Abstract: Under the trend of large-scale deployment of renewable energy, aiming at the increasingly severe transient power quality problems caused by high-frequency power electronic equipment such as inverters, a complex power quality disturbance identification method based on modified multi-fusion variational mode decomposition and improved support vector machine is proposed to improve power quality. Firstly, the improved variational mode decomposition algorithm uses the principle of the maximum total energy ratio of the mode to adaptively adjust the number of disturbance mode decomposition layers, thereby constructing the optimal disturbance mode eigenvector. Then, by constructing a multi-fusion combined kernel function structure, the effective mapping and fusion of the original disturbance signal and the disturbance mode feature vector is realized, and an optimized support vector machine model is formed to improve the accuracy of disturbance recognition. Finally, the results of multiple comparative experiments show that the proposed Multi-fusion MVMD-ISVM model can achieve the effective feature extraction and identification of disturbance signals, and has a higher recognition effect and anti-noise performance, which is suitable for accurate and fast recognition of complex power quality disturbance signals in modern power grids.