The wind generation process is very sensitive to uncertainties based on environmental effects due to the rapid growing wind capacity. In the situation where no means of grid connection exist, the variation in power could compromise safety of power system operations. In this paper, a model of the wind-energy storage system with a doubly-fed generator model is presented. Seven groups of operating states are determined based on the frequency dynamic response performance of the wind-storage system. To enhance the recognition of wind-storage operating states, a novel model (CTM-Net) is proposed based on the combination of multilayer Transformer and Convolutional Neural Network (CNN). When a wind-power failure occurs, the probabilities of change in value of output between 0 and 4 MW and 76 and 80 MW are considered and energy storage devices are allocated considering the chargeable and dischargeable ratio of the system. It has been demonstrated that the wind-storage operating states have obvious variations at different sampling times. The faulty shutdown state (ZT2) identification rate is 99.57% which encourages the effectiveness of state recognition significantly. The operating state of the wind-storage system can be identified accurately using multilayer adaptive Transformer.
