The invention discloses a novel 
photovoltaic inverter control method based on BP neural network and dual-mode structure 
repetitive control. According to the invention the output of the photovoltaic grid-connected 
inverter is connected to a power network after LCL filtering to carry out a sampling link. The phase theta extracted by the improved PLL is fed into the dqalpha beta transform. The difference between the reference 
voltage of the DC side and the actual 
voltage of the DC side is inputted into the PI controller; Obtaining the actual current 
signal ig through abcalpha beta transformationfrom the sampled current of the power network side. The difference between the 
reference grid-connected current ig and the actual current 
signal ig in the alpha beta coordinate 
system after the dqalpha beta transformation is subjected to the alpha beta abc transformation, and then is fed into the 
composite controller together with the actual current 
signal ig and the 
reference grid-connected current ig. The signal processed by the 
composite controller is sent to the notch filter for 
processing, and then the processed signal is sent to the SVPWM module, so as to generate periodic switching signals for controlling the grid-connected 
photovoltaic inverter, thereby suppressing the 
harmonics of the grid-connected 
photovoltaic inverter into the 
power grid. The 
compensation effect of the invention is superior to the traditional photovoltaic grid-connected 
inverter.