STATISTICA Blogs

Thomas Hill, Ph.D., VP Analytic Solutions for StatSoft Power Solutions, and Douglas Shuit, M.S., P.E., Process Performance Engineer at Alliant Energy's Edgewater Generating Station, will present a paper entitled "Robust Optimization Strategies Based on Historical Data to Achieve Improved Reliability and Lower Emissions" at the upcoming 2009 Electric Power Conference in Chicago (the presentation scheduled for 5/14/2009 to begin at 8:00 AM and end at 9:30 AM in room 29). The presentation will focus on the implications of historical data based boiler optimization for improved equipment reliability, availability of capacity, and reduced frequency of unscheduled outages.

Robust Optimization Strategies Based on Historical Data to Achieve Improved Reliability and Lower Emissions

Douglas Shuit, M.S. P.E.
Process Performance Engineer
Alliant Energy Inc.
Edgewater Generating Station, Sheboygan

Thomas Hill, Ph.D 
VP Analytic Solutions
StatSoft Power Solutions
Tulsa, OK  

September 24, 2008
Abstract
Data driven optimization strategies, based on historical performance data, (e.g., data mining) have been widely adopted in various industries (e.g. chemical manufacturing, mining, refining, pharmaceutical manufacturing, to name a few), where these methods have become indispensable tools for cost-effective performance and reliability optimization with existing equipment, instrumentation, and machines. This presentation will illustrate the approach and specific methodologies in various case studies involving cyclone and wall-fired furnaces to (1) achieve stable and sustainable Furnace Exit Gas Temperatures (FEGT), tube metal temperatures while simultaneously lowering emissions (NOx, CO), and mitigating the fouling effects of sticky fly ash in the back pass and (2) to improve overall the consistency (robustness, availability) of operations over a wide range of mill coal flows, unit loads, fuel qualities, etc., thus leading to less downtime (greater reliability of generation capacity), lower operational costs, and improved safety, while ensuring that the boiler is not operated in zones that have never been tried before. Even if the control system is of an older design (e.g. without split dampers), numerical suggestions can be obtained that will yield improved control strategies, or different control strategies altogether. These methods, which can be applied to virtually any boiler technology and plants with various degrees of instrumentation, and typically without requiring modifications to the control system, will also be contrasted with traditional analytic approaches such as DOE (design of experiments), CFD (computational fluid dynamics), as well as neural-networks and neural-networks based closed loop control and "optimization": None of those methods specifically pursue robust and reliable performance as a key process goal, nor can address complex systems with large numbers of parameters; they are also often costly to implement requiring new or significant augmentations to existing instrumentation and control systems.