T-S fuzzy identification for main steam temperature system using improved particle swarm optimization

For the main steam temperature system of pulverized coal fired boiler, the modeling precision is not quite satisfactory based on the traditional transfer function. Utilizing the nonlinearity of thermal process, the paper proposes the methodology of T-S fuzzy neural network for data fitting. The antecedent parameters are determined by selected centers obtained from simplified subtractive clustering method, and the number of `If-Then´ rules is automatically generated. Afterwards, the improved particle swarm optimization algorithm is proposed to assign the initial consequent parameters of rules which are then fine-tuned by BP algorithm. The simulation results show that the algorithm not only achieves the goal of higher precision, but also exhibits higher generalization ability with respect to the problem of identification and optimization of the main steam temperature system.