With the continuous development and expansion of supply chain finance business, it is necessary for policymakers to reduce the concentration risk cased by sharp fluctuations of price of single pledge and keep the flexibility of the supply chain finance business. To this end, portfolio methods has been successfully applied by financial institutions for selecting different pledges to optimize the multi-period loan portfolio in supply chain finance. As we all known, Copula technique is flexible to capture the nonlinear dependence structures among assets, which is very important for portfolio in practice. In this paper, a Copula-quantile regression method is proposed by employing quantile regression to fit marginal distribution of a single asset and Copula function to capture nonlinear dependence structures among assets. Our method is able to avoid the model specification errors without assumption of the distribution of random disturbance term. Most importantly, it is flexible and adapted to describe stylized facts in supply chain finance, such as asymmetry and nonlinearity. The Copula-quantile regression method for optimizing the multi-period loan portfolio in supply chain finance consists of two steps. The Copula-quantile regression method is firstly applied to predict the multi-period loan return. Then, a decision-making scheme is provided for the loan portfolio by minimizing the traditional Sharpe ratio and the generalized Omega ratio. To illustrate the efficacy of our method, an empirical research is conducted on the spot of aluminum and copper which are the most common form of the pledge in supply chain finance. At least two facts can be drawn from the empirical results. First, the t-Copula function in Copula-quantile regression method is always optimal for all periods in term of AICs, which indicates that the correlation between aluminum and copper is a fat tail version. Second, the Copula-quantile regression method outperforms the Copula-GARCH in that the former poses higher Sharpe ratio and generalized Omega ratio than the latter for all portfolios at different periods, and provides a more reliable decision-making reference for the healthy development of supply chain finance. In the future, considering more assets in a portfolio has practical significance for policymakers. To address this issue, our method can be extended to vine-Copula-quantile regression through combining vine-Copula approach with quantile regression model. It can be expected that vine-Copula-quantile regression method can effectively handle the problem of selecting more pledges to construct multi-period loan portfolio in supply chain finance. This is left for future research.