Parallel Accelerated Optimization Techniques for Large-Scale Power System Planning and Operation

• Author / Creator

  Huang, Shengjun

• Electricity is ubiquitous in modern life, nevertheless, its availability and reliability are not granted. Actually, providing reliable electricity is an enormously complex technical challenge even on the most routine of days, which requires trained and skilled operators, sophisticated computers and communications, and sufficient planning and designing. A broad set of interrelated decisions should be made for scheduling and operating various types of devices for electricity generation, transmission, and distribution subject to engineering, market, and regulatory constraints. Due to the nonlinear and non-convex power flow equality constraints, discrete control and decision variables, and large system scales, the decision-making process is extremely daunting that demands a high level of computational intelligence and speed. In addition, developments in the power industry, such as the introduction of renewable energy and smart grid, have brought new challenges on this tough issue, including uncertainty factors and real-time responses. Therefore, most conventional off-line optimization tools are required to be merged into online modes. On the other hand, rapid advances in digital computers, smart meters, and communication technologies, etc., provide great opportunities and powerful tools. In the context of coexisting challenges and opportunities, possibilities for the acceleration of various optimization techniques for large-scale power system planning and operation problems are investigated in this thesis via algorithm customization, framework development, and parallel processing. Eight fundamental problems are investigated with the coverage of all three major domains of power systems. Four categories are separated for the classification of them in this thesis. Planning of transmission system. Due to the boost of loading levels and the wide utilization of distributed generators, Transmission Expansion Planning (TEP) has regained its significance for investigation. In this thesis, a Multi-Group Particle Swarm Optimization (MGPSO) algorithm is proposed to solve DC TEP based on the multi-group co-evolution strategy and Linear Equation System (LES) transformation. Superiority over commercial software Lingo is established with case studies. In addition, based on the disjunctive model, Security Constrained TEP (SCTEP) problem is formulated into a Mixed-Integer Linear Programming (MILP) problem. Branch-and-Cut Benders Decomposition (BCBD) method is developed with the integration of BD into a B&C framework, resulting in better performance over MILP solver Cplex. Operation of transmission system. Alternating Current Power Flow (ACPF) analysis is one of the most fundamental tasks for the transmission system operation and optimization problems, such as Real-Time Contingency Analysis (RTCA). Exploration on the single ACPF solution has been conducted on Graphics Processing Unit (GPU) with Fast Decoupled (FD) method, where both Matlab and Compute Unified Device Architecture (CUDA) are employed for programming. Due to the limited performance, RTCA which comprises multiple ACPFs is addressed with Compensation Method (CM). Based on the sensitivity analysis of similar ACPFs, the number of matrix decomposition has been greatly reduced. Good performance on accuracy, convergence, and scalability of CM running on GPU with CUDA has been validated. Operation of generation system. Optimal operation of generation system dominates the economy and security of the whole electricity supply process. In this thesis, Security Constrained Unit Commitment (SCUC) and Real-Time Optimal Power Flow (RTOPF) are solved to determine the on-off status and the amount of active power output of each thermal generator, respectively. Potential of different Robust Optimization (RO) frameworks are fully discussed within the context of parallel computing. Minimization of the prediction error with the consideration of renewable energy is investigated for RTOPF based on the GPU parallel processing. Operation of distribution system. Due to the low voltage level of distribution network, significant power losses are encountered. In order to minimize them, two problems are investigated from different aspects: Distribution Network Reconfiguration (DNRC) and Real-Time Volt/Var Optimization (RTVVO). Two major concerns of DNRC have been addressed with the proposed efficient decimal solution encoding and decoding strategy and the acceleration of distribution network power flow (DNPF) process. In terms of RTVVO, detailed formulation of transformers and other devices is integrated into the Direct Approach (DA). Well-established data structure and thread organization pattern are also provided for GPU implementation.

• Subjects / Keywords

  • Parallel Processing
  • Large-Scale Power Systems
  • Graphics Processing Unit
  • Alternating Current Power Flow
  • Optimization Algorithms

• Graduation date

  Spring 2018

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R37H1F277

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.