Intelligent control applications in the chemical and power industries

Problems found in the electric power and petrochemical industries can be very challenging for control engineers. This paper focuses on two distinct types of problems; namely, sensor redundancy and set-point control. For the latter problem the workhorse of the industry is the 3-element proportional-integral-derivative (PID) controller which can be tuned to provide reasonable performance over a relatively wide range of operation and control problems. PID controllers can, however, become detuned over time as operators continually make minor adjustments. Solutions to these problems of sensor redundancy and self-tuning controllers are discussed using artificial neural networks (ANNs), which can learn in a static mode for the case of sensor redundancy, or dynamically (on-line) in the case of self-tuning adaptation