Sliding Mode Control Of STATCOM Reactive Power Compensation Based On Neural Network

Flexible AC Transmission Systems (FACTS) is a power transmission system,combined with modern power electronics, control theory, computer control systemsand signal processing theory，which can enhance voltage stability and improve powerquality of the grid effectively. STATCOM, an important one of FACTS devices, withits fast response speed, continuous adjustment of reactive power and maintainingvoltage stability advantages in reactive power compensation of grid, has been furtherresearch by electric power research institutes at home and abroad. In this paper, thecontrol system of STATCOM is studied, and the controller is designed based on RBFneural network sliding mode control. Also, the simulation verify the effectiveness ofthe controller.Firstly, the paper analyzes the principle of STATCOM, and establishes themathematical model of STATCOM based on the method of input-output modeling andKVL. Second, STATCOM model in three-phase static coordinate system is transformedinto two-phase synchronous rotating coordinate system by park transformation, andverify the correctness of the model through the principle of conservation of energy.Then two independent subsystem are obtained by state feedback exact linearizationmethod.STATCOM is a strong coupling nonlinear system. Its controller designed bysliding mode variable structure control, can significantly improve the dynamicresponse characteristics and robustness of the control system. This paper expounds adesign method of control system by sliding mode variable structure control, in whichexponential reaching law is used for weakening the high frequency chattering. Due tothe setting of parameters in exponential reaching law is lack of strict theoreticalsupport, thus lead to the reaching law parameters selection has certain blindness. So,the online learning ability of RBF neural network is applied to the exponentialreaching law parameters adjustment, and makes the parameters time-variant, which canmaximum weaken chattering and improve control effect.Simulation shows that the sliding mode controller which the reaching law gain isadjusted by RBF neural network simplifies parameters settings and further weaken the chattering. Compared with the conventional sliding mode control, the controller, in thispaper, has a better control effect in maintaining voltage stability, suppressing thecurrent mutation and reactive power compensation.