Date of Award

May 2014

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Engineering

First Advisor

Vishnuteja Nanduri

Committee Members

Swarnjit S. Arora, Roshan D'Souza, Hamid Seifoddini, Wilkistar Otieno

Keywords

Carbon and Water Taxes, Climate Change, CMDP, Multi-Objective Optimization, Reinforcement Learning, SPEA

Abstract

Climate change has been shown to cause droughts (among other

catastrophic weather events) and it is shown to be exacerbated by the

increasing levels of greenhouse gas emissions on our planet. In May 2013, CO2 daily average concentration over the Pacific Ocean at Mauna Loa Observatory reached a dangerous milestone of 400 ppm, which has not been experienced in thousands of years in the earth's climate. These levels were attributed to the ever-increasing human activity over the last 5-6 decades. Electric power generators are documented by the U.S. Department of Energy to be the largest users of ground and surface water and also to be the largest emitters of carbon dioxide and other greenhouse gases. Water shortages and droughts in some parts of the U.S. and around the world are becoming a serious concern to independent system operators in wholesale electricity markets. Water shortages can cause significant challenges in electricity production having a direct socioeconomic impact on surrounding regions. Several researchers and institutes around the world have highlighted the fact that there exists an inextricable nexus between electricity, water, and climate change. However, there are no existing quantitative models that study this nexus. This dissertation aims to ll this vacuum.

This research presents a new comprehensive quantitative model that studies the electricity-water-climate change nexus. The first two parts of the dissertation focuses on investigating the impact of a joint CO2 emissions and H2O usage tax on a sample electric power network. The latter part of the dissertation presents a model that can be used to study the impact of a joint CO2 and H2O cap-and-trade program on a power grid. We adopt a competitive Markov decision process (CMDP) approach to model the dynamic daily competition in wholesale electricity markets, and solve the resulting model using a reinforcement learning approach. In the first part, we study the impacts of dierent tax mechanisms using exogenous tax rate values found in the literature. We consider the complexities of a electricity power network by using a standard direct-current optimal power flow formulation. In the second part, we use a response surface optimization approach to calculate optimal tax rates for CO2 emissions and H2O usage, and then we examine the impacts of implementing this optimal tax on a power grid. In this part, we use a multi-objective variant of the optimal power flow formulation and solve it using a strength Pareto evolutionary algorithm. We use a 30-bus IEEE power network to perform our detailed simulations and analyses. We study the impacts of implementing the tax policies under several realistic scenarios such as the integration of wind energy, stochastic nature of wind energy, integration of PV energy, water supply disruptions, adoption of water saving technologies, tax credits to generators investing in water saving technologies, and integration of Hydro power generation. The third part, presents a variation of our stochastic optimization framework to model a joint CO2 and H2O cap-and-trade program in wholesale electricity markets for future research.

Results from the research show that for the 30-bus power grid, transition from coal generation to wind power could reduce CO2 emissions by 60% and water usage about 40% over a 10-year horizon. Electricity prices increase with the adoption of water and carbon taxes; likewise, capacity disruptions also cause electricity prices to increase.

Download

Share

COinS