Jerod Wagman of Omnia AI, Deloitte Canada’s specialized AI group and Dataiku partner, opened his talk at the Toronto roadshow of the Dataiku 2023 Everyday AI Conferences with an overarching and multifaceted question.
“What is our energy infrastructure going to look like in the future?”
He made note of several key elements to help answer this question. “Rapid electrification poses significant dynamics to our electrical infrastructure,” he said. Even with a range of alternative energy solutions gaining popularity and increasingly stronger footholds in the energy industry, there is an ever-growing list of questions that data scientists weigh heavily when it comes to the current electrical infrastructure.
The Speed of Progress Leads to Uncertainty
He spoke of geospatial uncertainty — not knowing where modifications or additions to the current infrastructure are needed or how industry leaders could optimize capacity building. He added capital uncertainty, or how utility companies could prioritize capital given limited budgets. He also added cost uncertainty. “If we build in using our current approach, electricity rates are likely to rise significantly,” he said. On top of all this, he added another layer in the form of regulatory uncertainty. Running under the assumption that energy regulations will constantly change, evaluating what can be done to accelerate permitting for the delivery of any infrastructure improvements is an important consideration.
At the same time, Wagman noted increased investments in specific areas to address the growing pace of electrification and the possible strain that would be placed on the existing power grid. He spoke of operations and systems investments like infrastructure and generation capacity, reliability including grid hardening and predictive maintenance, smart grid technology, people, and also accounted for as-yet-unseen considerations.
With this amount of uncertainty but also shifting investment strategy, there came a need to evaluate different scenarios to meet these new energy challenges and map out the future. From this need came a unique solution.
We’re building a digital twin of the grid. A simulated, virtual version, so we can be more intelligent about how we meet some of these rising challenges through a range of different capabilities.
-Jerod Wagman, Senior Manager, Data Science, Omnia AI, Deloitte Canada’s AI Practice
Virtually Identical Twins
Wagman described the innovative solution that he and his colleagues at Deloitte came up with, calling it a “platform upon which we’re going to build a series of use cases that we’ll be delivering with our clients to try to make an impact on this space.”
He then dove into the components of this simulated twin, starting with trust. “How can we do this in a repeatable way so we get reliable answers?” At its core, the system would need to be utilized by multiple stakeholders across potentially multiple industries, so consistency of performance is critical, proving that not only would the system work, but also provide solutions transferable to the actual electrical grid.
He also spoke about the speed of innovation.
We’re going to need to do this at a pace that hasn’t been seen before. We’re going to rely on our grid more and more to power technology. We’re going to need to go faster.
-Jerod Wagman, Senior Manager, Data Science, Omnia AI, Deloitte Canada’s AI Practice
Utility Data: Pivotal, but Siloed
Like many other industry leaders wanting to execute wider AI and ML implementations, the key data challenge appears: The problem isn’t a lack of data; it’s accessing it. “I was happy to hear that some energy regulators are moving to the cloud and are already there. But a lot of the key data that is necessary is on premises, so we can’t access them. How are we going to securely work with this data when we don’t know what the data looks like?”
His solution is to cast a wider net, making insights provided by the virtual grid more widely applicable. “We’re going to have a wide range of use cases to enable a number of capabilities,” he said. By selecting specific deliverables across an entire value chain, from energy generation to energy consumption, the virtual grid can be universally practical, providing insights to enable data-driven decisions across multiple industries.
This approach also allows the virtual grid to be adaptable over time, a hallmark of well-designed AI solutions. “One example is the electrification problem we were talking about,” he said. “We work with a number of clients that can do forecasting today. They’re aware of changes coming, particularly from electric vehicles, heat pumps, and solar panels.”
Wagman’s approach is again leaning on strength in numbers to solve this problem. “We’re working with a consortium of different partners to offer help.” This way, intelligence is built into the entire value chain, building in AI and ML into every step, from the energy production level, the transmission level, the server level — all the way to the wires that go into customer homes. “The models can help figure out where those technologies are going to show up,” he said. More importantly, the models can account for them.
By understanding this, we can find congestion where it doesn’t exist today. We can provide integrated resources planning by not only understanding what exists today, but what might exist tomorrow.
-Jerod Wagman, Senior Manager, Data Science, Omnia AI, Deloitte Canada’s AI Practice
Game-Changer: Teaming Up With Dataiku
The data security and technical interconnectivity that serve as the foundation for the digital twin remain challenges, but Deloitte’s partnership with Dataiku has provided critical solutions, according to Wagman.
“A lot of the data is on systems we can’t access,” he said. “We need to deliver this on the utility’s own data centers. We can’t do this using cloud technology.” He acknowledges that this means dealing with dozens of different datasets across organizations, but also stored in different kinds of data systems — a very complex environment to work in.
Dataiku, combined with Fortex AI, has provided Deloitte with their own safe, secure platform environment as a sandbox. From there they can, “work with clients and partners to understand environments where we’ll need to be able to work and any data security concerns.”
By building on a Dataiku foundation, Wagman’s team has built custom environments for clients that have all necessary security in place, but they also have the option to bring it into a client’s data center, providing true lift-and-shift capability. This allows their implementations to be much simpler for clients to use and adjust over time. “This becomes a data configuration exercise instead of customers asking how they’re going to change their Python and SQL to work.”
Instead of complex migrations that could take us weeks or months because of the different data systems involved, we now have a place where we can manage this all in one tool.
-Jerod Wagman, Senior Manager, Data Science, Omnia AI, Deloitte Canada’s AI Practice
With all of the uncertainty in the space around infrastructure, cost, and technology, Wagman sees opportunity. A secure data environment built with Dataiku has truly been a game-changer for Deloitte as they work with clients looking to build a future founded on electrification. It might also be argued that the steady growth of AI and ML in the utility industry provides insight into other industries that are powered by electrical grids, too.
Perhaps the “digital twin” model might lead to a family of similar implementations, pushing generative AI into other industries where these kinds of solutions haven’t yet been deployed or are slow to implement. The pace of technological advancement in all industries requires equivalent advancement in AI and machine learning, providing a roadmap for answering new questions and helping provide data-driven insights for the future.