In household electrical appliances, the number of nonlinear load appliances is gradually increasing. Traditional arc fault detection methods based on time domain features of current are unable to accurately identify fault phenomena. Therefore, this paper proposes arc fault identification method based on multi-dimensional features and MLP-SVM. Firstly, the Variational Mode Decomposition algorithm is used to decompose current to get the mode components. Then the fuzzy entropy of each mode component is calculated. Finally, a multi-dimensional feature vector is constructed with time domain and fuzzy entropy features, and the MLP-SVM model is used for classification decision to identify arc fault. Experimental results show that compared to other methods, the proposed method can achieve arc fault identification accuracy of up to 99%. This method is also suitable for low-voltage distribution arc fault identification with typical load and nonlinear load.