Seizing the AI Advantage in Utilities Now
By Adriana Karaboutis, Former Group Chief Information & Digital Officer, National Grid plc.; Independent Board Director AON & Perrigo
There is hardly a media channel you can tap and not hear about the disruption of Artificial Intelligence in any industry or geography. The opportunity is huge and the fear of the unknown relative to the power of this capability dominates a plethora of conversations. However, AI is here and adopted responsibly to help utilities in multiple ways: optimize energy distribution, predict equipment failures for improved and proactive maintenance, improve operational efficiency, raise the level of customer satisfaction through better engagements and interactions, and improve resource allocation and management. By examining past cases of funding approvals and denials by regulators, AI can even help identify mainstream and nuanced criteria that regulators prioritize –and by further leveraging economic and political datasets, utilities can better predict outcomes under various regulatory scenarios. In short, most processes, decisions, or interactions that utilities manage can be improved through the ingestion and modeling of vast amounts of data, a.k.a. big data. I think about this as a huge opportunity for ‘augmented intelligence’ (versus ‘artificial intelligence’). This is because humans are integral to running a successful, safe and secure utility and developing AI models to improve outcomes is also a human responsibility.
In short, most processes, decisions, or interactions that utilities manage can be improved through the ingestion and modeling of vast amounts of data, a.k.a. big data. I think about this as a huge opportunity for ‘augmented intelligence’ (versus ‘artificial intelligence’).
For utilities to realize the above benefits, it is necessary, however, to think differently about the vast amount of data they currently possess and the usefulness of this data. This forward-thinking requires utilities to operate across the enterprise (not in silos), to unarchive and expose data to a central or empowered data sciences team, and to identify questions and scenarios that must be addressed in order to achieve the next level of …resilience, efficiency, customer satisfaction…. Etc. Through an un-siloed and outcome-focused mindset, utilities will see how their data and external data can be merged/modeled to address key questions and desired outcomes. There will always be a technology or ‘system rollout’ that utilities await for next-level improvements: upgrade of smart meters, self-healing FLISR, and further voltage variation capability; however, vast benefits can be realized immediately by simply working with and valuing existing data.
How should this be done?
After defining the objectives and use cases, such as the ones above, an inventory of data sources is necessary. This includes data from SCADA systems, customer interactions, asset and maintenance logs, regulatory submissions, and rate cases, etc. Once these data sources are identified and aligned to a data model, the quality of data must be addressed. A drive to adopt a formal utility data model would be valuable to ensure quality results as AI models are trained. Currently, some utilities believe that the tomes of paper data they have are impossible to use or may be inaccurate. That challenge can be overcome with emerging technologies, innovative thinking, and an understanding of the value of the outcomes resulting from the AI models.
What technologies and capabilities make all this possible?
The onset of increased compute power (graphics processing units, etc.) that accelerates the speed of computations; availability of big data to facilitate machine learning algorithms; availability of cloud platforms that provide resources at scale for AI developers and AI applications; open-source frameworks that democratize development, as well as newer capabilities in edge computing for latency reduction and real-time processing at the source of data.
What is the biggest challenge to realizing the benefits of AI in utilities?
- Identifying the key questions, processes and outcomes desired;
- unleashing and unarchiving the data to an empowered team;
- driving innovation from the top of the house.
I know this from my experience at National Grid, where I witnessed the company make significant progress in the early phases of AI outcomes and GenerativeAI adoption by doing all of the above!