Hello, A new project on the LCRC cluster has been requested. Please forward the information on to the LCRC Allocation sub-committee. Applicant's name: Feng Qiu Applicant's institution: ANL Applicant's division: DIS Project Name: PSO Project title: Power System Optimization Associated funding: the U.S. Department of Energy (DOE) Office of Electricity Delivery and Energy Reliability Other Systems: A PC work station at Argonne Science: This LCRC project includes a series projects aiming to improve power system operations. The current projects are power system restoration optimization, and black-start resource selection optimization. The objective is to use mixed-integer linear programming techniques to help power systems to optimize their operations. Two earlier projects in this series have lead to two publications on IEEE Transactions on Power Systems. The restoration project and the resource selection project are also going to power system journals or operations research journals. Project description: Computational methods: Mixed-integer linear programming techniques, including cutting plane algorithms, branch-and-bound scheme, and so on. CPLEX, the solver we normally use, can run in a parallel fashion. That's why each of our computation jobs uses multiple CPU cores. Programming model: optimization models and algorithms are coded with C/ C++ or AMPL language. Algorithms are implemented within MILP solver framework. Scalability issues: Most of our models are highly combinatorial, therefore, in some cases, the branch-and-bound tree in MILP solver can grow very large, consuming a significant amount of memory when data size grows. In our previous experience, A job can run for two weeks until 64GB memory is used up. Therefore, we tend to run each instance exclusively on one node with 64GB memory. We will monitor the scalability issues and adjust the way the jobs are run. Software requirements: MILP solvers such as CPLEX (we have bought the license). C/C++ complier. Performance achieved either on LCRC or other systems: Our programs run for from a couple of minutes to a couple weeks. When the program involves practical data, the computational time tends to be longer(a couple of days, up to 10 days). Sizes of calculations run on and/or planned for LCRC or other systems: The proposed allocation is calculated in the following way: according to our past experience, a project last for 3 months on average. One third of the time is spent on debugging and computation. we tend to run a batch of 10 - 30 jobs with different parameter settings. Each job takes one node (to guarantee the job has sufficient memory). Suppose we run computations intensively for 20 days. Then the total time allocation is 24*20(day)*16(core)*15(job)=115200core hours. Efficiency of the code: The major part of the computation is done by solvers, which is out of our control. Making our code faster is an important goal of our research. Investigation on faster algorithm is part of the research. Expected project members: 1-3 Industry partnership: Project URL: Requested allocation: 460800 Q1: 115200 Q2: 115200 Q3: 115200 Q4: 115200 Justification: Storage requirements: The requester has used undetermined amount hours of their initial startup project. In addition to approving an initial amount, please specify a Category and Subcategory for this project. For a list of the current selection of approved categories, please see: https://wiki.lcrc.anl.gov/wiki/Processes/Categories Once the Allocation committee has approved the project, please go to the Project Management page to create it: https://accounts.lcrc.anl.gov/projects.php Thank You, The LCRC Accounts System