(328g) A Graph-Based Modeling and Optimization Framework for Complex Systems
PLASMO enables modularization of complex models, because the graph topology is independent of the node models. PLASMO also facilitates the construction of coupled heterogenous networks (e.g., supply chains and natural gas and electric networks), compartmentalizes automatic differentiation and model processing tasks, manipulates graphs to perform diverse model analysis tasks (e.g., network partitioning and aggregation) and manipulates the graph so that the model can be solved solved with standard solvers (e.g., Gurobi,Ipopt) or with structured solvers (e.g., PIPS-NLP  and DSP ). Furthermore, the virtual graph abstraction facilitates the creation of algorithms such as model predictive control , stochastic dual dynamic programming , and Lagrangian relaxation .
This presentation will focus on the modeling aspects of PLASMO and discusses how to use the platform to model and solve coupled infrastructure networks and multi-stage stochastic programming models. The talk will conclude with a motivating example on the centralized and decentralized control of a regional power grid and gas network . We also show how to create and solve multi-stage stochastic formulations for battery models.
 Bezanson, J., Edelman, A., Karpinski, S. & Shah, V. B. Julia: A fresh approach to numerical computing. arXiv 1â37 (2015).
 N. Chiang, C. G. Petra, and V. M. Zavala. Structured nonconvex optimization of large-scale energy systems using PIPS-NLP. In Proc. of the 18th Power Systems Computation Conference (PSCC), Wroclaw, Poland, 2014.
 K. Kim and V. M. Zavala. Algorithmic innovations and software for the dual decomposition method applied to stochastic mixed-integer programs. Optimization Online, 2015
Â J. B. Rawllings and D. Mayne.Â Model predictive control: theory and design. Nob Hill Publishing,Â 2009.
 Shapiro, A. Analysis of stochastic dual dynamic programming method. Eur. J. Oper. Res. 209, 63â72 (2011).