The hybrid permanent magnet (PM) variable flux memory machine employs high- and low-coercive-force hybrid PMs to simultaneously obtain a high torque density and regulable PM flux property. The operating trajectories and steady working-points of this type of machine switch back and forth in multiple regions. To identify operating modes under various PM magnetisation states and speed ranges, this paper proposes a mode recognition and coordinated control method based on probabilistic neural networks (PNNs). A PNN algorithm is trained offline and executed in real-time to obtain a certain operating mode with the maximum probability of current state. According to the mode recognition result, a coordinated control scheme for applying short-time pulses or continuous flux-weakening current in a reasonable manner is presented. The proposed method exhibits an excellent capability in real-time operating modes identification, hence achieving magnetisation current control of multiple modes easily. Its feasibility and effectiveness are validated by experimental results on a hybrid PM VFMM prototype.