Controller parameter optimization of direct-drive permanent magnet synchronous generator based on SGA-SVM classification

In this paper, a simulation model of direct-drive wind power generation system (WPGS) is established. The training samples of subsynchronous oscillation (SSO) are generated by changing parameters of the grid-side controller during time domain simulation. These samples are classified to be stable and unstable using nonlinear support vector machine (SVM) classification. The simple genetic algorithm (SGA) is used to optimize the hyperparameters of SVM, which improves the accuracy of prediction. Simulation analysis shows that the proposed model provides the possibility to predict the stability of system accurately with the parameter settings of grid-side controller. When the system is predicted as unstable, the grid-side controller parameters can be changed to recommended values to ensure the safety and stability of power system operation.