Review of control strategies employing neural network for main steam temperature control in thermal power plant

Abstract

Main steam temperature control in thermal power plant has been a popular research subject for the past 10 years. The complexity of main steam temperature behavior which depends on multiple variables makes it one of the most challenging variables to control in thermal power plant. Furthermore, the successful control of main steam temperature ensures stable plant operation. Several studies found that excessive main steam temperature resulted overheating of boiler tubes and low main steam temperature reduce the plant heat rate and causes disturbance in other parameters. Most of the studies agrees that main steam temperature should be controlled within ±5 Deg C. Major factors that influenced the main steam temperature are load demand, main steam flow and combustion air flow. Most of the proposed solution embedded to the existing cascade PID control in order not to disturb the plant control too much. Neural network controls remains to be one of the most popular algorithm used to control main steam temperature to replace ever reliable but not so intelligent conventional PID control. Self-learning nature of neural network mean the load on the control engineer re-tuning work will be reduced. However the challenges remain for the researchers to prove that the algorithm can be practically implemented in industrial boiler control.