Date of Award

December 2020

Document Type


Degree Name

Doctor of Philosophy (PhD)


Industrial Engineering

Committee Member

Scott J Mason

Committee Member

Mary E Kurz

Committee Member

Harsha Gangammanavar

Committee Member

Yongjia Song


With globally increasing energy demands, world citizens are facing one of society's most critical issues: protecting the environment. To reduce the emission of greenhouse gases (GHG), which are by-products of conventional energy resources, people are reducing the consumption of oil, gas, and coal collectively. In the meanwhile, interest in renewable energy resources has grown in recent years. Renewable generators can be installed both on the power grid side and end-use customer side of power systems. Energy management in power systems with multiple microgrids containing renewable energy resources has been a focus of industry and researchers as of late. Further, on-site renewable energy provides great opportunities for manufacturing plants to reduce energy costs when faced with time-varying electricity prices. To efficiently utilize on-site renewable energy generation, production schedules and energy supply decisions need to be coordinated. As renewable energy resources like solar and wind energy typically fluctuate with weather variations, the inherent stochastic nature of renewable energy resources makes the decision making of utilizing renewable generation complex. In this dissertation, we study a power system with one main grid (arbiter) and multiple microgrids (agents). The microgrids (MGs) are equipped to control their local generation and demand in the presence of uncertain renewable generation and heterogeneous energy management settings. We propose an extension to the classical two-stage stochastic programming model to capture these interactions by modeling the arbiter's problem as the first-stage master problem and the agent decision problems as second-stage subproblems. To tackle this problem formulation, we propose a sequential sampling-based optimization algorithm that does not require a priori knowledge of probability distribution functions or selection of samples for renewable generation. The subproblems capture the details of different energy management settings employed at the agent MGs to control heating, ventilation and air conditioning systems; home appliances; industrial production; plug-in electrical vehicles; and storage devices. Computational experiments conducted on the US western interconnect (WECC-240) data set illustrate that the proposed algorithm is scalable and our solutions are statistically verifiable. Our results also show that the proposed framework can be used as a systematic tool to gauge (a) the impact of energy management settings in efficiently utilizing renewable generation and (b) the role of flexible demands in reducing system costs. Next, we present a two-stage, multi-objective stochastic program for flow shops with sequence-dependent setups in order to meet production schedules while managing energy costs. The first stage provides optimal schedules to minimize the total completion time, while the second stage makes energy supply decisions to minimize energy costs under a time-of-use electricity pricing scheme. Power demand for production is met by on-site renewable generation, supply from the main grid, and an energy storage system. An ε-constraint algorithm integrated with an L-shaped method is proposed to analyze the problem. Sets of Pareto optimal solutions are provided for decision-makers and our results show that the energy cost of setup operations is relatively high such that it cannot be ignored. Further, using solar or wind energy can save significant energy costs with solar energy being the more viable option of the two for reducing costs. Finally, we extend the flow shop scheduling problem to a job shop environment under hour-ahead real-time electricity pricing schemes. The objectives of interest are to minimize total weighted completion time and energy costs simultaneously. Besides renewable generation, hour-ahead real-time electricity pricing is another source of uncertainty in this study as electricity prices are released to customers only hours in advance of consumption. A mathematical model is presented and an ε-constraint algorithm is used to tackle the bi-objective problem. Further, to improve computational efficiency and generate solutions in a practically acceptable amount of time, a hybrid multi-objective evolutionary algorithm based on the Non-dominated Sorting Genetic Algorithm II (NSGA-II) is developed. Five methods are developed to calculate chromosome fitness values. Computational tests show that both mathematical modeling and our proposed algorithm are comparable, while our algorithm produces solutions much quicker. Using a single method (rather than five) to generate schedules can further reduce computational time without significantly degrading solution quality.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.