Utilities typically do not lack the data to make important improvements in network efficiency and stability. The data is there, but often the value of that data is lost in a petabyte-sized forest of telemetry, phasor, and business data. The software to analyze it was probably there too. What is missing is the expertise to know how to ask the right questions of that data.
Python, R, and Seeq have become standard analytical tools for utilities. But their potential to improve operational efficiency, reduce outages, and shorten root-cause analysis works only when users can work at an advanced level. Python automates data pipelines and machine learning, R handles statistical load forecasting, and Seeq makes historian time-series data searchable. Utility-specific training is what turns owning these tools into getting results from them.
Fred Knops, SVP of Solutions at Data Society, leads the company’s Energy and Utilities Sector work, which includes both custom training and software solutions. He has spent more than 30 years moving between engineering and executive roles, and he works with utility teams on exactly this problem. Today, we’re focussing on the Data Society’s training programs for utilities.
How did utilities end up running on Python, R, and Seeq?
For a long time, utility analytics lived in spreadsheets, Excel macros, and licensed desktop packages such as MATLAB, SAS, and SPSS. That worked when most operational data arrived in a daily or hourly cadence, and queries were mostly about the past. As grids added sensors and the data started arriving by the second (or fractions of a second), the limits of those tools became clear. Spreadsheets cap out, macros break when a tag name changes, and per-seat licenses make it expensive to put analysis in more hands.
Python and R filled that gap. Both are free, both can handle very large datasets, and have very capable libraries of statistical, forecasting, and machine learning capabilities. Among energy data scientists, they are now the default. When the Energy Systems Catapult surveyed data science practitioners in the energy sector, every one of the 42 respondents said they used Python or R, or both. Excel, MATLAB, SAS, and SPSS still showed up, but in smaller numbers. A systematic review of data analytics across the electricity value chain, published in Energy and AI, offers that modern analytics techniques are now applied at every stage, from generation through to transmission, distribution, and consumption, as well as trading.
Why is grid data so difficult to work with?
Grid data is hard because there is so much of it, it arrives so fast, and it is rarely in perfect condition. A modernized grid can generate billions of data points a day. Smart meters typically report usage relatively slowly — typically every 15 minutes to an hour — but there may be hundreds of thousands, or millions, in the network. Modern SCADA systems log equipment status continuously. Inverters at solar and battery sites report their output continuously, and PMUs take measurements dozens of times every second.
Almost all of this is time-series data: a value, a timestamp, and a tag that says where it came from. Time-series data is replete with issues. Sensors drop out and leave gaps. Some get stuck and report the same value for hours. Clocks drift, so two readings that should line up do not. Tag naming conventions differ between vendors. None of this is unusual, but you have to handle each problem before the data can answer a question reliably.
More data doesn’t fix this on its own. A utility can collect every reading from every asset and still miss a transformer running hot if no one has built a way to notice.
What does each tool actually do for a utility?
Python, R, and Seeq each handle a different part of the work, and strong utility analytics teams use all three.
For utilities, Python does most of the analytical automation. A Python pipeline can regularly pull readings from the historian, clean out the gaps and stuck values, combine them with other data structures (e.g., maintenance records and weather data), and feed the result into a machine learning model that scores each asset’s risk of failure. Once built, it can run without anyone touching it, which makes it possible to monitor thousands of assets at once.
Clearly, Python is a powerful tool, but having clear expertise in its use is important, and Data Society offers the tools to move utility analysts to the right level of proficiency. “While Python is often viewed as a bit less complex than R and the basics are considered easier to master, there exist many options for assembling useful and informative courses and pathways.” Data Society has created courses ranging from introductory to advanced Python, to advanced courses that address the overall environment in which good Python models live, such as LLM fine-tuning, LangGraph and ReAct Agents, and Model Context Protocol. “Every client is a bit different and has a different portfolio of questions they have an urgent need to answer, so each time we do training, it’s a bit different. We build our pathways around client use cases and tailor them to their specific engineering environment and needs.”
Fred Knops, SVP of Solutions, Data Society
Statisticians built R, and it is especially strong for forecasting and measuring uncertainty. A utility planner might use R to build a probabilistic load forecast: rather than a single number for next July’s peak, a range of likely peaks with the odds attached to each, based on weather scenarios. That kind of analysis also supports rate cases and long-range capital planning, where regulators want to see the reasoning behind a number as well as the number itself.
“When we teach topics in R, we are often asked to address Multiple Linear Regression, Decision Trees, Ensemble Methods, and related topics like MLOps.” Bonneville Power, for example, worked with Data Society to offer these types of courses to their analytical team. “But each client is different, and the combinations of topics are limitless. Data Society can, for example, include topics such as how to create coding assistants for data scientists in our training pathways.”
Fred Knops, SVP of Solutions, Data Society
Seeq arrived to solve a specific problem. Most utilities already store years of sensor data in historians, the databases that record readings from SCADA and PMU systems and equipment. Getting useful answers out of a historian used to mean exporting data, cleaning it by hand, and lining up timestamps in a spreadsheet. Seeq connects directly to the historian and lets engineers search, clean, and compare that data visually, without writing a pipeline first.
Seeq is where engineers go when something has already happened, or is starting to. After a breaker trips, an engineer can use Seeq to pull every related tag from the hours before the event, line them up, and see what changed first. The same approach works for spotting conditions that tend to come before failures, then setting up monitoring to catch them next time.
“Seeq is an interesting case. We are Seeq’s designated training partner for the energy and utilities sectors. When we train Seeq, it is often using customer use cases and can also be based on the utility’s data. So our training is quite different from Seeq’s out-of-the-box training. We really focus on making sure that our utility customers get the most out of the environment, focusing on ensuring adoption and finding efficiency gains during the class. Our Seeq training program for Dominion is a good example of this approach.” In terms of adoption, Data Society works with the client to make sure that the various departments and processes that use the tool are brought together around at least a well-defined set of analytical tools, and create a consistent approach across the organization. Efficiency is the result when Data Society works with client analytics teams, in the classroom, to automate time-consuming calculations and data preparation. Their work stayed the same, but they can now do it faster and more efficiently.”
Fred Knops, SVP of Solutions, Data Society
What’s the difference between data literacy and data fluency?
Data literacy means someone can read a dashboard, follow an analysis, and run a script that a colleague wrote. Data fluency is achieved when they can build the analysis themselves, adapt it when conditions change, and recognize when the output doesn’t make sense. Training programs often stop at literacy. Working with today’s grid data requires fluency.
The difference shows up in day-to-day work. A literate analyst can run next year’s load forecast with this year’s data. A fluent one notices that the model never saw a succession of heat waves like this summer’s and adjusts it appropriately. A literate engineer can open Seeq and plot a few historical analyses. A fluent one can build a condition that flags every time a transformer’s temperature rises faster than its load, across the whole fleet, and route it to the right crew. A literate team member can rerun a Python script. A fluent one can fix it when a vendor renames half the tags.
When a team stops at literacy, the gaps show up as blind spots in monitoring, root-cause investigations that take weeks instead of hours, and outages that the data could have predicted. When the analytical team reaches fluency, the organization can be assured of the speed, precision, and accuracy of their output.
Not sure where your team falls between literacy and fluency? Fred is happy to talk it through. Email him at fred@datasociety.com.
What should data training for utility staff include?
Most Python and R courses usually are based on retail sales data, marketing clicks, or public datasets that are not time-series data and have nothing to do with the grid. Staff finish the course knowing the syntax but still unsure how to apply it to a historian full of (for example) messy SCADA tags. The skills don’t transfer because the problems don’t look the same.
Training that works for utilities starts with time-series utility data: historian events, SCADA events, meter reads, outage records, and load curves. It covers all three tools, because real problems often cross between them. And it goes past the introductory level, into building pipelines, validating models, and setting up monitoring that runs on its own. It also helps to learn from instructors who understand grid operations, so a question about a protection scheme or a dispatch decision doesn’t stall the class.
Data Society offers live, utility-specific training on Python, R, and Seeq, from foundations through the most advanced levels. Teams bring their own operational problems into the sessions and leave with work they can use the next day. You can read more about the program on our Data and AI Upskilling page.
Where should a utility start?
Before investing in more software or more training, it helps to answer three questions.
First, which tools has your utility deployed, and which ones are people actually using week to week? It’s easy to own a Seeq license that only three engineers open once a month.
Second, where is the bottleneck? For some teams it’s synchrophasor diagnostics that drag on for days or weeks. For others it’s forecasts that miss during extreme weather, or data pipelines that only one person knows how to fix.
Third, who on your team does everyone go to when the analysis gets hard? If the answer is a small cadre of busy staff, that’s a clear indication of business risk, and where training can provide a broader pool of talent that the business can draw upon.
The answers usually point to a clear starting place: a team, a use case, and a business outcome. That is where Data Society can focus their training across all the relevant software environments.
About Fred Knops
Fred Knops is Data Society’s Senior Vice President leading Data Society’s work in energy and utilities, telecommunications, and manufacturing. He has spent more than 30 years moving between engineering and the executive suite, including as Senior VP and Partner at Booz Allen Hamilton and CEO of LendPro and. He has advised major energy, telecom, manufacturing, and high-technology firms, as well as federal agencies including the Departments of Homeland Security, Justice, and Energy.
Fred holds a degree in aerospace engineering from the University of Michigan and an MBA from George Washington University, and has completed postgraduate work at the University of Cambridge’s Judge Business School. That background means he can talk through a historian data problem with your engineers and a training plan with your executives in the same meeting. He has been an operator as well as an advisor.
Talk to Fred about your team
If you’re working through these questions at your utility, Fred would like to hear from you. Email him at fred@datasociety.com with a few lines about which tools your team uses and where you’re getting stuck. The conversation is a good way to figure out what kind of training, if any, would make the biggest difference for your team.
Frequently Asked Questions
Utilities use Python to automate data pipelines, combine data from different operational systems, and build machine learning models for tasks like predicting asset failures and detecting anomalies.
R is used for statistical work in the power sector, including probabilistic load forecasting, reliability analysis, and the econometric analysis behind rate cases and capital planning.
Seeq is an analytics application for industrial time-series data. Utilities connect it to their SCADA and process historians to investigate events, monitor equipment, and speed up root-cause analysis.
Most utility analytics teams get the most value from all three, because each covers different work: Python for automation and machine learning, R for statistical forecasting, and Seeq for historian data analysis.
Data Society offers live, utility-specific training on Python, R, and Seeq, from foundational to advanced levels. To talk about training for your team, email Fred at fred@datasociety.com.
