Research On Generalized Nonlinear Predictive Control And Online Optimization Of Complicated Thermal System

Thermal power automation is an important technique to assure the safety of equipments, increase the economy of units, and reduce labor intensity and improve the working conditions. It is extremely difficult in modelling and control upon nonlinear thermal power system, which is a very complicated control system with delay, inertia, time-varying and multi-variable characteristics.Generalized predictive control, as a real-time control algorithm with the features of low model accuracy requirements, robust, capable of overcoming the system time delay and easy to implement, has been successfully realized in industrial process control, and shows great vitality. Researching the generalized predictive control and its online optimization algorithm for the complicated nonlinear thermal power system has certain theoretical significance and value to build on the smart grid afterward. Therefore, fast and stable online optimization algorithm and its application in nonlinear multi-step predictive optimal control are focused on, and the following researches have carried out:1. Online optimization algorithmSeveral new intelligent optimization algorithms are researched:chaotic optimization, particle swarm optimization, and novel small world optimization algorithm. A novel small world optimization algorithm based on real-coding is proposed on the basis of binary-coding and decimal-coding small world algorithm which may cause the problem of long running time and cumbersome because of the complicated encoding and decoding calculations.Simulation results show that small world optimization algorithm based on real-coding had better efficiency and accuracy than chaotic optimization algorithm and particle swarm optimization, so it is suitable for online rolling optimization and solution generalized predictive control.2. The application of generalized predictive control in complicated nonlinear thermal power systemFirst of all, a novel generalized nonlinear predictive controller combined with BP neural network identification, small world optimization algorithm based on real-coding and multi-objective optimization predictive control is proposed in order to solve the control problem of multi-objective optimization of nonlinear unit power plant load system. Simulation results show that the variables of power unit have been obtained real-time control satisfactory as AGC load command changes.And then, for large inertia, delay and nonlinear characteristics of boiler overheated steam temperature system, nonlinear predictive control based on chaos genetic algorithm T-S fuzzy model and small world optimization algorithm is implemented. It indicates that the proposed generalized nonlinear predictive control with novel online rolling optimization algorithm has superiority in dealing with large delay and nonlinear system through simulation results.Finally, two control systems mentioned above are added with the constraint of the implementing agencies, taking into account that actual industrial process systems are always restricted by soft and hard constraints. Simulation results show that two kinds of generalized predictive control of nonlinear systems achieve good control effect in actual working conditions.