报告题目:A stochastic dual dynamic programming approach for solving two-stage distributionally robust constrained optimization problems
报 告 人:童小娇
报告时间:2023年5月6日下午3:30-4:30
报告地点:数学楼311会议室
摘 要:In this paper, we provide a stochastic dual dynamic programming (SDDP) method to solve a class of two-stage distributionally robust constrained optimization problems which comes from the power system or revenue management. Through Shapiro’s duality theorem, a robust counterpart for the two-stage problem is derived. Then we construct a constrained-based approximation model and an objective-based approximation model which is an convex implicit semi-infinite programming under some moderate conditions. A stochastic dual dynamic programming method is proposed to solve the objective-based approximation model. To circumvent the difficulty of calculating the subgradient of the objective function, we propose a CVaR-based sample average approximate method for the robust counterpart and a linear approximation for optimal value functions. Convergence analysis for the approximation approaches are provided and numerical tests are given to show the effectiveness of the SDDP algorithm.
报告人简介:
童小娇,湖南第一师范学院 博士、省二级教授,国务院政府特殊津贴专家,湘潭大学博士生导师,湖南师范大学、长沙理工大学硕士导师。是中国运筹学会第十一届副理事长、中国工业与应用数学学会第七届常务理事。是国家首批一流专业“数学与应用数学”负责人、湖南省双一流数学学科负责人。以第一主持人获湖南省自然科学二等奖2项、湖南省教学成果一等奖1项。主持国家自然科学基金面上项目6项、天元专项4项。曾任湖南第一师范学院校长,湖南省第十三届人大常委会委员,政协湖南省第十一届常委,湖南省运筹学会第一、二届理事长。