专家简介：
邓天虎，博士，副教授。目前就职于清华大学工业工程系。2013年于美国加州大学伯克利分校获得工业工程与运筹博士学位，2008年于清华大学工业工程系获得学士学位。目前研究方向侧重智慧供应链。以第一作者和通讯作者在Manufacturing & Service Operations Management、Operations Research等国际高水平学术期刊和学术会议发表论文20余篇。
报告简介:
We study the energy consumption minimization problems of natural gas transmission in gunbarrel structured networks. In particular, we consider the transient-state dynamics of natural gas and a compressor's nonlinear working domain and ramp constraints. We formulate the problem as a two-level dynamic programming (DP), where the upper-level DP problem formulates each compressor station as a decision stage and each stage's optimization problem is further solved as a lower-level DP by setting each time period as a stage. Both dynamic programming problems are high-dimensional. We propose an approximate dynamic programming (ADP) approach for the upper-level DP and an exact approach for the lower-level DP. Finally, we numerically compare our ADP approach to the existing simulated annealing (SA) heuristics based on real-world data in China. The results demonstrate that the ADP saves about 4.1%∼7.5% energy consumption than the SAs. Furthermore, the solution found by the ADP within the first 20 seconds dominates the solutions found by SAs in more than ten minutes. The advantages in solution performance and computation time support the proposed ADP algorithm in practice.
邓天虎，博士，副教授。目前就职于清华大学工业工程系。2013年于美国加州大学伯克利分校获得工业工程与运筹博士学位，2008年于清华大学工业工程系获得学士学位。目前研究方向侧重智慧供应链。以第一作者和通讯作者在Manufacturing & Service Operations Management、Operations Research等国际高水平学术期刊和学术会议发表论文20余篇。
报告简介:
We study the energy consumption minimization problems of natural gas transmission in gunbarrel structured networks. In particular, we consider the transient-state dynamics of natural gas and a compressor's nonlinear working domain and ramp constraints. We formulate the problem as a two-level dynamic programming (DP), where the upper-level DP problem formulates each compressor station as a decision stage and each stage's optimization problem is further solved as a lower-level DP by setting each time period as a stage. Both dynamic programming problems are high-dimensional. We propose an approximate dynamic programming (ADP) approach for the upper-level DP and an exact approach for the lower-level DP. Finally, we numerically compare our ADP approach to the existing simulated annealing (SA) heuristics based on real-world data in China. The results demonstrate that the ADP saves about 4.1%∼7.5% energy consumption than the SAs. Furthermore, the solution found by the ADP within the first 20 seconds dominates the solutions found by SAs in more than ten minutes. The advantages in solution performance and computation time support the proposed ADP algorithm in practice.
