To effectively assess the transient overvoltage (TOV) risk of renewable energy stations in multi-DC sending-end systems under DC blocking fault scenarios, this paper proposes a TOV magnitude prediction method that considers the mutual interactions among renewable energy stations. First, an analytical expression for the TOV magnitude at the grid-connection points of renewable energy stations in the sending-end system caused by DC blocking faults is derived. Then, an approximate analytical expression is proposed to characterize the relationship between the multiple renewable energy stations short-circuit ratio (MRSCR) and the TOV magnitudes at the grid-connection points of renewable energy stations. Given that MRSCR can quantify the coupling degree (i.e., the mutual interaction) among renewable energy stations, a TOV magnitude prediction method for renewable energy stations under DC blocking scenarios is developed based on a knowledge-embedded neural network. By incorporating a regularization term associated with MRSCR into the loss function, the proposed model ensures that the TOV magnitude prediction adheres to the physical constraints of power systems, thereby improving prediction accuracy. Finally, the proposed method is validated on a practical power system in a region of China. The results demonstrate that, compared with conventional TOV magnitude prediction methods, the proposed method incorporating the mutual interactions among renewable energy stations can significantly enhance the prediction accuracy.