Tuesday, February 8, 2022

Navigating Complexity

Share this article:

An EPRI Incubatenergy project highlights the benefits of simulations in improving grid resilience in the face of complex disasters.

One of the most important lessons Gary Rector learned during his time as a wilderness guide was that accidents are rarely the result of just one factor. “We focused on multiple events,” said Rector, a data scientist with the Arizona utility Salt River Project (SRP). “If there is an accident in the wilderness, it is almost always due to multiple causes, like human error, the environment, and equipment failure. It’s the combination of two or more things that causes the problem.”

That insight was one reason Rector was eager for SRP to participate in one of EPRI’s Incubatenergy Labs 2020 pilot projects. The project brought together EPRI, SRP, and Edmonton, Canada-based technology company RUNWITHIT Synthetics (RWI), which combines artificial intelligence (AI) and machine learning with large and diverse sets of data to simulate complex events. For instance, in the past, RWI simulated an earthquake in Santa Clara, California, and demonstrated how sensors could be used to detect gas and water leaks, identify if a utility pole was down, and facilitate accurate disaster communication with citizens.

In this project, RWI was tasked with modeling the potential impacts of a 7-hour power outage during a pandemic on a 15.6-square-mile district in SRP’s Phoenix and Tempe, Arizona, service territory. The goal was to simulate the impacts of these two simultaneous disasters on the 18,249 homes and 2,000-plus businesses in the area. SRP sought a better understanding of how simulations can improve utility decision-making and planning to prepare for and respond to disasters and improve overall grid resilience and outage management.

“We plan for outages, and we understand a bit about predicting outages due to asset failures, which usually happen with transformers and distribution lines,” said Rector. Like wilderness accidents, the complexity of outage response and grid resilience flows from the multitude of forces—everything from weather conditions and their impact on grid infrastructure to people’s decisions and priorities—that intermingle and evolve during an event.

The scenario examined in Phoenix underscores just how complex these dual disasters can be. Phoenix in July is scorching hot, with an average high temperature of above 106°F, according to the National Weather Service. While such a high temperature alone can cause grid equipment failures, Phoenix can also flood during the summer when short bursts of torrential rainfall hit ground so baked by the heat that it cannot absorb the water. In fact, Phoenix set a record for most rainfall during a single July day this year, when 0.8 inches fell on July 23.

The experience of COVID-19 has shown that a pandemic alters lifestyles and work patterns, changing energy consumption habits and adding even more complexity to the situation. “There were lots more people at home and very different load curves for certain neighborhoods, and that can affect transformer health,” said Rector. “We had difficulty putting our arms around that. We want a predictive model that says, ‘Replace these transformers before they fail’ or ‘These distribution or transmission lines are likely to fall in the next big storm.’ Some of that we can do, but to pull all of these multiple events together as a data scientist and allow for predictive modeling is extraordinarily difficult.”

Grid Relience Complexity

Using Synthetics to Model Reality

RWI was one of just ten startups chosen from more than 130 applications to participate in the 2020 EPRI Incubatenergy Labs Challenge. The challenge provides opportunities for startups to demonstrate innovative technologies that can benefit everything from grid resilience and reliability to energy management and decarbonization. Other startups selected in 2020 included Kognitiv Spark, which developed an augmented reality tool to digitally connect utility workers with subject matter experts to teach them new skills and help troubleshoot problems, and Sharc Energy Systems, which provides wastewater heat recovery systems for multifamily buildings.

EPRI and a group of utility advisors from SRP, American Electric Power, Ameren, and Xcel Energy provided RWI with weekly guidance and industry expertise throughout the 10-week project. “We sought to establish parameters for RWI’s modeling,” said Omar Siddiqui, an EPRI senior program manager who worked on the SRP project. “Since RWI had no prior utility experience, we advised them on what types of data would be key for their analysis.”

Initially, SRP planned to provide information about customer load curves, the location of grid infrastructure, and municipal emergency response services within the specified modeling district. However, the project’s aggressive schedule did not allow sufficient time to go through the necessary protocols that safeguard certain types of utility and customer data.

That wasn’t a problem for RWI, though, because the company is an expert in simulating complex events through the use of synthetics. “Synthetics are AI-based entities that react, respond, learn, and are context-aware,” said Myrna Bittner, CEO, and co-founder of RWI. “They can be models of people, businesses, consumption, devices, controls, weather, all types of infrastructure or events.”

RWI had to quickly collect large amounts of publicly available data to create highly accurate and localized synthetics. “RWI used Google Earth and public documents to construct a representation of the utility infrastructure and other critical infrastructure in the area,” said Siddiqui. To better understand load curves during the pandemic, RWI used data from a study by the Smart Energy Consumer Collaborative, which tracks consumption trends and projects future consumer energy investments in technologies like solar photovoltaics and electric vehicles based on customer demographics.

To help understand the residents and businesses in the area, RWI assembled census, property, and taxation data from Maricopa County and health information from the Arizona Department of Health Services. “We also dove into data and sales reports of generator companies to see which houses and businesses would be likely to have backup generators as well as the type of fuel and size of the generation capacity,” said Bittner.

In total, RWI assembled and connected over 200 models to create a synthetic representation of this particular district in SRP’s service territory and then validated aspects of it with SRP and EPRI. “The load curves they came up with matched almost identically the ones our dispatchers used,” said Rector.

However, what was unique was how RWI could model human behavior and decisions during an outage based on health, socioeconomic, demographic, and other data. This allowed RWI to simulate the decisions homeowners and businesses would make at different times throughout an outage and also track their attitudes towards the utility and their willingness to pay for improved resilience in the future.

These individual decisions have cascading impacts. For instance, a business or homeowner with a backup diesel generator will contribute to a spike in greenhouse gas emissions during the outage but will not need to be prioritized in a utility’s grid repair efforts. Utility decisions and communication efforts can also impact customers’ decisions and attitudes. This simulation provides utilities with insights into how individual households and businesses will respond to actions taken by a utility during a dual-disaster event.

“The synthesized population is an active system during an outage,” said Bittner. “Some people are at home trying to work; some have health conditions to consider and rely on oxygen. All of these households are impacted by stress in different ways during an outage. When we turned the lights out and moments ticked by, we could see the key decision points and stress levels. Those were especially high when someone in a house had COVID or was relying on oxygen that was disrupted because of the outage.”

The Value of Simulation for the Entire Industry

The fact that RWI was able to put together an accurate synthetic environment without using utility data or collecting information using sensors and other equipment is important. It means that simulations that can guide utility decisions and investments related to grid resilience don’t have to risk data breaches or require long-drawn-out regulatory approvals.

This proof-of-concept project will help Rector make a case for the use of simulations to make better decisions. “The value is showing to management that a simulation is extremely useful,” Rector said. “Such a simulation engine would let us anticipate grid resiliency problems and let us anticipate staffing and financial problems across the business. All these areas are addressed with the same simulation technology because all of the pieces are interconnected.”

For the industry as a whole, RWI’s inclusion of the human component in resilience planning may very well be the most important lesson of this project. For example, RWI produced a heat map during the simulation that identified the most vulnerable populations—particularly those whose health could be most negatively impacted by an outage—which the utility could use to make decisions about which repairs to prioritize.

Sara Mullen-Trento, strategic issues lead at EPRI who also worked on the project, believes that a modeling and simulation approach that acknowledges the role of human behavior and attitudes is an especially valuable tool for resilience planning.

“There are so many questions when it comes to improving resilience that people don’t know where to start,” said Mullen-Trento. “This is a great way to test out research questions around how customers will respond to utility programs or whether they’re likely to adopt technology on their own. The ability to integrate not only models but also visualize that complex environment is eye-opening; it underlines the many things we need to think about as we plan for the future and the need to approach planning as holistically as possible.”

EPRI will continue to explore how synthetic simulations can support decision making for grid resilience. EPRI plans to continue working with RWI, member utilities, and other stakeholders to advance synthetic simulation as a decision support tool for a range of utility use cases.

“There are so many open questions when it comes to improving resilience planning that it’s hard to know where to start,” said Mullen-Trento. “This approach helps us answer a lot of questions and opens new avenues to see how different events and decisions play out. We want to do a lot more of that.”

Key EPRI Technical Experts:

Omar Siddiqui, Sara Mullen-Trento
For more information, contact techexpert@eprijournal.com.

Artwork by David Foster and Craig Diskowski