Whatever improves the employees' productivity, enhances the employer's profits. For industries, such as the electric power utility industry, a desirable form of productivity is the timely extraction of intelligence from endless streams of data across dozens of domains.
Power utilities across the world are in a constant battle against the imbalance of energy supply and demand. They engage in this tug-of-war using an arsenal of protection devices, forecasts, and operational planning, and the skill of talented engineers and crew members. In this fast-paced, dynamic game, there is nothing more valuable than the right information at the right time, available to the right person.
In recent times, the business of delivering reliable power to customers has been complicated further by infusion of unpredictable energy resources, customer-owned solar panels and storage systems, customer choice, and electric vehicles. On top of these everyday challenges, the increasing frequency of inclement weather events like Hurricane Harvey and the looming threat of a crippling cyber-attack are making utility executives lose their sleep.
While protection devices, like relays, reclosers, and circuit breakers, are the first line of defenses against an unforeseen event - planning for emergencies has always been in the DNA of utility operations. Plans made by utilities range from estimating energy demand to strategies and protocols for restoring downed lines following a significant storm.
Such contingency plans are made in advance, by a team of experts. At the moment of crisis, circumstances almost always change beyond the variables considered in the planning models. The field operator, or the first-responder, has limited capabilities to make last-minute changes to the model based on his latest insights to make the best possible decisions.
The fact that Artificial Intelligence can help decision making in a data-driven industry like the power utilities is not a secret. Many utilities are actively using AI to solve problems related to decision making or demand forecasting. Yet, most such solutions stand in isolation to one another, resulting in a lack of reusability, and heavy dependence of internal or consultant R&D teams for leveraging the power of AI models. On the other hand, there are several utilities in the US and the rest of the world, who cannot afford the human horsepower needed for running the AI engines.
This silo-ed way of using AI also limits the capabilities of the number of people who can use or contribute to the models. Underengagement with the AI model, in turn, undercuts the performance of the AI, and it runs its course in a vicious downward spiral, till the AI solution is rendered unacceptable, and a new silo-ed solution is rebuilt. Often, the same AI model is used for multiple years before it gets an upgrade because of a prevalent belief that a reliable model is worth more than the most recent one. (we aren't saying its wrong). However, it leads us to contemplate what keeps AI in silos?
We hypothesize that the sophisticated metrics churned out by the ML models, cloud the reasoning by which a model arrived at a decision. Fundamentally, the human factor is missing. A dynamic AI-on-demand platform can transform the tacit intelligence of complex machines into better story-telling and empathetic listening partners, rather than byte-gurgling black boxes.
A correct AI solution for large organizations, in essence, should not just focus on intelligence extracted from data exclusively. Instead, subtle, non-numeric, purely human experiences and expectations should drive the AI.
Let's consider a few examples of human collaboration in intelligent environments. People in vegetation management may have some observations, which might contribute to outage prediction models - yet, they cannot directly interface their field intelligence to an expensive locked-up AI model in the OMS. An economist may have some insights into the load demand of feeders in specific service areas. Still, it might be costly and time-consuming to interface to code that prediction into a demand forecasting model for the fiscal year in progress. Decarbonization is a priority, but which segments will be most responsive first? Automated analysis of multiple service areas of the utility might yield a reliable answer faster and more comprehensively.
The key principle of the method is using and integrating different ways of knowing.
There are a number of benefits arising from the application of this method. Causal layered analysis increases the range and richness of scenarios; leads to inclusion of different ways of knowing among participants in workshops; appeals to wider range of individuals through incorporation of non-textual and artistic elements; extends the discussion beyond the obvious to the deeper and marginal; and leads to the policy actions that can be educated by alternative layers of analysis.
The utilities are large organizations with vast diversity in skills and experience. It runs by the coordinated operations of a variety of teams - many of which communicate with one another rarely. Despite the proven potential of AI in power utilities, it is neither sufficient nor efficient to add one isolated solution to the other. Thus, AI in utilities needs to be accessible and open to contribution as a human-centric collaboration platform. Such a platform would soak in cognitive experience across all teams and colleagues, keep models alive with dynamic input of data, and tailor AI's insights and analytics to jobs, roles, and specific queries.
Many times, decisions taken by executives and engineers alike, are made under the pressure of extremely high workload, complex compliance requirements, and long working hours. An assistive AI technology, which can use without extensive involvement of R&D teams or writing complex queries, could mean the difference between power delivered, or not. Or the difference of millions of dollars over the years. Data is enriched, and data-driven decisions are empowered by adding multiple layers of context, a task where AI excels.
'Sync' provides such a platform, which makes any utility employee a contributor to the AI modeling efforts of the organization. The platform also makes any utility engineer access any AI insight, specific to the job/task she is engaged in, at any time.
An "AI on-demand, for anybody, any time, anywhere" platform, such as Sync, can also be used to tackle another challenge utilities are undergoing - attrition and the aging workforce. Valued employees, with their decades of real-world experience, are on their way to retirement, while a new generation of engineers is joining the utility industry in a much lesser number than what's the need of the hour. The protocols of operation are teachable, but the new engineers may not be as efficient in many tasks as the seasoned worker he is replacing, only due to the lack of experience. If a collaborative system, or some form of AI, could be developed to archive the ingenuity of the human experience accrued by the retiring workforce, it would make the new engineer's work much more efficient, eventually improving the utility's bottom line.
No single individual, not even the CEO, within a large organization, knows the entirety of problem space within which the utility operates. However, a collaborative input and intelligence platform can put together a complete picture by seaming together perspectives across all domains. It can help anyone ask better questions, discover higher impact problems to solve.
Will the adoption of AI-on-demand replace jobs of engineers and R&D teams within the utility? No, a resounding "No!". Machine learning systems rarely replace crew, technicians, engineers, and scientists. Instead, it adds polish, depth, and value to their work while freeing their minds for more creative problem solving, which machines cannot do yet. And Sync's AI-on-demand platform is no exception. Sync is a force multiplier for anyone working in the utility industry in any role. By using an AI-on-demand platform, the engineer can improve their productivity, a crew dispatcher can route the fleet efficiently, and tariffs can be set more fairly. Ready-access to information and intelligent insights will lead to better employee experience and retention rates.
Millions of years ago, human civilization began based on interpersonal communication and the exchange of ideas. Eventually, when someone outside the immediate family could see a predator, they would get on top of a tree and scream a warning for the rest of the villagers. Millenia later, they would use a megaphone. Now, they might use Twitter. The means and the machines of communicating intelligence change with time, the need for humans do not. Instead, they improve productivity and progress. By recreating a fundamental human need - collaboration and intelligent communication - Sync's AI-on-demand platform is one step towards the next level of productivity for utility workers, and our transition into resilient smart cities of the future.