A robust fault identification method is proposed to address the challenges of insufficient fault feature extraction and low identification accuracy in distribution networks with noisy environments and complex disturbances, which integrates multidimensional entropy features with a cooperative optimization model. Firstly, the zero-sequence current signal is denoised by using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN). It innovatively integrates eight heterogeneous entropy features to construct a high-dimensional complexity space that can deeply describe the nonlinear dynamic characteristics of faults, thereby enhancing the intrinsic anti-disturbance ability of the features. Secondly, a collaborative diagnostic model is developed, in which Kernel Principal Component Analysis (KPCA) is employed to extract the most fault-sensitive low-dimensional features, while an Improved Sparrow Search Algorithm (ISSA) is utilized to jointly optimize the penalty factor and kernel function parameters of the Support Vector Machine (SVM). Finally, a fault classification model for distribution networks is established based on multidimensional entropy feature fusion using KPCA-ISSA-SVM. Simulation studies on the IEEE 33-bus system demonstrate that the proposed method achieves superior fault identification accuracy under various signal-to-noise ratio conditions, thereby validating its effectiveness and advancement.