NRCan's Digital Accelerator team applies big data and AI to answer policy questions

Natural Resources Canada (NRCan) is using advances in artificial intelligence and related digital technologies to deliver insights to policymakers. 

The Digital Accelerator — an in-house team of data scientists, business and policy experts dedicated to building data-driven software tools — is applying AI across NRCan "where we can have breakthrough results," according to Dr. Vik Pant (PhD), Chief Scientist and Departmental Science Advisor at NRCan. 

The model is also inspiring other science-based departments and agencies (SBDAs) to think about new ways to leverage the masses of data they collect and drive a much-needed digital transformation in government.

Pant said science, including data science, has always fed the policymaking process at NRCan, which employs more than 2,000 scientists and researchers and spends more than $500 million annually on science and technology.

“What’s different now is machine learning," he said in an interview with Research Money. "We are supplementing and complementing the insights coming from traditional AI with insights from machine learning, then sharing those results and helping to answer policy questions."

Current projects include:

• analyzing geospatial data to support flood mapping and emergency management in flood zones;
• modelling forest ecosystems to improve forest resource management, and;
• forecasting load on the power system and demand for charging stations as electric vehicles become mainstream.

Each one is directly tied to a departmental priority, such as completing all flood maps in Canada, operationalizing the planting of 2 billion trees over the next 10 years and installing up to 5,000 additional charging stations along major roadways. 

In an interview with Research Money, Vidya ShankarNarayan, Chief Information Officer (CIO) and Assistant Deputy Minister, Information Systems at Agriculture and Agri-Food Canada, said she is watching the Digital Accelerator closely as she sets her department’s digital and data strategy.

“The number one thing I think we can leverage from the work Vik’s team is doing is AI—applying AI where we can. How do we leverage the data we collect on the science front to inform policy, not only in agriculture and agri-food but across the Government of Canada writ large?” she said.

To that end, she has started convening CIOs from the SBDAs for quarterly meetings to talk about issues and opportunities such as cloud strategy, data sharing and standardization, information management and governance. “The pandemic definitely showed that we are now in the digital future… and our role as CIOs has pivoted. In the past, we were looking at our own departmental systems, but now our role is to be looking enterprise-wide,” she said.

Rise of electric vehicles

In addition to applying AI in the natural resources sector, the Accelerator aims to increase digital literacy and digital capability within the department. Another key objective is building strategic partnerships with the private sector, academia and other science-based departments and agencies.

“We can't do any of the first two if we don't have a good partner ecosystem,” said Pant. “That's really the foundation of what we do.”

The Accelerator has collaboration agreements with Google and Microsoft and partnerships with CIFAR, MaRS and several of Canada’s AI institutes. The flood mapping project was conducted with NRCan's Canada Centre for Mapping and Earth Observation, the University of Winnipeg and MILA – Québec Artificial Intelligence Institute. The electric vehicle (EV) load prediction project includes several research centres within NRCan, academic partners, Transport Canada, Environment and Climate Change Canada, and Microsoft Canada. 

The latter project — to evaluate whether the power grid can handle an influx of electric vehicles as more Canadians choose to plug and drive — is an example of the real-world problems the Accelerator team is focused on. The results will give policymakers and electric utilities the knowledge they need to prepare for the rise of electric vehicles, guiding decisions about how to reduce load and optimize the use of public charging stations.

Electrification of the transport sector is multidimensional with technical, economic and social dimensions, said Dr. Evgueniy Entchev (PhD), a senior scientist who works at NRCan's CanmetENERGY Research Centre. “Modelling this complex system is extremely difficult. With the Digital Accelerator, we are employing all these wonderful artificial intelligence tools, which are giving us some of the answers to forecast the load on the grid," he said.

Data scientists at the Digital Accelerator are analyzing more than a million data records capturing charging behaviour for a fleet of electric vehicles over the course of a year. They are trying to understand when and where Canadians are charging their vehicles and how variables like charging rates, weather and vehicle range influence charging behaviour as well as trying to predict the demand this behaviour will place on the grid and charging infrastructure.

Dr. Alexei Lapouchnian (PhD), a digital solutions architect at the Digital Accelerator, has been working on the electric vehicle grid readiness project for more than a year. In an interview with Research Money, he said having a dedicated budget and compute capacity allows the Accelerator to be efficient.

“We’re able to come into a project, help, and hopefully transfer knowledge…. Many of the researchers we work with have deep domain expertise and some expertise in building models for data analysis. But they lack the knowledge about modern, state-of-the-art tools. One of the things we’re trying to do is empower them," he said.

Challenges include recruitment and measuring impact

Despite a string of early successes, a deep project pipeline and widespread departmental support for the Digital Accelerator, it does face several challenges. A major one is recruiting talent at a time when data scientists are a hot commodity inside and outside government. Pant said the Accelerator has received a “tsunami” of proposals, but the team's size is limiting how many projects they can take on. 

The problem is not insurmountable. Lapouchnian himself was a researcher and instructor in computer science at the University of Toronto before moving to NRCan, which was his first job in government. He said he was drawn to the Digital Accelerator because of Pant’s thought leadership and the opportunity to use AI to tackle problems of national importance.

Another challenge is measuring impact. Individual projects have a well-defined set of deliverables, but it is harder to measure success at changing departmental culture, Pant said. “As a team of data scientists, the first thing we want to do is say, is this working?” he said. “But it’s hard to put a specific metric on something like culture.”

Scaling the solutions coming out of the Accelerator beyond the prototype or proof of concept stage may also prove a challenge. However, Pant said they have tried to engineer the organization of the Digital Accelerator to avoid the issue. When it comes time to scale, projects move out of the Accelerator to NRCan’s Chief Information Officer (CIO) and Chief Digital Officer.

“If the client who's a scientist or a policy person inside the department likes what we've done and they want to scale it, then we seamlessly hand it over to the CIO branch. They are experts at taking systems and scaling them in a reliable, performant way," he said.

R$