How to Monitor and Predict Operational Performance with Digital Analytics [PODCAST]

Credit to Author: Aaron Larson| Date: Wed, 26 Jun 2019 21:33:48 +0000

Power plants are capturing operational data in ever-increasing amounts. However, analyzing all the data can be challenging. A number of tools are available that can help.

In this episode of The POWER Podcast, two experts from one technology provider explain how big data can be analyzed to identify trends and create actionable information to solve production issues. Their company’s solution allows users to troubleshoot problems, and monitor processes and assets, in real-time, so operators can make better decisions, faster.

“We want the user to have a really good springboard to be able to jump into analytics,” said Nick Petrosyan, data analytics engineer with TrendMiner. “If you’re going to be doing some type of analytics or you’re troubleshooting something in your process, it’s usually because one of your KPIs [key performance indicators] are off. So, we would like to have those KPIs somewhere very easy for them to access and see, so that they start doing more analytics in their day-to-day job, and troubleshooting and solving more problems.”

Petrosyan said there are more than 30 power plants across the U.S. connected to TrendMiner and the tool is paying dividends. As one example, he mentioned a plant that was experiencing a decrease in efficiency over time, but it had been hard for operators to identify the cause due to the wide range of variables involved.

“The first thing we did was we flattened the dataset with regards to ambient temperature, using TrendMiner’s big data searches,” Petrosyan said. “So, finding periods where ambient temperature was relatively stable and wasn’t varying a whole lot—that you’re able to do in TrendMiner in a matter of seconds—and then from there you can start to layer these really stable periods in terms of ambient conditions on top of each other and start performing comparison analytics.”

Petrosyan said different plant parameters were then added and compared to see what besides efficiency had changed over time. “The first thing we confirmed was that efficiency was actually changing, and then we were able to find something very, very subtle in their process, a malfunction in their air supply system that was causing this decline in efficiency,” he said. “Once we were able to narrow down and pinpoint the actual root cause of the efficiency decline, they were able to repair it really quickly.”

Another thing Petrosyan said TrendMiner has done for customers is “fingerprint what a really good startup looks like.” The company does that by taking multiple startups that were performed well, combining them together, and generating a fingerprint of an ideal sequence. Then, when future plant startups are performed, they can be monitored against the model conditions.

“They were able to really reduce the amount of bad startups that they had and reduce the frequency that their heat recovery steam system was experiencing thermal stresses,” Petrosyan said. In the end, engineers developed multiple fingerprints based on the length of time the plant had been offline, and operators could choose the appropriate model for any given situation.

Thomas Dhollander, CTO and co-founder of TrendMiner, said there are generally four ways in which customers experience and quantify the return on investment for digital solutions. They are in time savings, solving unsolved cases, avoiding abnormal situations, and knowledge capturing and sharing. “The fourth one is maybe harder to quantify but strategically very important,” Dhollander said.

For more power podcasts, visit The POWER Podcast archives.

Aaron Larson is POWER’s executive editor (@AaronL_Power, @POWERmagazine).

The post How to Monitor and Predict Operational Performance with Digital Analytics [PODCAST] appeared first on POWER Magazine.

http://www.powermag.com/feed/