In the research of identification of shaft orbit of hydropower generating unit, the selection of feature parameter in traditional SVM system is not adaptive, which results in lower classification performance and long computation time. Aiming at the problems above, this paper proposes a novel method to identify the shaft orbit based on HABC-SVM. Artificial bee colony is introduced to the solution of SVM identification optimal model, and the search strategy, food source and update equation of artificial bee swarm are improved. Through the simulation experiment, four typical samples of shaft orbit of hydraulic turbines are obtained, the 19 kinds of feature parameters extracted from shaft orbit and parameters of SVM are optimized synchronously, and the improved HABC algorithm is compared with PSO-SVM algorithm and GA-SVM algorithm. The results show that HABC-SVM has good adaptability and classification accuracy, can acquire the optimal solutions of SVM parameters and feature subset synchronously, enhance the performance of classifier, and improve the precision of identification of shaft orbit, which has some guidance significance to fault diagnosis of hydropower generating unit.