Research and Application of Predictive Control Based on Emran in Superheated Steam Temperature Control System

The temperature of superheated steam of thermal power plants is characterized by large inertia and time delay. Its dynamic characteristics vary with the unit load. General strategy for the temperature control doesn't satisfy the performance requirement. We propose a predictive control approach based on extended minimal resource allocation network to address this issue. In brief a neural network model based on on-line identification of superheated steam temperature is proposed to predict future plant behavior. A receding horizon optimization of the predictive control is finalized with a on-line one-dimensional golden section algorithm, yielding the optimal control actions at each sampling time point. The simulation study shows the proposed control method has excellent control performance and enhanced self-adaptability, thus fits well the superheated steam temperature system.