Data: Automated

The following is an excerpt from our Enriching Data, Empowering Action whitepaper. You can access the full report here.

Download Whitepaper

Having access to the right data can mean the difference between uptime and downtime, customer cost savings and rate hikes, growing margins, and crippling expenses—even life and death. Knowing about damaged components in a timely manner, aging equipment, or vegetation encroachment gives you an opportunity to mitigate the issues before they turn into big, expensive problems.

Right Data, Right Time

Automating data is the process of generating the desired outputs from a set of provided inputs with little to no human intervention. This requires finding the optimal mix of human and machines that requires the least amount of human interaction while still meeting the quality standards of the utility.

In the PrecisionHawk solution, artificial intelligence combines with human skills to create a streamlined end-to-end data workflow that increases accuracy and efficiency without incurring unreasonable costs.

Veg encroachment PA

Humans are involved at the beginning and end of this workflow. At the beginning, data engineers research, design, and implement the machine learning models. At the end, data analysts analyze the outputs of the model for quality control before sending them along to the client. Artificial intelligence and machine learning handle the middle phases to create a manageable dataset that includes only the most relevant information.

Building, Testing, and Improving the Models 

PrecisionHawk uses three metrics to score the performance of its machine learning models:

  1. Recall is a measure of how well the model performed in finding all of the true targets in the data. This metric takes into account the number of true positives and false negatives; in other words, did the model find everything it was supposed to find? Recall is important when missing something comes with a high cost; for example, when looking for anomalies that might cause damage or downtime.

  2. Precision takes into account the number of true positives and false positives, but it also takes the measure of all of the targets that the model detected. How many of them were correct? Precision is important when a high level of insight is needed, for example when identifying types of equipment.

  3. The F1 Score is a combination of Precision and Recall, which is useful for comparing a model with other models. F1 is very important when you're trying to find the right balance between precision and recall.

While all three of these metrics are important for different needs, Recall is the key metric for PrecisionHawk’s utilities clients since utilities are looking to identify anomalies that could lead to equipment failure, network outages, fires, or permanent damage to the network.

How AI Powers Thermal Tagging to Prevent Overheating

During the summer of 2020, in the middle of the COVID-19 pandemic, electric utilities faced scorching weather as well as increased and unpredictable demand—which heightened the risk of overheating components.

One solution: PrecisionHawk’s inspection process that incorporates AI to help utilities identify overheating parts before they fail.

  1. Once thermal images of electrical equipment are collected, a model in PrecisionAnalytics reviews the images and detects items of concern, from connection points to insulators.

  2. The ML algorithm produces anomaly reports, selecting only those images that show potential damage.

  3. Thermographers review the subset of images for thermal issues. 

  4. The utility company quickly identifies thermal issues and ranks them by severity so they can be pushed to work orders and repaired.

Since the machine learning algorithm culls through the data to pull out only the most relevant images, it enables the utility to collect massive amounts of data—increasing accuracy and decreasing the risk of thermal issues.

Data that Drives Results

Accelerating, automating, and amplifying data helps utilities collect the vast amounts of data needed to ensure accuracy and usefulness, because AI and machine learning can filter that data down to the most relevant information. The balance of automation and human intervention makes the process cost effective, as well.

With the right data, utility personnel can now make the best decisions to increase safety, customer satisfaction, and cost savings—decisions that will benefit the utility today and tomorrow.

The bottom line: by deploying PrecisionHawk’s data solution, enterprises can improve every step of their inspection process—from collecting data to deploying repair crews.

Download the full Enriching Data, Empowering Action whitepaper today to learn how data-powered decision-making can help you reduce downtime, increase safety, and lower costs.

Download Whitepaper