Abstract: To test and evaluate the decision-making ability of intelligent ships during navigation under different collision risks and increase the probability of encountering situation scenarios with controllable collision risks. This paper proposes a method for generating test scenarios of ship collision risk based on conditional variational autoencoder (CVAE). Based on the data of the Automatic Identification System (AIS) for ships, the relative motion parameters of ships are calculated. Combined with the "International Rules for Collision Avoidance at Sea" and the fuzzy rule base, the ship encounter situation and the ship collision risk level are classified to construct the ship encounter - risk data set. The CVAE model is trained using the dataset. The relative motion parameters of ships are taken as the input features of the model, and the encountering situation and the risk level of ship collision are taken as the condition variables of the model. After the model training is completed, input the condition variables and the initial state of the test vessel to generate the corresponding test scenarios. At the same time, with a step size of 10 seconds, dynamically display the nearest encounter distance, the shortest encounter time and the collision risk level of the vessel during navigation. A comparative experiment was conducted with the variational autoencoder (VAE) and random sampling methods to verify the accuracy and similarity of the generation results of different models. The results show that under the given conditions of encountering situations and collision risk levels, the accuracy rate of the test scenarios generated by the CVAE model that simultaneously meet the two targets is 93.54%. Compared with the VAE model (0.08%) and the random sampling method (0.83%), the accuracy rates have increased by 93.46% and 92.71% respectively. Therefore, the accuracy rate of the encounter situation proposed in this paper has been improved in terms of effectiveness, diversity, and authenticity in the generation of various encounter situations and ship collision risk scenarios.