Umetrics Suite Blog

MSU Uses Data Analytics to Uncover Cost Savings and Reduce Power Plant Emissions

June 2, 2020

Looking for ways to improve the efficiency of its power plant operations while reducing costs and environmental emissions, the Department of Power and Water at Michigan State University (MSU) began a study using multivariate data analytics that led to some surprising findings. The results have implications that could help other operators of large-scale power facilities reduce their carbon footprint and improve power plant operations.

boiler optimization msu case study umetrics

A collaboration between the University of Michigan and Sartorius data scientists uncovered surprising ways to reduce fuel costs, improve power plant efficiency, reduce environmental emissions and extend the life of equipment.

A public research university located in East Lansing, Michigan, MSU operates its own large-scale power supply with multiple industrial steam boilers and turbines. The university–one of the largest in the United States—spends over $20 million USD a year on fuel (natural gas). MSU operates four steam boilers and one recovery steam generator, and power is generated with five steam turbines and one combustion turbine.

“If we can save just 1% of fuel costs from going up the chimney, that is a huge impact,” said Nathan Verhanovitz, Performance Engineer with the Department of Power and Water at MSU. “1% is literally hundreds of thousands of dollars for us.”

Working with data scientists at Sartorius Stedim Data Analytics and making use of the vast amounts of data that the facilities department collects in conjunction with their equipment and power generation, MSU was able to develop a new approach to large-scale power and equipment management using multivariate data analysis (MVDA).

“We have a lot of data coming from many different systems. Our stack emissions computer is recorded separately from the burner monitor and other systems," said Verhanovitz. “Using MVDA, we were able to create virtual models of our power plant systems that identify and project which variables have the most impact.”

Using historical data, MSU was able to see the impact, for example, of how room temperature of a monitoring system affects emissions or how turning the combustion turbine off and back on would increase fuel efficiency without having to run months of real-world experiments.

Beyond saving money and reducing MSU’s carbon footprint, the MVDA project also supports the university’s three pillars: teaching, research and outreach. “This not only lets us provide reliable energy for the university at the lowest cost possible, but also lets us to do great things toward environmental stewardship,” added Verhanovitz. “By sharing what we learned, we’re fulfilling the outreach part of the university mission.”

Read more: From emissions to fuel efficiency: what is your power plant data hiding?

Using Existing Data to Gather Insights

For MSU, the first step in optimizing power production involved gaining a detailed picture of all parameters that affect the operation and performance of the existing plant units and their interplay. This data included variables such as fuel consumption, temperatures at different points, steam pressure, emission data, generator field amps, wattage, time of day, daylight hours, ambient temperature, and many more. It came from various process monitors, performance databases, as well as the continuous emission monitoring system (CEMS) data.

Using multivariate data analysis, MSU created a model based on historical data, and was able to use this to discover and predict the factors that have the greatest impact on emission rates, fuel usage, operating efficiency and expected length of service for equipment. Some of the results were surprising.

“In the old days, to determine the impact of changes in operating settings or input variables, you would have to do a controlled test,” said Verhanovitz. “But using the MVDA models we have developed, we were able to pinpoint relationships that might have taken days or weeks, as well as needing decades of expertise, to uncover. Using the MVDA models, we can get answers in just a few hours.”

Optimizing the power production process and emissions control using software is far less costly than purchasing new equipment and doesn’t require permitting the way purchasing and installing new equipment, such as scrubbers, does.

How much could you save using MVDA?

Surprising Findings from MVDA

One of the great strengths of multivariate data analysis and modeling is that it shows interrelations between factors that the human operator or engineer, and in some cases even the equipment manufacturer, may not have known about. The impactful findings MSU discovered from multivariate data analytics included:

Emissions Reduction

■ Pressure affects NOx. It’s not just the boiler feed water temperature that is related to NOx emissions, pressure is as well. When the pressure is high N0x is low. This is a case where a fairly inexpensive hardware addition makes sense: if the feed water is being supplied from a pressurized feed water tank, it will reduce NOx emissions.

■ Air temp affects CEMS sensors. It was discovered that the HVAC / air conditioning unit on the continuous emission monitoring system (CEMS) shelter caused artifacts in the NOx measurement and affected control. The plant runs with a 10 ppm emission certificate. It was possible to achieve 8-9 ppm on a 24-hour rolling average. Stabilization of the control circuit, solving the air conditioner problem and changing the plant operation mode accordingly has led to consistently 7 ppm NOx emitted.

■ Warmer gas reduces NOx. Turning the fuel chiller off reduces NOx. Although gas and therefore energy density will drop, the warmer gas burns faster and more efficiently, reducing NOx production. Whereas, a colder natural gas will require more time to burn completely.

Fuel Reduction and Efficiency

■ Factors affecting steam uncovered. Using coefficient plots, the team identified the factors that positively and negatively affect steam production. Eliminating negative factors means an annual fuel savings of $500,000 USD.

■ Temperature related factors identified. Analyzing the performance of a feed water heater and removing inconsistencies in temperature control could save an additional $250,000 USD.

■ Shut off increases efficiency. Shutting down the gas turbine over the weekend has led to a fuel use improvement of 2.3% —equivalent to a washing cycle— saving around $70,0oo USD in fuel (this was a new learning that even the manufacturer of the turbine didn’t know about). An additional efficiency improvement will be attained by running the turbine at the highest allowable gear box vibration rate.¹  Also, as the turbine rotor has a specified run time (in this case of 30,000 hours), the weekend shutdown should extend its life by at least 20%.

Predictive Maintenance

A positive side effect of the feed water heater analysis was that a feed water pump that was nearing failure was discovered. Its failure would have caused a major upset soon by taking the plant offline. This would have resulted in a cost of more than $250,000 USD in sourcing power externally and replacement of the feed water heater. Instead, the cost is limited to that of a new pump.

Extending Equipment Lifetime

Monitoring the operation of equipment using the experience gained from the historical data (determining optimal settings, safe operating ranges, etc.) it will be possible to extend the lifetime of equipment beyond the manufacturer specifications. In many cases 3 to 5 years are possible, delaying necessary fixed capital investment in new equipment accordingly.

Boiler secondary img

Read more:  Meeting common challenges in balancing emissions and boiler efficiency

More Effective Training for Operators

Running a plant efficiently and safely also requires experienced operators. Creating unambiguous operating models based on MVDA helps experienced operators work more efficiently, as well as providing actionable knowledge to better train the incoming generation of operators. New operators learn best practices more quickly with graphic visualizations of historic data that shows specific cause and effect relationships between variables.

Verhanovitz said that phase two of the MVDA project at MSU will be to implement real-time data monitoring. “We’re bringing process optimization right into the control room,” he said. “So operations will be synonymous with meeting efficiency goals. This is a new way of thinking…it’s where innovation lives. The pharma companies have been doing it for decades, but in the power industry, it’s innovative.”

Summary and Conclusions

Multivariate data analysis has many useful applications in pollution reduction, fuel efficiency and overall operational enhancement. Every boiler, turbine and furnace will have different variables that influence their production and pollution efficiency. Mapping the interplay of all the equipment and variables can only be done using powerful MVDA methods and tools.

The insights from the data analysis models created savings worth over $1 million USD per year. The return on investment was realized in less than one month. The benefits achieved were:

■  Fuel cost savings in the range of 2–6 % annually
■  Reduced emissions in compliance with permits /environmental regulations
■  Predictive maintenance for equipment that saves time and money
■  Increased equipment lifetime (3 to 5 years for several units)

In addition, using software to optimize the processes used in power generation and emissions control is far less costly than purchasing new equipment, such as scrubbers. Running software doesn’t require permitting the way purchasing and installing new equipment does. Thus, a moderate investment in MVDA software (online and offline) can save large amounts on fuel and maintenance.

Even for operations that don't have access to big data and historical data sets across multiple systems, like MSU did, valuable knowledge can still be gained from CEMS data. Multivariate analysis of CEMS data could increase cycle efficiency and help identify opportunities for emissions rate reductions.

Software for MVDA

SIMCA is a multivariate data analysis software that transforms your data into visual information for easy interpretation. This enables you to make decisions and take action quickly and with confidence. Find out more.

See a Boiler Operations Example

Michigan State University , which operates a its own power production facility running multiple industrial steam boilers and turbines, used SIMCA to uncover key optimization factors that led to major cost savings and emissions reductions.

Get Case Story


How much could you save by optimizing your power plant? 

Use our free ROI calculator to find out.

Try the Boiler Benchmark 



1. Until now this vibration has proven harmless. Nevertheless, should it shorten the lifetime of the gear box, replacement will be cheaper than to waive the savings that can be attained through running the turbine in this efficient operating mode.

Topics: Multivariate Data Analysis, Data Analytics, Power plant

Stefan Langner

Written by Stefan Langner

Stefan Langner is Market Manager for Process Industries (Food & Beverage and Chemical) markets at Sartorius Stedim Data Analytics.

Search the Blog

    Subscribe to the Blog

    View the:

    Data Analytics Glossary of Terms

    List of Webinars

    Get a free trial