Combustion Optimization Using AI and Knowledge Capture

The highly complex and sensitive nature of combustion makes it one of the most challenging processes to control in a thermal power plant.

Taber International's Combustion Optimization System (COS) built within the Griffin AI Toolkit® platform recognizes that all units are distinct and custom builds each application to meet the specific needs of every unique unit.


circuit icon


We customize each COS application to fit the unique traits and requirements of the system and unit we're focusing on. By blending machine learning with captured expert knowledge, we can optimize various goals simultaneously. Plus, our system can adapt to prioritize different goals based on current conditions.

We have successfully simultaneously optimized:

  • NOx Emissions (10%-40%) and CO Emmissions (40%+)
  • Heat Rate Improvement (0.7%-1.5%)
  • Temperature Control (swings reduced 25%+)
  • Tube Metal Temperatures (exceedances decreased by 80%+)
  • Loss-on-Ingnition (LOI) / Carbon-in-Ash(CIA)
  • Boiler Slagging of Tube Sections

As an adaptive, intelligent solution, Taber's COS is compatible with a wide variety of equipment and unit configurations.

  • Tangentially-fired, wall-fired, or cyclone units
  • Forced or natural circulation
  • Sub-, super-, and ultra-supercritical units
  • With or without low NOx burners
  • Works in concert with SCRs or SNCRs, if present
  • Few or many automated hardware components

NOx and CO Emissions

MATS Rule and EPA, Clean Air Act Title V regulations require that power plants take steps to reduce and optimize their emission rates of certain pollutants, including nitrogen oxides (NOx) and carbon monoxide (CO).

One of the most cost-effective methods of realizing this is through a real-time combustion optimization system.

Taber's COS has been implemented at more than 40 units for NOx and CO control and reduction and has seen tremendous success, with results ranging from 10%-40% NOx reductions and 40%+ improvement in CO emission rates.


graph icon


One long-term study demonstrated continual improvements in observed NOx emission rates quarter-to-quarter, ultimately achieving an average improvement of 22.5% across the unit's load range.

Graph: continual improvements in observed NOx emission rates quarter-to-quarter

Heat Rate Improvement

Graph: Heat rate improvements

A typical unit implementing Taberss COS for Heat Rate improvement will see, on average, an 0.7% improvement, while improvements as high as 1.5% have been realized.

The COS improves the heat rate through more efficient overall combustion and energy utilization, resulting in lower controllable losses such as lower average excess O2, reduced spray flow, and lower auxiliary power usage.


graph icon


Heat rate improvements across the load range are achievable, often with the greatest opportunity at low- and ultra-low loads as units become more flexible to meet ever-changing grid demand.

Temperature Control

Increased unit flexibility to meet challenging grid demands has placed an emphasis on improved temperature control through combustion units. Maintaining consistent temperatures below limits increases unit health and reliability, decreasing maintenance costs and improving overall performance, contributing to efficiency improvements and unit dispatch order.


graph icon


Steam temperatures across the unit's load range are more consistent, and improved temperature setpoint realization is achieved. This is espeically important at low- and ultra-low loads where heat rate often suffers.

Chart: Steam temperatures across the unit's load range are more consistent

circuit icon


Self-Learning and Self-Adapting

The artificial intelligence deployed within the combustion optimization system learns the unique relationships between all the operating parameters of the unit, recognizing best practices for optimal performance in the current unit situation. The system is also configured to learn and update itself in real-time, meaning that as the unit changes over time (e.g., equipment degradation), the system is automatically learning the effects this has on the unit and adapting its decision-making in response. Sites have even been able to avoid the installation of SCRs through the NOx emission improvement performance achieved using real-time combustion optimization.

Expert Knowledge Capture

The artificial intelligence deployed within the combustion optimization system learns the unique relationships between all the operating parameters of the unit, recognizing best practices for optimal performance in the current unit situation. The system is also configured to learn and update itself in real-time, meaning that as the unit changes over time (e.g., equipment degradation), the system is automatically learning the effects this has on the unit and adapting its decision-making in response. Sites have even been able to avoid the installation of SCRs through the NOx emission improvement performance achieved using real-time combustion optimization.

optimize the complex, automate the mundane
talk to an expert