SOLVER: Transmission Suite
Reduce cost of maintaining the transmission grid while improving its reliability.
Through a series of proof of concepts, apply predictive data science techniques across the business to find measurable value in the data.
A prominent East Coast electric utility faced a common problem: it wanted to put historical transmission-grid data to use to improve grid reliability, safety, and asset management, but lacked the data science staff, expertise, and budget to jump feet first into a major predictive analytics engagement.
The utility believed data in its systems could unlock better ways of doing business, but it didn’t have a blank check to go find out– or the appetite to take on an arduous integration project.
Instead, the utility turned to TROVE, who, through an ingenious and affordable set of Proof of Concepts, delivered the utility not only millions of dollars of savings and ROI, but produced a suite of transmission-focused predictive data science Solvers™ it would continue to use and scale across its Transmission operation.
Starting Small Pays Off Big
To get started, TROVE sat down with the Transmission team to identify areas across operations where predictive data science could impact reliability, safety and asset management. Together, they then prioritized the list, factoring in cost, projected returns, and impact on the business.
Both teams wanted a quick win that would deliver tangible business results – they found one in maintenance work orders.
Maintaining transmission and substation assets is a constant and costly challenge for electric utilities, one that can leave managers wondering how there can be such high variability in labor costs for nearly identical work. TROVE saw an opportunity for savings if it could surface commonalities buried in years of structured and unstructured data in the company’s maintenance logs.
TROVE deployed machine-learning modelling on this archival data to systematically group maintenance activities that impacted labor hours, including factors such as weather, traffic, and whether the asset was in a dense urban environment or rural mountainous terrain.
The resulting outputs helped predict the optimal steps, crew size, number of hours required, and cost of future work, empowering the utility’s managers to precisely define “units of work” and aggressively negotiate new maintenance contracts accordingly, resulting in over $1 million in annual savings.
Off to a fast start – the maintenance work order proof of concept delivered results in just 90 days, required no IT integration, and turned a low-six figure investment into a seven-figure return – the Transmission team and TROVE prioritized another one, two, and, eventually, five additional proof of concepts, which TROVE methodically completed over the next 15 months. Following is an overview.
Capital-Project Component Scoping
TROVE and the Transmission team then turned their sites on the problem of scoping component parts more quickly and accurately for new designs and rebuilds in the transmission grid (towers for example). This had been a time-consuming process for the Transmission group due to a lack of cohesive standards and the absence of component data linked to completed structures currently in service.
TROVE’s Science Squad, its best-in-class data science team, developed new machine-learning software which provides engineers usage likelihood for components, factoring in geolocational and grid-specific attribute information. Using the new Solver, the utility can determine whether, for example, lightduty poles could be installed in a portion of a line, considering proximity to the substation, environmental costs, right of way, and labor costs.
Through successful prediction of what components will be needed in future jobs, the utility dramatically reduced its projectplanning times – streamlining the installation of more robust structures – while increasing budget certainty.
Avian, Lightning, and Surge Analytics
Next, the teams turned their attention to a major set of transmission-grid reliability challenges: bird contacts, lightning strikes, and power surges. Again, with TROVE, each challenge was met with creative problem-solving and the rapid development and deployment of new Solvers.
1. Birds - Bird contacts with, and nesting on, transmission infrastructure can cause power interruptions and damage network assets. Although avian guards are generally available, substantive approaches to cost effectively place, test and adapt such devices are not.
To address this problem, TROVE evaluated the relationships between power outage, avian incident, environmental (e.g., altitude and ecological data), and infrastructure data. It developed statistical models to understand and predict transmission structures and conditions at greatest risk for avian incidents, showing that 80% of incidents occurred at only 8% of transmission-line locations. Such analyticsdriven insights are enabling the utility to pinpoint the placement of avian guards (saving time and money) and improve power-reliability ratings.
2. Lightning - Lightning strikes are a major cause of power interruptions which can damage transmission infrastructure. TROVE helped the Transmission group cost-effectively mitigate and respond to lightning by developing statistical models to understand where and when lightning strikes are most intense, what strike conditions (location and amplitude) are most probable to cause a service interruption, and specific infrastructure types that are most vulnerable to lightning strikes.
Similar to the avian guard challenge outlined above, the resulting Solver helped the utility more strategically – i.e. at less cost and with more precision – place lighting arresters across its transmission grid.
3. Surges - For electric utilities in general, and for this utility in particular, maintaining and updating aging transmission-grid infrastructure is an expensive, but necessary, cost of doing business. TROVE developed a new Solver that predicted the probability of surge-arrestor failures and survival ages by combining statistical modeling with surge-arrester lifespan and failure data, historic lightning strike data, and transmission infrastructure data. These predictions enhanced the Transmission group’s infrastructure replacement strategies, reduced maintenance and upgrade costs, and improved overall power-reliability.
Incident and Safety Reporting
Frequent and remote project work is common to the utility sector and can lead to a significant number of incident and hazard reports collected over time. Understanding reoccurring incidents, including safety events, motor vehicle incidents, improper use of equipment, and incorrect system settings became a priority at this Transmission group, in hopes that such understanding would contribute to a safer workplace with fewer errors and improved operational performance.
Yet incident- and hazard-report data often takes the form of unstructured text entries that describe situational conditions and corrective actions pertinent to analytical learning. TROVE predictive software overcame these hurdles to analyze nearly ten years of incident records, follow-up investigation results and action plans, discovering patterns and trends that indicate examples of successful mitigation strategies and recurring human performance errors. The new TROVE Solver is helping the Transmission group observe key trends and root causes to foster continuous safety improvement at the company.
From Proof to Platform
By focusing on – and solving – a series of specific business problems in the Transmission group that impacted grid reliability, safety, and asset management, TROVE highlighted the power and versatility of its predictive data science software and proof of concept methodology. Now, with 50 more use cases to go, the utility has graduated to an enterprise license with TROVE, enabling the Transmission group to develop and deploy its own predictive data science models on the TROVE platform while continuing to harness the creativity and skill of the TROVE Science Squad.
How much should we pay for each substation maintenance job?
Which components should we include in a new design or rebuild?
How do we improve reliability at lower cost with targeted preventive measures, such as avian guards, lightning arresters, and surge protectors?
How can we improve safety for our maintenance teams?