Abstract: Abstract: In order to evaluate the performance prediction accuracy of single-channel pump under computational fluid dynamics, a single-channel pump was taken as the study object. The standard k-ε, RNG k-ε (renormalization group k-ε), standard k-ω and SST k-ω (shear stress transport k-ω) were used to predict the performance of single-channel pump numerically. Meanwhile the grid dependence was checked by employing 5 sets of meshes to improve the computational accuracy. The results of steady and unsteady numerical simulation were compared with that of the experiment on the head and efficiency. The unsteady results were closer to the experimental value. Therefore, the unsteady simulation method should be applied to simulate the internal flow of the single-channel pump. The head, efficiency and power were predicted under 3 different discharges (0.6 Qd, 1.0 Qd, 1.4 Qd, Qd is flow rate on design operating conditions, 220 m3/h) by using CFX (computational fluid dynamics X) 14.0. Energy performance prediction error was analyzed by compared with experimental results. The results showed that there was different degree error between the performance prediction values of the different turbulent models and experimental values. For head prediction of the discharge of 0.6Qd, the standard k-ω model, compared with the other 3 models, had higher prediction accuracy and the head error was 0.008%, followed by the SST k-ω model. However, for efficiency prediction of the discharge of 0.6 Qd, the SST k-ω model had the minimum error value. Therefore, for the simulations at low flow rates, head, efficiency and power errors were 0.38%, 3.12 percentage points and 5.59% respectively with the SST k-ω model. At the design condition, while the head calculation results of the RNG k-ε model were closer to the experimental results than other turbulent models, the standard k-ε model got the best results for the efficiency calculation. The efficiency error of the RNG k-ε model was about 0.1 percentage points lower than the standard k-ε model, so the RNG k-ε model was applied to performance prediction under the design operating condition. When the single-channel pump was operating at the large flow condition (1.4Qd), the RNG k-ε model possessed the highest precision of the head prediction and the standard k-ε model was the best in the efficiency prediction. For comprehensive evaluation of the data, among all 4 turbulent models the RNG k-ε model was the best one to predict the performance of single-channel pumps at the large flow rate. For the internal flow simulation of single-channel pump, the flow separation existed on the blade pressure surface under various operating conditions. As the decrease of the flow rate, the flow near the inlet edge tended to be disordered, especially with a wide range of backflow at the low flow rate. At the same time, there was large backflow on interior areas of the blade pressure surface. There were stagnation points near the inlet edge of blade suction surface and on the backboard. However, serious separation and recirculation flow would occur within the flow passage of the blade pressure side under the low flow conditions. Low pressure on the ring surface of impeller outlet was upstream to the blade outlet and near the outlet. The conclusions in this paper will provide a reliable performance prediction data and practice basis for the single-channel pump, also point the way for developing turbulent models for further study of single-channel pumps.