In recent years, China's local governments have issued numerous bonds to support the country's economic development. However, as total debt accumulates, the pressure on debt repayments is gradually increasing. To increase the sustainability of local government debt, we propose a multi-period stochastic optimization-based approach to determining the portfolio composition of issued bonds, with the goal of minimizing the expected cost under the constraints of liquidity risk and cost deviation risk. Liquidity risk is measured by conditional payment-at-risk (C Pa R), and cost deviation risk is measured by conditional value-at-risk (C V a R). By bounding C V a R and C Pa R, local governments can control the levels of cost deviation risk and liquidity risk. To alleviate future liquidity risk, which is caused by the issuance of a large number of long-term bonds to deliberately reduce repayment pressure within a debt planning horizon, we consider an extended liquidity planning horizon to manage both current and future liquidity risk. Based on this, we analyze the efficient frontier and portfolio compositions of issued bond under the constraints of different C V a R and C Pa R levels. Compared with actual Chinese local government bond portfolios, the efficient frontier performs better for different issuance strategies.