Internal model controller for CFBC boiler using neural networks

Abstract

Superheater temperature control for a circulating fluidized bed combustion (CFBC) is really challenging during plant operation as lots of factors affect the temperature such as main steam flow, drum pressure, spray water flow, flue gas flow and temperature. This paper, deals with a single input single output (SISO) system, where spray water flow is the manipulated variable & superheater temperature is the process variable. To operate the plant at higher efficiency, a precise control of superheater temperature (540°C± 5) control is required. Superheater temperature control through a simple PID controller is not satisfactory due to large process lag and process non-linearities and presence of external disturbances. An Internal Model Control (IMC) using artificial neural network (ANN) is procedurally easy to design and accounts for process non-linearities. ANN is obtained by training available plant input-output data of 125MWe CFBC boiler at NLC, Barsinghsar, India. Parameters for ANN are optimized by Particle swarm optimization (PSO) algorithm. Results obtained by PID, Linear IMC and non-linear IMC structure are compared.