Application And Research On Neural Network PID Control Optimized By Chaotic Particle Swarm In Voltage-source PWM Rectifier

The instantaneous power is controlled in the voltage-source pulse width modulation (PWM) rectifier direct power control (DPC) strategy. Both the power grid voltage and current information are included in the instantaneous power, whose value can’t be influenced by coordinate transformation and is constant in steady state. The characteristics of DC control are reflected in DPC system. However, the traditional DPC have the disadvantages of variable switch frequency, larger system fluctuation and DC voltage static error existed. The stator flux concept is introduced from induction motor to PWM rectifier, combined with the space vector pulse width modulation (SVPWM) technology and the instantaneous power can be gotten using the method of virtual flux estimating. The power grid voltage sensor is removed, the DC voltage static error is eliminated and the switch frequency is fixed.The neural network PID control has advantage of controller parameters self-adjusted, but the performance of the controller seriously reliance on the initial network weight coefficient. The basic particle swarm optimization (PSO) algorithm and chaos theory are combined in chaotic particle swarm optimization (CPSO) algorithm which global optimization ability is strong and convergence speed is fast. The performance of controller can be effectively improved when using the CPSO algorithm into initial weight coefficient optimization of neural network PID controller.This paper focus on DPC strategy and target to improve the performance of rectifier. The research for application of neural network PID control based on CPSO in voltage-source PWM rectifier is carried on. Following is the major works:Firstly, the mathematical model of three phase voltage-source PWM rectifier is established based on instantaneous power theory and the principle of DPC is stated. A kind of improved switch vector table is introduced for the problem of big fluctuation of instantaneous power, DC voltage and AC current caused by too much zero vectors is used in traditional switch vector table and the performance index of rectifier is improved.Secondly, the stator flux concept is introduced from induction motor to PWM rectifier. The virtual flux is used to orientation control and instantaneous power estimating. A kind of DPC strategy without power grid voltage sensor is gotten and the cost of system construction is reduced. Combining SVPWM technology, a kind of fixed frequency voltage-source PWM rectifier control scheme which have the advantages of DPC, virtual flux estimating and SVPWM is put forward in order to solve the problems that variable switch frequency and DC voltage static error existed in DPC.Then, the neural network PID controller based on CPSO is designed. The initial weight coefficient of back propagation (BP) network is optimized through offline way by CPSO algorithm in the controller. The control parameters are adjusted by BP neural network PID controller through online way and the Jacobian information of controlled plant is real-time identified by radial basis function (RBF) neural network.Finally, the neural network PID control based on CPSO is used for DC voltage control in voltage-source PWM rectifier. The simulation and comparison results show that the response of DC voltage, the harmonic distortion rate of AC current and others performance index are all the best. The feasibility of applying neural network PID based on CPSO in voltage-source PWM rectifier is verified.