Unit commitment in power systems

• 主要作者: Johnston, Thorfinn James
• 其他作者: Ponnuthurai Nagaratnam Suganthan
• 格式: Final Year Project
• 語言: English
• 出版: 2017
• 主題: DRNTU::Engineering::Electrical and electronic engineering::Electric power::Production, transmission and distribution
• 在線閱讀: http://hdl.handle.net/10356/70933
• 機構: Nanyang Technological University

A worldwide increase in energy consumption is being observed and the task of meeting this demand at the lowest possible production cost is becoming much more important. The financial savings which can be achieved, and the need to preserve the planet’s depleting fossil fuels are key motivators for optimizing generation. The Unit Commitment Problem is the resultant algorithm which encapsulates the aspects of meeting demand at the lowest cost whilst simultaneously adhering to several other constraints. These constraints relate to the system as a whole, and also to the individual constraints unique to the generating equipment being used. A schedule of generators is created based on demand before optimizing the quantity of generation by each committed unit. Different optimization methods are investigated at the outset and the chosen method implemented in this project is Differential Evolution. This optimization technique begins by initializing a population before iteratively generating offspring populations through three steps; mutation, crossover and finally selection of the best solution – chosen from the most optimal of parent or offspring population. This project implements several variants of Differential Evolution in the context of the Unit Commitment Problem using MATLAB. Different problem dimensions are also applied and results obtained to allow for evaluation of the approaches used. Both manual adjustment of control parameters is undertaken to demonstrate the effect of this as well as implementing adaptive algorithms. Comparing the results obtained using the different methods show the most effective strategies implemented to solve this real-world problem to be SHADE, DE/rand/1 and DE/rand/2 – all of which, in this context, provided more optimal results compared to several other methods including a proposed two subpopulation strategy. To gain a wider appreciation of other recent optimization methods, research is conducted for comparison with the results obtained and possible ways of further improvements are able to be identified.