| NSF Org: |
CCF Division of Computing and Communication Foundations |
| Recipient: |
|
| Initial Amendment Date: | August 26, 2015 |
| Latest Amendment Date: | August 26, 2015 |
| Award Number: | 1539462 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Rahul Shah
CCF Division of Computing and Communication Foundations CSE Directorate for Computer and Information Science and Engineering |
| Start Date: | January 1, 2016 |
| End Date: | December 31, 2018 (Estimated) |
| Total Intended Award Amount: | $164,307.00 |
| Total Awarded Amount to Date: | $164,307.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
2550 NORTHWESTERN AVE # 1100 WEST LAFAYETTE IN US 47906-1332 (765)494-1055 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
155 S. Grant St. West Lafayette IN US 47907-2114 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | CyberSEES |
| Primary Program Source: |
|
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.070 |
ABSTRACT
While extensive attention has been given to sustainability in the energy systems, including the subsystems of electricity, petroleum, and natural gas, an oft-overlooked aspect is the interdependence between energy and other infrastructure systems, such as water and transportation systems, and the potential adverse impacts to economics, reliability, and sustainability caused by such interdependence. For example, regulations in the water sector to preserve freshwater may restrict water usage in the power sector, likely causing reduced available generation capacities and hence jeopardizing the reliability of power systems. On the other hand, environmental policies only focused on the power sector, such as those encouraging retrofitting or installing carbon dioxide capture and sequestration capabilities to existing and new coal plants would further constrain the water system as coal plants with carbon sequestration are among the heaviest users of water. Thus, there is a clear need to better understand and manage the interdependence of critical infrastructure systems to promote sustainability across all systems, while not undermining economic and reliability considerations. This proposed work aims to address this need through the theory, modeling and computation of large-scale, interdependent complex systems by way of distributed, highly scalable computing. The results will be widely disseminated through publications and seminars. Further, the project team will leverage established institutional outreach programs to the general public, especially to high-school students and teachers, such as through the Engineering Projects In Community Service program and Purdue?s Energy Academy.
The grand vision of this project is to promote sustainability across interdependent systems, as well as to achieve economic efficiency and to maintain reliability through decentralized yet coordinated management of individual systems by establishing a complete modeling, analytical, and computational framework based upon the general class of augmented Lagrangian methods originating from convex optimization. While the augmented Lagrangian method is not a new algorithm, the current implementation of such algorithms has not taken advantage of its distributed feature, which would be particularly suitable to deal with large-scale, interlinked systems. One of the major goals of this work is to establish the theoretical foundations of distributed Lagrangian methods and to implement the algorithms on supercomputer clusters to demonstrate the benefits of distributed computing. This research aims to pave the way for cloud computing such that the algorithms can be used by decision-makers even without access to supercomputers. Another contribution is that the augmented Lagrangian method algorithms will be extended to incorporate stochastic data, both in terms of theoretical issues such as algorithm convergence as well as practical implementation. The computational methods will be tested and validated through real-world models of interdependent power and water systems.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
The aim of this research project is to help identify least cost options (in terms of short-term system operations and long-run infrastructure investment) to achieve both sustainability and reliability in the water and electric energy system, through quantitative modeling and computational approaches. This work is directly motivated by the concern that policies or market rules that promote sustainability in one system may have negative impact on the other interconnected systems, such as regulations in the water sector to preserve freshwater may restrict water usage in the power sector, likely causing reduced available generation capacities and hence jeopardizing the reliability of power systems.
While one potential solution to solve the interdependency issues is to have an overarching agency overseeing all the interdependent infrastructure systems. Such a solution is not only infeasible from the perspectives of government rules or federal versus state regulation, it is also infeasible from decision-support perspective, as a multi-system model with high fidelity representation of each infrastructure system would lead to an extreme scale problem that is computationally infeasible to provide good policy analysis. This collaborative research is to directly address this issue by designing and implementing distributed algorithms; i.e., a process to let each system (e.g, electric power and water) repeatedly perform their own usual operations and planning optimization. Along the iterative optimization process, however, a set of carefully designed information is to be shared among all the interdependent systems. Under proper market designs (i.e., algorithm rules), such an iterative process (aka a distributed process) can lead to cross-system optimal solution. Significant efforts of the PI at Purdue University have been dedicated to design such algorithms, with provable convergence to a cross-system optimal solution, and with scalability so that they can be applied to solve extremely large-scale problems over distributed computing infrastructures (such as through GPUs, cluster computing, or cloud computing).
Two such algorithms have been developed through the course of this project, with different emphases. The first algorithm, representing a significant extension of an existing algorithm, is referred to as the asynchronous PCPM (predictor corrector proximal multiplier) method. Its key feature is that while multiple systems conduct their own optimization (for operations or planning), they do not have to wait for all other systems to share the needed information (such as pricing of electricity and water usage of all power plants) before moving along to the next iteration of optimization, hence the name of “asynchronous.” Such a feature is key for implementing a distributed optimization algorithm over interdependent systems where different systems operate at different temporal resolutions. The PCPM algorithm is also extended to solve decision making problems under future uncertainties. More specifically, a new decomposition approach based on the PCPM algorithm is applied to stochastic optimization problems with recourse actions, and it can decompose an extremely large-scale problem into a large-number of low-dimension (or even one-dimension) optimization problems, which will be amenable to high performance computing, and hence has the potential to solve extreme-scale problems that cannot be solved existing algorithms.
The other algorithm, named the modified augmented Lagrangian method (or the MAL method), is designed to solve a broader class of problems, termed the generalized Nash equilibrium problems (GNEPs), in which each “player” (such as the power system itself viewed as a player, or the individual power plants) aims to minimize their own costs, while knowing that both their costs and feasible set of actions are affected by other “players” actions as well. The multi-system joint optimization problem is a special case of a GNEP. The MAL algorithm has been proven to be able to find a Nash equilibrium for broadest class of (convex) GNEPs. In addition, the algorithm naturally leads to parallel implementation, again making the algorithm scalable to solve large-scale problems.
In addition to algorithm development, the PI has been actively working in creating a joint water and electric energy system test case based on real-world data representing California. A website is being constructed to share such a test case and data, with the purpose of helping broad community of researchers to conduct their own research in improving joint systems’ sustainability and reliability. Research findings resulted from this grant have been broadly disseminated to research communities both within the U.S. and internationally.
Last Modified: 04/04/2019
Modified by: Andrew L Liu
Please report errors in award information by writing to: awardsearch@nsf.gov.
