A new practical way of modelling direct-driven permanent magnet synchronous generator (PMSG) wind turbines is proposed. The model emphasizes on the wind-rotor-to-PMSG-to-converter-to-grid system, which is the main energy flow system of the direct-driven wind turbine. In this article, a new wind-rotor back propagation neural network is proposed, which consists of a four-layer network and is used to describe the wind-rotor aerodynamic characteristics. According to the orthogonal experimental method, 1200 sets of wind-rotor aerodynamic data, which are calculated based on combining blade element momentum-modified theory with a dynamic stall model, are adopted as the sample data; and the wind-rotor neural network is trained using the Levenberg–Marquardt algorithm. Then, the coupling dynamic models of the wind-rotor and PMSG, and AC–DC–AC converter model are established, respectively; the control strategies for the generator-side and grid-side converters are constructed, too. The mechanical model, electric model, and control model are integrated into the whole simulation model, and the numerical simulation are carried out. The research results show that all the wind-rotor aerodynamic characteristics, electrical characteristics, and control characteristics can be obtained quickly and efficiently from the constructed model, and are helpful for optimization design and control for large-scale direct-direct PMSG wind turbines.