Enerjisa Üretim
            

Enerjisa Üretim was established in 1996 in Istanbul to champion a sustainable balance in power generation. The company has a diversified and efficient portfolio and is now the market leader in Turkey’s private sector electricity generation. With 850 employees the company has 21 power plants, a 3.607MW installed capacity, and 56% of its power generation comes from renewable energy.

Accelerating with self-service data analytics

Turkey is certainly not the only country in the world making an energy transition. And Enerjisa Üretim isn’t the only company looking to meet ambitious sustainability goals. But as a privately-run, agile corporation, it might be one of only a few using self-service data analytics to accelerate meeting its goals. Enerjisa Üretim has 3 wind power plants, 12 hydroelectric power plants, 2 solar power plants, 3 natural gas power plants and 1 lignite power plant. The ongoing devaluation of the lira makes its provision of affordable clean energy an essential part of its business plan. And like any other power plant, better asset management is key to operational efficiency and cost savings.

PI Vision gets TrendMiner boost

In its hydroelectric power plant, Enerjisa has huge turbines responsible for producing electricity. While in its natural gas plant, coolers are vital to condense the steam. Data around both these processes are critical to ensuring that the asset operates efficiently. Enerjisa knows this, which is why it is using a web-client visualization tool (PI) to show time-series data in dashboards. This used to help with simple diagnostics. But plant managers were starting to get frustrated: there were several irregularities in equipment performance or operating anomalies causing latency or downtime without anyone really knowing why. But management didn’t think they could make their data work any better for them.

“Do we really need this?” Emin Sahin from Enerjisa Üretim remembers saying at his first meeting with Software AG’s TrendMiner analytics team. “We’ve got data available to us in PI, you reckon we can interrogate that data better? And by ourselves without data scientists?” Turns out we could.

4 compelling use cases in under a week

TrendMiner data analytics engineers have a workflow that they follow with customers to results fast. In phases 2-discovery and 3-diagnostics, hypotheses can be tested, actions analyzed, and root causes determined. Given a 6-month pilot at Enerjisa, the results came in under a week. “Using TrendMinder, we were quickly able to compare a relationship between our chosen tag of interest, and potential influence factors” says Emin. “Based on time-series patterns, we could fingerprint optimal process performance. And monitor it in real-time to continuously check for deviations—setting alerts. In phase 4-predictive, we started using past performance to calculate future behavior incredibly accurately.”

That’s because TrendMiner was able to conduct powerful analyses of combined data sources from remote monitoring, internal systems, and time-series data from the past 3 years. In just one week, Enerjisa operators (not data scientists) had established 4 groundbreaking use cases for optimizing operations:

• Use Case 1: Correlation analysis of cooler. Goal: To analyze the influence of meteorological parameters on cooler performance. Results: Using scatter plots, operators found a strong negative correlation (-80%) between the ambient temperature and the cooler’s power output. They identified 12 cases over the past year where the cooler had not even kicked in despite extreme weather conditions.
• Use Case 2: Correlation analysis of steam turbine. Goal: As above but with steam turbines. Results: Contrary to previous belief operators established that atmospheric pressure did not influence the process as much as humidity does.
• Use Case 3: Optimize pump operations: Goal: To understand operational impact of pump failure. Results: The team compared periods where all pumps were operating to periods when one wasn’t. Now, they have alerts to ensure that pump downtime is avoided.
• Use Case 4: Axial shaft position. Goal: To analyze the evolution of the axial shaft position (slowly displaced over time) and predict trips based on this displacement. Results: Established a predictive maintenance schedule.

“Our new tool is now used by site engineers on a daily basis. It’s user friendly, easy to integrate with other data sources, and intuitive. It can handle noisy (fluctuating) data at high volumes and isn’t a burden on our IT resources as we have a price agreed on volume of data/users,” said Emin Sahin.

Empowering operators with Machine Learning

Currently, self-service analytics from TrendMiner are enabling three Enerjisa Üretim plants to achieve operational efficiency through remote monitoring, and predictive maintenance. Interpreting this intelligence are plant operators, who can play around with their dashboards and data just like any data genius. On the horizon for Enerjisa Üretim is to use its new tool to establish measures that tie into its sustainability strategy and ESG compliance.

Engineers will soon start using TrendMiner’s Notebooks functionality. This new intelligence layer sees all that complex operational data fed and analyzed by Machine Learning. Using Python, it’s possible for developers to program and to create new smarter dashboards for operators to interact with. This has only just started, but at Enerjisa the future is as bright as its now smart operators.