The invention discloses an
inverter repetitive control design method based on a neural network. The method includes the following steps: 1, analyzing that traditional
transfer function of the inverterby using the
control theory, selecting an appropriate
control algorithm for the bottom control of the
inverter, identifying the theoretical model offline by use the neural network
algorithm, and obtaining the initial parameters of the model by selecting the basic structure of the model; 2, taking the obtained initial identification model as a reference for on-line identification learning, and obtaining the actual identification model of the traditional control
closed loop of the
inverter; 3, constructing an
inverse transfer function by utilizing that identification model information to replace the repeat control compensation link; 4, storing that parameters of the identify model in real time, and importing the latest data of the learning model of the
system every time the identification model is put into the device. The method of the invention saves a large amount of tedious work of
parameter design and selection, avoids the influence of parameter drift on
repetitive control, improvesthe stability and robustness of
repetitive control links, and can be popularized and used in inverters and related fields.