Abstract: For the high-voltage direct current（DC） sending system containing large-scale wind power, faults such as commutation failure and DC blocking may lead to transient overvoltage on alternating current（AC） bus at the sending end, leading to the tripping risk of renewable energy units and posing a serious threat to the safe and stable operation of power grid. To characterize the impact of DC faults on the safe and stable operation ability of the system, the transient overvoltage amplitude prediction method for the DC sending-end systems based on knowledge-data fusion model is proposed. Firstly, an improved decision tree algorithm is proposed by modifying the branch quality measurement index during the decision tree splitting process. The mapping relationship between the transient overvoltage and the key electric quantities is fully explored, and the evaluation performance of the overvoltage amplitude prediction model is improved. Secondly, based on the system mechanism model, the correlation between the power exchange situation of the sending-end system and the AC bus voltage is analyzed, and the model driven analytical expression for the transient overvoltage amplitude of the converter station bus is derived. Finally, combining theoretical analysis methods, a knowledge-data fusion model is established to quickly correct the theoretical calculation results of the overvoltage amplitude and improve the robustness of the fusion model in the case of insufficient training samples or improper input feature selection. The simulation results verify the effectiveness of the proposed transient overvoltage amplitude prediction method for the DC sending-end system with transient overvoltage problem.