to leverage the diagnostic power of AI to make solar power plants more efficient, the Cleveland, Ohio-based university reported in January 2021.

The project aims to use computers to better analyze data from a large number of neighboring PV systems to help quantify their short- and long-term performance. Those machine-learning methods will be used to overcome data-quality issues affecting the individual plants.

The work, funded by a three-year, $750,000 grant from the U.S. Department of Energy (DOE), is part of a broad $130 million solar-technologies initiative announced by the DOE in 2020 including $7.3 million specifically for machine-learning solutions and other AI for solar applications, the university noted.

In 2018, DeepMind and Google started applying machine learning algorithms to 700 megawatts of wind power capacity in the central U.S. DeepMind is an AI firm and Google affiliate.

“Using a neural network trained on widely available weather forecasts and historical turbine data, we configured the DeepMind system to predict wind power output 36 hours ahead of actual generation,” DeepMind noted in a 2019 blog. “Based on these predictions, our 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 (i.e. can deliver a set amount of electricity at a set time) are often more valuable to the grid,” DeepMind said in the February 2019 blog post.

“Although we continue to refine our algorithm, our use of machine learning across our wind farms has produced positive results. To date, machine learning has boosted the value of our wind energy by roughly 20 percent, compared to the baseline scenario of no time-based commitments to the grid,” it said.