OCR Text |
Show In the electric power industry, as in many other combustion-dependent industries, these characteristics are accentuated. The methods used to improve these process~s must be especially non-intrusive. A common approach for some has bee~ para~etnc testing, which is usually employed by changing one control variable at a time while holding the remainder of variables constant. This approach not only excludes process influences that cannot be controlled but also ignores the interactions between control variables. In order to meet the special requirements of a production system, an optimization solution needs to exhibit the following advantages: • Can be tailored to fit a wide variety of different processes. • Allows for specification of multiple process objectives, including economic value. • Creates accurate predictive models from data taken from the process (uses empirical methods). • Is non-intrusive in the method of collecting process data. • Must be effective in discovering new, optimal operating conditions away from standard conditions. • Must be available for continuous use in order to respond to changing operating conditions and evolving business objectives. • Must be easy to use so an operator can employ the technology as a support system. Two Distinct Functions: Modeling and Optimization There are two stages to applying an effective optimization technology. One stage is modeling, and the other is optimization of the model (Figure 1). These are two separate functions, a distinction that is not always recognized. Optimization is often attempted based on the model created with historical data taken under standard operating procedures. While the models created from such data may be quite accurate and predict well for the range of data modeled, they may not provide the extrapolation capability necessary to discover optimal settings that are outside the range of the historical data. This is often overlooked when using neural network modeling methods. The models may be very accurate if the optimal region has been established, but additional search technology is needed to discover an optimum that is distant from the region of historical data. In this paper we will examine the performance of the UL TRAMA~ Method of Sequential Process Optimization (SPO) as it is applied in three electric power generation examples. A few other combustion industry processes will also be mentioned for consideration. Electric Power Industry Challenges A growing number of utilities are including boiler tuning in strategies to comply with the Clean Air Act Amendments of 1990 (CAAA), as well as to prepare for a more competitive business environment. Combustion tuning (optimization) is a logical first step in meeting both these challenges. Power generation plants must learn to operate at the least cost as well as to comply with NOx emission regulations. This UL TRAMAX Method of SPO has been applied to scores of processes in recent years'. In particular, special focus has been on the combustion process of fossil-fuel power generation boilers to reduce NOx emissions, improve fuel usage and reduce the carbon content of fly ash2 • SPO, provided to the electric power industry by Ultramax Corporation and the Electric Power Research Institute (EPRI), is being utilized as a low- 2 |