NeuCo’s CombustionOpt® system is broadly known for its use of adaptive Neural Networks for both modeling and optimizing the combustion process to improve boiler performance.  More recently you’ve heard me talk about the tremendous value customers have obtained from CombustionOpt implementations that combine Model Predictive Control (MPC) and Neural Network technologies. Today I want to focus one of the ProcessLink Platform’s lesser-known technologies -- one that is becoming increasingly important to CombustionOpt installations: Condition-Based Expert Rules.

Expert System Evolution
Condition-Based Expert Rules have always promised benefit for combustion optimization software systems, but earlier solutions -- including early versions of CombustionOpt -- were typically limited with respect to their ability to fully exploit the "Expert Systems" branch of artificial intelligence.
When designing our sootblowing optimization and equipment health detection & diagnostics systems, NeuCo realized that scripting and implementing expert rules in a flexible but sophisticated manner was required to produce viable solutions. This capability was embedded in the ProcessLink Platform and has become one of the primary optimization methods used for both SootOpt® and MaintenanceOpt®. 

SootOpt, for instance, uses a "Propose-Apply" method to evaluate how applicable and optimal the many combinations of sootblowing rules are under different operating conditions, based on unit measurements (e.g. steam temperatures, attemperation spray flows, economizer exit gas temperatures, etc.).

CombustionOpt Benefits From Expert Rules
More recently, CombustionOpt installations have adopted and substantially benefitted from Condition-Based Expert Rules - and we have just begun to scratch the surface. 
In some cases, simple rules can produce dramatic performance improvements. One customer, for example, noticed that stratification across a split windbox often arose in a manner that was not well understood and not addressed through the control logic in the DCS. 

A simple rule that opened one windbox damper and closed the other when the differential pressure in one versus the other exceeded a threshold, substantially reduced such incidences of stratification. It also resulted in reduced NOx, improved heat rate, and better control of CO. While this same logic could be implemented through the DCS, adding such rules through ProcessLink is simple and can be easily modified as operating experience and conditions require.

A more elaborate example is the scripting of a rule to automatically swap in and out the most recently- trained and validated neural network models for a particular set of mills in service. This is particularly valuable for boilers where the mills in service change infrequently and/or where combustion behavior is highly sensitive to mill configuration.

Another exciting application of the ProcessLink Rules Engine for CombustionOpt involves dynamically swapping entire Optimization Profiles based on applicable goals and/or operating conditions. An Optimization Profile essentially comprises the “marching orders” for the optimizer: the entire set of objectives, constraints, relative priorities, step sizes, etc. 

For instance, one CombustionOpt customer experienced significant variation in the coal quality attributes affecting ash fusion temperature and the propensity for slag formation. Using available on-line signals that indicate whether the current coal quality is going to present a risk of slagging, NeuCo configured two alternative Optimization Profiles.

In situations where slagging is a risk, the Optimization Profile focuses on keeping furnace exit gas temperature below, and boiler O2 above, conservative maximum and minimum respective thresholds.  In situations where slagging problems are unlikely, the Optimization Profile focuses on maintaining a lower NOx limit and more tightly controlling steam temperatures.   

Other examples of on-line measurements that can be exploited by Condition-Based Expert Rules include laser-based in-furnace combustion measurements; coal flow as indicated by velocity, mass, and/or level; ultrasonic based indications of coal fineness; and on-line carbon-in-ash or loss-on-ignition as indicated by a variety of measurement methods.
Just Scratching the Surface
But as I said, we have just begun to scratch the surface here. I firmly believe that the ability to wed human expertise, inductive modeling and optimization methods into a single integrated application is going to profoundly change the way that fossil-fired generating units are operated, even those without anything more than traditional instrumentation.  The resulting improvements in efficiency, emissions, and availability are going to be made even larger with the emerging advanced measurement technologies described above. 

The next few years are going to provide for exciting times as we work together to take advantage of these paradigm-shifting advances in technology.