2009 Poster Sessions : Reservoir Management Based on Approximate Dynamic Programming

Student Name : Zheng Wen
Advisor : Benjamin Van Roy
Research Areas: Information Systems
Abstract:
Petroleum reservoir management is in general a large-scale, nonlinear dynamic optimization problem. In this project, we propose computational methods to solve this problem based on Approximate Dynamic Programming (ADP). Specifically, we generate features of the approximate cost-to-go function based on Proper Orthogonal Decomposition (POD) and compute the weights of features based on Temporal-Difference Learning (TD-Learning) and Reduced Linear Programming (RLP). Some simulation results based on this algorithm are also demonstrated.


Bio:
Zheng Wen is mainly interested in mathematical modeling and optimization. He is currently working with Prof. Benjamin Van Roy on Reservoir Management based on Approximate Dynamic Programming. He is currently a second-year PhD student in Stanford, majoring in EE.