Aiming at the current problem that it is difficult to diagnose early soft faults in DC-DC circuits. An early fault diagnosis model for circuit soft faults based on CPO-VMD is proposed. Firstly, for the problem that the decomposition performance of VMD is affected by the number of decomposition layers K and the penalty factor, the Crown Porcupine optimization algorithm is used to optimize the number of decomposition layers K and the penalty factor of VMD; secondly, The fitness function is established by integrating the minimum envelope entropy with the minimum sample entropy to form a composite criterion, and the sensitive modal components are selected according to the principle of the minimum fitness to form the new eigenvectors; lastly, the new eigenvectors are inputted to the SVM classification model for diagnosis. Experiments show that the method can adaptively determine the optimal parameter combinations for VMD decomposition. By comparing and analyzing with VMD, EMD and EEMD, it has better fault diagnosis performance and can realize early soft fault diagnosis of DC-DC circuits.