Now is the time for electric utilities to make a transformative change in their operations. By implementing data science and analytics as the driver of strategic thinking, utilities can reduce traditional reliance on the subjective and local and gain a more profound understanding of our grid as it truly operates.
Utilities are the ultimate data generator – our operation of the system creates valuable daily grid data through outage records, asset records, work orders, maintenance notifications, device operations, and more. Traditionally, utility strategy has centered on the collection and storage of data, but not on the analysis and mobilization of it. At Avangrid, we are modernizing our approach to data – we’re transforming our foundational data infrastructure, so it is primed to maximize a vast, yet previously untapped, resource. We’re doing so by integrating data analytics and data science into our daily operational and strategic methodologies, shaping the future of utility data policies.
For utilities, it’s necessary for us to reprioritize data as the critical resource. Our ultimate mission is to provide safe and reliable electric service to our customers. This can be realized through the representation and understanding of the electric grid through data. For example, utilities have information on the age, health, risk and location of each individual piece of infrastructure on the grid. Investment decisions about the grid’s millions of disparate assets, and the millions of variables dependent on these assets, can only be done through a mature analytical framework. Moreover, our data is not limited to asset location information; we have outage data, SCADA data, smart meter data, and much more. The complexity of our questions, and how we answer them, will only grow as we gain access to more datasets.
However, the prioritization of investment must not only be driven by but also communicated through data. To truly evaluate the risk to our grid, we must look at each asset as an individual point of potential failure. With this in mind, how does a utility quantify and manage the associated risk of each point? The infrastructure is the still the primary asset, but it is the data and corresponding analytics that bring operational and decision-making value. We can then exponentially increase the level of complexity when evaluating risk on the system by expanding data-driven decision making across all assets. Pairing disparate asset types with information like historical outages, customer density and load will give us a systemic view of risk. To take it one step further, we can add in external factors, such as vegetation density, weather, and motor vehicles. When considering all these potential factors, how could a utility ever effectively understand each reliability and resiliency point of risk at scale and manage the necessary investment without data?
THE INDUSTRY’S HISTORIC DATA CHALLENGES
Unfortunately, taking advantage of a utilities’ vast operational data is not always simple. Electric grids are incredibly complex systems that were built decades before data was mobilized as a critical resource. As a result, large swaths of operational data are hidden. Further still, data can be aggregated or stored in a series of Excel spreadsheets or broken, patched together databases that are decades old. The challenge of this structure may seem overwhelming, but the solutions are readily achievable by the utility.
Another challenge is ensuring that every level of the organization sees the value of data—from the CEO all the way to the person entering each record. After all, if a worker puts the fault location of an outage in its record and the fault location is never used, there is no value in them entering that information. Likewise, if this worker does not put the fault location in an outage record, who is there to catch it? If this missed data point is caught by some studious supervisor, do they correct the outage record as a “check-the-box” practice or because the worker and supervisor see the value of quality data for analytics? Unless the data is being used in ways that the operational personnel clearly see and associate with their work, that connection is not nurtured. At Avangrid, our Operations leadership has extracted tremendous value from the applied analytics of our Operational Performance team, not only demonstrating the “why” of data quality as a result of personal ownership but driving future data policies through a shared vision of what is possible.
Once a culture of prioritizing and valuing data is established, the next question that must be answer is who bridges the gap between generation of data and its operational benefits? Utilities take different approaches to this question with varying degrees of success. There are the utilities where the answer is a series of analysts who do not have much control over the systems they operate, but instead battle manual processes to their reports. This approach bottlenecks the utility’s reporting capabilities causing only cosmetic changes to be made, such as a different visualization or another slice at the data. They are not able to ask deeper questions. For example, we know what our SAIFI is today, but do we know why? We know our outage statistics around tree contacts, but do we know how to use these statistics to give us operational insight or rethink strategy? We need to dive deeper into the data, requiring it to be well-structured and thoroughly understood. Though analysts may understand the data, they do not necessarily have control of the structure.
Other utilities take a different approach, expanding their capabilities by bringing in departments that do have underlying structure control—they understand the data systems, but they do not fundamentally understand what the data means. So, they pair an electrical engineer with an analyst to generate value for operations out of the data. This reaches the end state of data analytics for electric utilities. We have reports, we have data collection systems, and we have some engineering benefit. Anything more ambitious will require a purchase order.
At Avangrid, we have made the ambitious use of data core to our business and data science an integral part of our own internal operations through the Operational Performance organization. Our data scientists and data engineers fundamentally understand what the data represents by working daily with line supervisors, arborists, dispatchers, and power engineers. Data engineers who understand how a substation is designed is critical for developing substation asset tables and data scientists who understand the variables that lead to the degradation of a circuit breaker is critical for developing AI models around predictive failure. The scientist who develops machine learning models sits next to the engineer designing underground systems, and they learn how to apply the other’s knowledge to their work for the benefit of our customers.
REIMAGINING DATA SCIENCE IN OPERATIONS
In the past, utilities have placed an overt reliance on external vendors to manage their data. This lack of internal investment in data science as a core business for electric operations has resulted in a profoundly weak data culture throughout our industry. But this industry atrophy is unnecessary– the utility generates most of the data itself, the data processes are the utility’s, the subject-matter experts are the utility personnel, and the decisions that need to be made through data must be made by the utility. The requisite skills to do the challenging data work are not confined to well-coiffed lawns in Silicon Valley - if anything, people with strong data science, data engineering, and analytical backgrounds may be more abundant than strong power engineers with a utility background. We have recruited incredible talent from leading technology universities right in our own backyard in Western New York.
To transform decision-making for today and tomorrow, it is foundational and necessary for a utility to invest in a data science organization that is aligned with the needs and priorities of its electric operations. Data engineers can build well-designed, well-scoped database infrastructure around datasets such as reliability, asset, and geospatial data, so that depending on the question, an analyst can put together an answer in minutes. Data scientists that share the priorities and needs of electric operations, who work with not only operational data daily but also with operational personnel daily, gain invaluable knowledge which improves their scientific models. For example, they learn why conductors not only have different sizes but can be covered or uncovered, or they learn to differentiate between the characteristics of underground versus overhead construction. With this collaboration among data science and operations, the utility puts itself in a position that is liberated from the “use-case” mindset that so easily limits analytic projects to a narrow band, built backwards from the end-goal with little flexibility. Instead, the utility builds capacity, knowledge, and personnel for the operations leadership to rely upon so they say, “before we make a decision, let’s take a look at the data.”
Data teams allow the utility to truly see—for the first time—the system as it is, not as we feel it to be. It transforms the subjective to the objective. Data science models for reliability, resiliency, electrification, and distributed generation will be critical throughout an operational organization to make strong decisions. The ability to answer questions and drive strategy through data is not only a necessity for a well-managed utility, but an expectation from regulatory authorities. How can we tell the story of our grid—past, present, and future—without the basic building blocks data provides? The utility that invests in data science as a part of its operating model, and not as an interesting side project to fulfill corporate goals of innovation, will be the utility that positions itself to optimize grid investment, tell a compelling story to customers and regulators, and meet the incredibly complex needs of the future. I am very proud that we are doing this at Avangrid and see our work as only the beginning of how data science can shape the grid of the future. The utility industry is a long-term industry, and the only way to be prepared for that long-term is to see it and understand it. Data science is the vision electric operations need to meet tomorrow.