Improvement Of Control Algorithm Based On Extended State Observer

Extended state observer(ESO) is the core component of active disturbance rejection control. The pivotal idea of ESO is that the uncertainty of the system and the external disturbance are considered as a total disturbance, and it can be taken as a new extended state. Then the new extended state can be estimated in real time and compensated for feedback. So the effects of total disturbance to the system can be eliminated. Finally, the good control performance can be achieved. Because the necessary information of ESO is the input and output of the controlled object, the dependence of the mathematical model can be reduced obviously.Compared with other previous predictive control method, generalized predictive control(GPC) has more remarkable robustness. But the uncertainties of predictive model and high computational complexity are the intrinsic drawbacks of GPC.In this thesis, ESO is combined with proportioned controller and GPC, and several control strategies are proposed. The main work of this thesis can be summarized as follows:(1) Up to now, for the strong nonlinear coupling of the forced-circulation evaporation system, few valuable control methods exist. In order to cope with this problem, a kind of linear time-invariant active disturbance rejection decoupling control is proposed. Meanwhile, a fitness function is designed to evaluate the degree of decoupling of the closed-loop system. And particle swarm optimization algorithm is employed to tune control parameters. The controller can not only decouple the loops effectively but also ensure performance robustness under external disturbances.(2) For typical nonlinear systems-chaotic(hyperchaotic) systems, a kind of fast linear generalized predictive control algorithm based on ESO is proposed. The calculation complexity and the dependence on predictive model can be reduced effectively.(3) In order to further reduce the calculation complexity and improve robustness of the above control strategy, a sort of improvement method is presented. In this method, the integrator of CARIMA model is removed to enhance the stability of the control system. Moreover, the approach provides a unifying framework for GPC to control linear and nonlinear systems without distinction.The control problem of nonlinear systems can be solved effectively by employing three linear time-invariant methods above. The dependence on the accurate mathematical model of controlled plants is relatively low, and the performance robustness can be ensured.