I saw news of a combustion optimization and simulation project award for a 385 megawatt power station in the U.S. It reminded me of Emerson’s Jeff Williams presentation on the topic of emissions reduction equipment optimization at the last Emerson Exchange. He also presented at the ISA Power Industry Division Symposium on the topic of applied statistical analysis for performance calculations.
In power producer’s quest for cleaner, greener operations, these statistical optimization methods are showing great applicability. Statistical tools like Principal Component Analysis (PCA) help to discover which process variables have the most influence on heat rate distribution. For those like me that are unfamiliar with the term heat rate, I found this definition:
A measurement used in the energy industry to calculate how efficiently a generator uses heat energy. It is expressed as the number of BTUs of heat required to produce a kilowatt-hour of energy…
It’s often the case that mechanical problems or incorrect loop tuning cause most energy losses.
Jeff describes the process to find optimization opportunities. It starts with mining the automation system’s historical data. He shared results from 200MW coal-fired generating units that were identical in design. The SmartProcess team took 9 months of performance data from an Ovation system.
The PCA analysis on twin 225MW units provided fast identification of the greatest effect on heat rate increase (reduced efficiency.) The two major causes were wide variability of reheat steam temperature when Unit A was at low load and variability on the condenser unit of Unit B.
New correction curves (heat rate in BTU/kWh versus reheat steam temperature in degC) were established. These were created through empirical modeling of the heat rate based on historical data. The model is created using tools such as linear and nonlinear regression, neural networks, and hybrid methods. A model of heat rate is created based on the main input operating parameters of the unit. A calculation of gradients is performed to generate these new correction curves. These curves were tested for reheat temperature.
These statistical methods were also applied to oxygen concentration in flue gas. Higher O2 leads to increased flue gas temperatures, which increases the unit heat rate. Optimizing the O2 concentration improved the heat rate.
A final example was with the unit’s feedwater temperature control. The PCA analysis showed high variation of feedwater temperature across various generator loads. It caused a significant energy loss due to the lower feedwater temperature. Once identified, the team took actions to correct the control valves and loop tuning that was causing the excessive variability.
These statistical methods showed the greatest sources of energy losses without having to devote extensive engineering efforts to build accurate thermal models for comparison to the actual plant operational data. Jeff notes that this also means less ongoing maintenance is required for these applications.