The invention discloses a design method for an LQG controller of an LPV system based on data driving. The method comprises steps of S1, building a system equation, carrying out the equal-time-intervalsampling of sensor inputs under different parameter conditions, carrying out the order reduction of the sensor inputs through a POD method, and obtaining an order reduction basis; S2, under the condition of different parameters, a system filtering algebra Riccati equation and a controller algebra Riccati equation being calculated respectively, and a filter gain and a controller gain being obtained; S3, a database being established, reduced-order bases of all prior parameters and the two gains obtained in the S2 being in one-to-one correspondence and stored in the database, the reduced-order bases serving as learning bases of a machine learning classifier, a proper machine learning model being selected, the learning bases in the database serving as a training set, the two corresponding gain parameters serving as labels, and the classifier model being designed; wherein the classifier model is the LQG controller. The controller can be used online in real time. According to the method, the stable reaction speed and control precision of the system are effectively improved, and the interference can be greatly reduced.