Papers by Peter Yun Zhang
Journal of Scheduling
This paper presents a three-stage algorithm for resource-aware scheduling of computational jobs i... more This paper presents a three-stage algorithm for resource-aware scheduling of computational jobs in a large-scale heterogeneous data center. The algorithm aims to allocate job classes to machine configurations to attain an efficient mapping between job resource request profiles and machine resource capacity profiles. The first stage uses a queueing model that treats the system in an aggregated manner with pooled machines and jobs represented as a fluid flow. The latter two stages use combinatorial optimization techniques to solve a shorter-term, more accurate representation of the problem using the first stage, long-term solution for heuristic guidance. In the second stage, jobs and machines are discretized. A linear programming model is used to obtain a solution to the discrete problem that maximizes the system capacity given a restriction on the job class and machine configuration pairings based on the solution of the first stage. The final stage is a scheduling policy that uses the solution from the second stage to guide the dispatching of arriving jobs to machines. We present experimental results of our algorithm on both Google workload trace data and generated data and show that it outperforms existing schedulers. These results illustrate the importance of considering heterogeneity of both job and machine configuration profiles in making effective scheduling decisions.
Lecture Notes in Computer Science 7874 (CPAIOR Proceedings 2013)
The wind farm layout optimization problem is concerned with the optimal location of turbines with... more The wind farm layout optimization problem is concerned with the optimal location of turbines within a fixed geographical area to maximize energy capture under stochastic wind conditions. Previously it has been modelled as a maximum diversity (or p-dispersion-sum) problem, but such a formulation cannot capture the nonlinearity of aerodynamic interactions among multiple wind turbines. We present the first constraint programming (CP) and mixed integer linear programming (MIP) models that incorporate such nonlinearity. Our empirical results indicate that the relative performance between these two models reverses when the wind scenario changes from a simple to a more complex one. We also propose an improvement to the previous maximum diversity model and demonstrate that the improved model solves more problem instances.
Proceedings of the ASME International Design Engineering Technical Conferences (2012)
Wind farm design deals with the optimal placement of turbines in a wind farm. Past studies have f... more Wind farm design deals with the optimal placement of turbines in a wind farm. Past studies have focused on energymaximization, cost-minimization or revenue-maximization objectives. As land is more extensively exploited for onshore wind farms, wind farms are more likely to be in close proximity with human dwellings. Therefore governments, developers, and landowners have to be aware of wind farms' environmental impacts. After considering land constraints due to environmental features, noise generation remains the main environmental/health concern for wind farm design. Therefore, noise generation is sometimes included in optimization models as a constraint. Here we present continuous-location models for layout optimization that take noise and energy as objective functions, in order to fully characterize the design and performance spaces of the optimal wind farm layout problem. Based on Jensen's wake model and ISO-9613-2 noise calculations, we used single-and multiobjective genetic algorithms (NSGA-II) to solve the optimization problem. Preliminary results from the biobjective optimization model illustrate the trade-off between energy generation and noise production by identifying several key parts of Pareto frontiers. In addition, comparison of single-objective noise and energy optimization models show that the turbine layouts and the inter-turbine distance distributions are different when considering these objectives individually. The relevance of these results for wind farm layout designers is explored.
Conference Presentations by Peter Yun Zhang
International and regional guidelines:
M.A.Sc. Thesis by Peter Yun Zhang
Wind farm layout optimization (WFLO) is the design of wind turbine layout, subject to various fin... more Wind farm layout optimization (WFLO) is the design of wind turbine layout, subject to various financial and engineering objectives and constraints. The first topic of this thesis focuses on solving two variations of WFLO that have different analytical aerodynamic models, and illustrate deep integration of the wake models into mixed-integer programs and constraint programs. Formulating WFLO as MIP and CP enables more quantitative analysis than previous studies could do with heuristics, and allows the practitioners to use an off-the-shelf optimization solver to tackle the WFLO problem. The second topic focuses on another version of WFLO that has two competing objectives: minimization of noise and maximization of energy. A genetic algorithm (NSGA-II) is used. Under these two objectives, solutions are presented to illustrate the flexibility of this optimization framework in terms of supplying a spectrum of design choices with different numbers of turbines and different levels of noise and energy output. ii I am incredibly fortunate to have Professor Cristina Amon as my supervisor, who has provided unfaltering support and guidance through my MASc journey. I am grateful for her welcoming me into the ATOMS lab and allowing me to pursue my academic interests freely. Her work ethics and ability to discover important research questions are inspirational. Since my first day at ATOMS, Dr. David Romero has been a supportive and resourceful mentor. Without David's confidence in me, I would not have appreciated research so much. David's depth and breadth of knowledge, along with his effective supervision style, are instrumental for my research progress. Sitting beside him brought us many interesting discussions from research to culinary art. Professor Timothy Chan has been an academic and career mentor to me through the many courses and projects that we are both part of. Thank you for fostering my passion for operations research and re-kindling my love for mathematics. Discussions with Professor Chris Beck through various research projects and the CP&LS course honed my critical thinking skills, and brought me one step closer to being a good researcher. have taught me many things in research: work ethics, communication, and research collaboration. Michael Kim, unknowingly, inspired part of my work through his award-winning paper. He probably still does not know if he has not read this paragraph. The Hatch team, especially Michael Morgenroth and Joaquin Moran, brought insights from the industry, without which I would not have been able to write this thesis.
Uploads
Papers by Peter Yun Zhang
Conference Presentations by Peter Yun Zhang
M.A.Sc. Thesis by Peter Yun Zhang