In order to explore the impact of uncertainty in recycling quantity, demand quantity, and their associated risks on the power battery closed-loop supply chain network structure, a conditional-risk-at-value (CVaR) measurement method is adopted with the maximum profit as a goal, a mixed integer programming model under certain environments and a two-stage stochastic programming model under risk-neutral and risk-averse situations are constructed, respectively. The scenario method is used to deal with parameter uncertainty, and an improved genetic algorithm based on hybrid coding is proposed to solve models. The effectiveness of the stochastic model and the algorithm is verified through solving examples. The results show that：under a certain environment, profits increase as the recycling utilization rate increases； in the risk-averse situation, the increase in the degree of risk aversion reduces profits. Therefore, decision makers need to consider how much profit to sacrifice to deal with risks. Comparing the results of three situations, it is found that considering the uncertainty of parameters and their adverse effects will reduce network profits and change the network structure.