Stability Control For Power Systems Using Deterministic Learning

The fast identification and control of faults is very important for the operation of power system.The failure of timely control may lead to large-scale power failure,or even power system paralysis.With the increasing scale of power system,the types and uncertainties of faults become more and more complex.The rapid identification and control of power system has become the focus of power system research.If the fault can be expected,the fault can be identified quickly and controlled.Deterministic learning theory is a fast and effective method to approach the periodic trajectory or quasi periodic trajectory in unknown system.Based on the deterministic learning theory,two control methods are designed for the power system model.The main work of this paper is as follows:1.Power system stability control based on deterministic learning and experiential learning.(1)Using radial basis function neural network to train the unknown correlation term of power system fault,accumulate the knowledge of power system fault dynamics,and store the learned knowledge in the knowledge base,so the fault can be identified quickly.(2)for the control of expected fault,this paper designs a new control strategy by introducing derivatives of affine terms based on the study of Lyapunov theorem,deterministic learning theory and back-step control method.This strategy can avoid the problem of controller singularity without the requirement of integral-type Lyapunov functions.(3)Based on the unexpected fault,the traditional adaptive control method is adopted for control.Finally,the simulation results show that the control strategy designed in this paper has good control performance when the power system suddenly fails.2.ISS modular adaptive neural network control of power system.In the control method based on deterministic learning and experiential learning,the derivative of affine term is introduced,which increases the difficulty of training.In this paper,ISS(input-to-state stability)modularization and adaptive neural network are combined,and a simpler scheme is designed to avoid the possible control singularity problem without the use of any restriction on the derivative of affine terms.(1)The mathematical model of power system is simplified into two subsystems,one is the state error subsystem,the other is the neural network weight estimation error subsystem.(2)The back-step control method is used to control the simplified subsystem.The interconnected term is introduced into the nonlinear function to compensate the interconnected term and overcome the restriction on the derivative of the affine terms.(3)The persistent excitation condition of localized radial basis function networks is still satisfied for subsystems.In the last two steps,the deterministic learning theory is used to approach the unknown interconnected terms.The learned knowledge of system dynamics is stored in a constant radial basis network for reuse.Finally,the stability of the power system is verified,and the effectiveness of the method is verified by simulation experiments.