Umetrics Suite Blog

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

October 28, 2019

Optimizing the function of boilers, turbines and other capital equipment used to generate power requires a careful balance of fuel, heat, pressure, operator proficiency and many other variables. Managing the process on a day-to-day, or minute-to-minute basis, is like performing a skilled and well-orchestrated dance—partly based on data, but also based on operator expertise. Yet, adding more accurate information to the equation can potentially save millions of dollars, cut emissions significantly and even expand the working life of your equipment. 

Boiler hero img

MVDA lets you understand the affect that different variables in power plant operations, such as air temperature or tank pressure, have in factors such as emissions and efficiency.

So why isn’t everyone doing it?

That’s a great question. Surprisingly, in spite of the rising cost of fuel and growing body of regulations around energy efficiency and emissions limits, many operators of major power plants—ranging from municipal utilities to large industrial companies—are still not using the data they have already collected in the most efficient way possible.

By incorporating multivariate data analysis (MVDA), power and heat generating companies can uncover ways to optimize their processes—without expenditures on new capital equipment or obtaining new permits—and gain significant increases in energy output, reduction in emissions, and extend the life of their equipment.

Reducing power plant emissions

Industrial heat and power plants can be found in many places, not only in the energy and utilities sector. Large manufacturing companies, such as those in the food and beverage, chemical, pulp and paper industries throughout the world, often operate their own boilers or turbines, either because their energy requirements are large or because independence from the grid makes it worthwhile. Altogether they have a noticeable share in the world’s fossil energy consumption and pollution of the atmosphere.

For that reason, they must comply with a variety of emissions regulations around the world imposed by governments, as well as mandates to reduce fossil fuel consumption and increase efficiency.

In 2007 the European Council set targets to reduce CO2 emissions and improve energy efficiency by 20% (compared to 1990), as well as ensuring 20% of energy comes from renewable resources (the 20-20-20 goals) by 2020.

For many power operators, the data needed to reduce emissions and optimize output are already sitting in their facilities and could be used without the need for new or expensive equipment (like scrubbers) in order to make emissions reductions. Variables such as air temperature, tank pressure or amount of steam can impact the NOx and CO2 emissions (nitrous oxides and carbon dioxide). But determining how these variables interact with each other and which have the most significant impact is not easy to determine using a cause-and-effect or iterative testing method.

From univariate to multivariate data analysis

Plant managers today often look at factors suspected of causing a problem (or considered as a potential way to make improvements) one at a time. They are investigating optimization in a univariate way. They choose one point, and compare it against another point to look for a solution or improvement.

But with multivariate data analysis, it’s possible to look at millions of data points all at once to discover the impact they might have on each other. And the analysis can be done without having to shut down the equipment or lose operation time. In addition, the solutions uncovered are about optimizing the existing equipment, not focused on new capital expenditures. That means they are faster, less expensive and don’t require permitting to implement.

The data needed already exists. Power plant operators have been collecting it for years: in operational historians, continuous emission monitoring systems (CEMS), health and safety records, personnel files, financial reports on revenue and expenditures.

data sitting in emission monitoring systems (CEMS)

You have the data already. It’s sitting in your operational historians, continuous emission monitoring systems (CEMS), health and safety records, personnel files, financial reports on revenue and expenditures. Are you using it to optimize your operations?

Turning your data into meaningful insights means gathering it together in the right format. Depending on the specific type of equipment used, this might include data such as fuel consumption, temperatures at different points, steam pressure, emissions data, generator field amps, wattage and many more. By using the data to build a multivariate statistical model, you can find out which of these factors or variables are actually important in your plant operations. Additionally you will know how these data correlate and also see sources of variability or instability of operation.

Read more: View case studies

Consider a boiler facility example

Let’s consider the impact that various boiler operations parameters have on emissions using data available in a CEMS monitor. The CEMS data might include:

  • GASFLOW (HSCFH)         
  • GFLW#HR (KLB/HR)       
  • GFLWKCFH (KCFH)         
  • HEAT (MMBTU/HR)       
  • NOX#MM (LB/MMBTU)              
  • NOXR (PPM)     
  • PILOTGAS (mcfh)            
  • PRESS (PSIA)     
  • STEAM  (KLB/HR)            
  • STKPRESS (INHg)             
  • STKTEMP (DEGF)             
  • Calculated Variable | GCE - STEAM  (KLB/HR)/HEAT (MMBTU/HR) – General Cycle Efficiency
  • Calculated Variable | CO2 (PERCENT)/HEAT (MMBTU/HR)

If we look at two years of CEMS data, and match it against one emissions variable (NOx), we might find (to our surprise) that the variable most likely to have an impact on emissions is air temperature. In other words: higher air temperature reduces NOx (this is shown as a positive influence).

Here’s an example of a plot of this data from plant using four steam boilers:


The plot of data from a power plant with four steam boilers shows higher air temperature reduces NOx.

The entire analysis (using measurements every 15 minutes) contains 290,000 data points. How long would it take you to test this theory without a multivariate data analysis approach?

Some additional information we are able to see looking at the coefficients affecting NOx production: The more output, the higher the NOx; the hotter the stack is, the less NOx is produced.


The data from power plant operations shows higher gas flow and steam pressure creates more NOx emissions, while high temperature and CO2 reduces NOx.

Then, we can consider all the contributing factors—combining operational (historic) data with hourly NOx emissions to see the correlations.

From these we can learn some surprising things for this specific boiler. For example:

  1. It’s not just the feed water temperature that is related to NOx emissions. When the pressure in the tank is high, the NOx is low.
  2. Any heat will make steam, however the geometry of the pilot flame will determine NOx content of the flue gas.  A short wide pilot flame will contribute more to NOx content than a narrow elongated one. Turning the flame on and off can help save fuel and CO, while reducing NOx.

Predicting maintenance requirements and equipment lifetime

The knowledge gained from data analytics can be used for more than fuel savings or emission control. If you have suitable sensors in place, such as vibration sensors, and have appropriate historic data at hand, you can also use the modeling process to obtain additional data on your equipment, which will allow you to predict the failure of the relevant parts and schedule maintenance and replacement in time, but no earlier, provided you use online monitoring.

You will also be able to predict the remaining lifespan of your machinery: if run under constant conditions the model will be able to tell you how long your plant will last. For example, the data from the operations above showed that a turbine could be operated safely for another three years beyond the 12 years lifetime specified by the vendor. This means capital investment in a new turbine could be delayed for 3 years.

You can also use MVDA to uncover the factors that lead to wear or earlier than expected equipment failure.

Watch webinar: Predictive maintenance of machines and equipment

Training new operators more quickly

Running a plant efficiently and safely also requires experienced operators. Often, proficient operators make adjustments to equipment based on years of trial and error that gives them situational insights. But armed with empirical data, even less experienced operators have the necessary knowledge to make adjustments that will keep power operations running smoothly.

So this is where the real power of MVDA is realized. It can be used to enhance operations, manage environmental compliance, keep operators updated on a minute-by-minute basis, and give newer operators years of experience through technology.

Applications and results

Multivariate data analysis software has many useful applications in power and heat generation facilities. For example it can uncover ways to:

  • Reduce fuel costs as much as 2-6 % without hardware changes.
  • Decrease and control emissions to comply with permits or environmental regulations.
  • Avoid unplanned downtimes and reduce maintenance costs, while increasing unit lifetime 3-5 years beyond manufacturer specification.
  • Gain metric-based operational standards that can be used to train new operators

Every boiler, turbine, and furnace will have different variables that influence their production and emissions efficiency. Yet, multivariate data analysis provides great potential for pollution reduction, fuel efficiency and overall operational enhancement.

Watch webinar: Predictive maintenance of machines and equipment

Software for MVDA

SIMCA is a multivariate data analytics 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

Learn more about Umetrics Suite

Discover other applications and uses for Umetrics Suite of tools for multivariate data analytics.

Watch a recorded webinar




Topics: Multivariate Data Analysis, Data Interpretation & Analysis, SIMCA

Stefan Langner

Written by Stefan Langner

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

Leave a comment