An improved Predictive Optimal Controller with elastic search space for steam temperature control of large-scale supercritical power unit

Predictive optimal control (POC) combined with artificial neural networks (ANNs) modeling and advanced heuristic optimization is a powerful technique for intelligent control. But actual implementation of the POC in complex industrial processes is limited by its known drawbacks, including the oscillation resulting from random search direction, difficulty in meeting the real-time requirement, and unresolved adaptability and generalization ability of the ANN predictive model. In resolving these problems, an improved Intelligent Predictive Optimal Controller (IPOC) with elastic search space is proposed in this paper. A new simpler and high-efficiency Particle Swarm Optimization (PSO) algorithm is adopted to find the optimal solution in fewer epochs to meet the real-time control requirements. The system output error in each control step is fed back to adjust the search space dynamically to prevent control oscillation and also make it easier to find the optimal solution. An improved recurrent neural network with external delayed inputs and outputs is constructed to model the dynamic response of the highly nonlinear system. The proposed IPOC is used to superheater steam temperature control of a 600MW supercritical power unit. Extensive control simulation tests are made to verify the validity of the new control scheme in a full-scope simulator.