In view of the possible human errors in the development process of maritime autonomous surface ship (MASS), the identification method and evaluation model of human error risk in L2 level MASS collision avoidance process are proposed from the perspective of risk identification and risk assessment. Firstly, the critical tasks of the remote operators and crews in the MASS collision avoidance process were obtained based on the hierarchical task analysis method (HTA), and the problem solving and decision making model (IDAC) was used to establish the cognitive model to decompose each operation stage of collision avoidance task and identify the possible human errors in the collision avoidance process. Secondly, the failure mode and impact analysis (FMEA) method was used to model the risk level of the identified risk factors by combining evidence reasoning (ER). Finally, the risk of human error in MASS collision avoidance process is quantified and ranked, compared with the existing research results, and necessary suggestions are put forward. The results show that the biggest risk factor is the uncoordinated human-machine system interaction in the decision-making stage during L2 level MASS collision avoidance process.