Advanced analytics is a collection of data analytics techniques, such as machine learning and predictive modeling, used by businesses to improve their decision-making.
The use of advanced analytics techniques is growing exponentially among electric utilities. Large public utilities, especially, are pouring sizable amounts of resources into data science professionals and tools to gain operational efficiencies and improve resource allocations.
Smaller utilities, co-ops, and municipalities, meanwhile might be wondering how and where to start.
In this article, we’ll outline some of the basics to address prior to embarking on your advanced analytics journey. Then we’ll examine some of the ways advanced analytics through machine learning can pay dividends to your organization, particularly when they help inform data-driven decisions that lead to utility asset maintenance and replacement.
Before starting your advanced analytics journey, take some time and gain an understanding of your organization’s level of maturity in these areas:
In my experience, organizations almost always believe their data is in better shape than it really is. A key prerequisite for machine learning is consistent, clean data. Take time to review or establish a data governance process so that your analytics goals can be achieved given the current and future state of your data. A common misstep for organizations is bringing on high-salaried data scientists only to have them spend most of their time performing data cleansing activities.
With all the buzz surrounding machine learning, executives can get into the mindset that it is a “magic bullet” to cure all their problems. At first, they may appear as if they are all in. But once faced with a choice to trust machine learning output versus what they’ve always done, they still might go with their gut. Analytics-focused departments need to feel empowered to change the status quo and not be vetoed by higher-ups within the organization. Additionally, proving value can take time, in some cases years. Executive buy-in should come with the understanding that analytics value might not be realized in the short term, especially with asset maintenance use cases.
Each organization is unique in how it allocates resources for people and tools. What’s not unique is needing to address the commitment to personnel and tools prior to making your investment in an advanced analytics program. An advanced analytics program can only be successful if your data is good, and leaders support the model output. It’s also worth mentioning that machine learning can have a negative connotation in some cases. People often think it means their job could be in jeopardy, which can cause friction within organizations. It’s very important to involve the people who might be impacted at an early stage to level-set goals and expectations. Analytics projects should not only streamline decision-making, but also solve employee pain points wherever possible within processes.
No organization is 100% mature in each of these areas, but a strong internal analytics community can drive improvement. Grassroots, cross-department communities increase the success of getting your advanced analytics program off the ground successfully.
Yes, contractors can be a great supplemental resource for best practices and industry standards, but internal communities unite various business units, promote best practices, and improve data literacy for all.
Utility organizations are filled with talented data-focused people in every business unit (Finance, IT, Utility Operations, etc.), you should seek these gems out and bring them together in a common forum.
Now that we’ve covered some of the basics, it’s time to dive into a common machine learning use case – Asset Health Analytics.
Here is the scenario…
Currently, your organization has a very standard process for prioritizing transformer replacement. Either the circuit runs to failure which prompts replacement, or it’s replaced in order by its age.
This existing process has been sufficient, however with customer tolerances for outages becoming narrower, there’s a higher emphasis on replacing equipment prior to failure to preset outages, thereby improving SADI, SAFI, CADI metrics. This warrants an updated approach to calculating asset health and prioritizing proactive replacement.
Ideally, a predictive model would be more effective here, but what type of model should we execute? What data do we need? A scalable solution such as IKE Insight can assist with these use cases, potentially saving your organization millions in overhead by negating the need to hire an entire data science staff and incur corresponding technical debt.
What we’re looking for is a solution that can process all of our data, determine which data is actually important to our target variable, and then run customizable rules to provide actionable output.
In addition to the predictive asset health model, we seek powerful image processing tools that can run anomaly detection on any imagery we’ve taken from our data acquisition device in the field to determine if a transformer is damaged and needs to be replaced outside the new maintenance cycle.
These two components together can yield a material improvement in KPIs for reliability, safety, and financial efficiency.
Matt Fernandez is a Project Manager and Certified Scrum Master for ikeGPS. He has eight years of experience with a large public utility, performing various roles from managing IT projects to capital planning. He holds a bachelor’s degree in Information Systems Management from Duquesne University and an MBA from The University of Akron.Learn More about IKE Insight
I’ve worked in the field collecting pole data and I’ve worked in the back...Read More
With the media’s daily bombardment of new information on artificial intelligence’...Read More
Get the latest insights on data acquisition and structural analysis from the ikeGPS experts.