Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
7 results
Search Results
Item OPTIMAL SCHEDULING OF RESIDENTIAL DEMAND RESPONSE USING DYNAMIC PROGRAMMING(2019) Moglen, Rachel Lee; Gabriel, Steven A; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Electricity price volatility in the Electricity Reliability Council of Texas (ERCOT) poses significant financial threat to many players in the electricity market, though most of the financial burden falls on the retail electric providers (REPs). REPs are contractually obligated to purchase and provide all electricity that the end-user wishes to consume, even when the purchase of this electricity brings them financial losses. Electricity prices in ERCOT can increase from typical ranges ($30/MWh) to $3,000/MWh in as little as 15 minutes, causing REPs to seek mitigation techniques to avoid paying price spike prices for electricity. We explore two techniques for REPs to schedule mitigation techniques: a price prediction-based heuristic approach, as well as an optimal scheduling algorithm using dynamic programming (DP). We aim to optimally schedule these mitigation techniques which shift load from high price periods to times of lower prices, called demand response (DR). To achieve this load shifting, REPs remotely manipulate internet-connected thermostats of residential customers, thereby controlling a fraction of residential HVAC load. We found that the price prediction approach was highly unreliable, even for predicting prices as near as 5 minutes out. We therefore chose to rely on the DP as the primary scheduling model. By applying the DP deterministically to historical electricity price and weather data, the load-shifting technique is shown to potentially improve REP profit margins by 10% to 25% per customer annually. Most of these savings come from a few crucial events, highlighting the usefulness of the DP and the importance of accuracy in the timing of DR events. Due to the uncertainty in electricity prices, we apply a multi-objective approach considering the REP’s conflicting objectives: maximizing savings and minimizing financial risk. Results from this multi-objective formulation point to shorter duration DR events in the evening being the least risky, with additional savings possible through riskier short midday events. To ensure that REPs could apply our DP formulation for use in near real-time decision-making applications, the computation speed was verified to be under one second for 24 stages (i.e., 1-hour intervals for one day.)Item Decision Making Under Uncertainty: New Models and Applications(2018) Jie, Cheng; Fu, Michael C; Applied Mathematics and Scientific Computation; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In the settings of decision-making-under-uncertainty problems, an agent takes an action on the environment and obtains a non-deterministic outcome. Such problem settings arise in various applied research fields such as financial engineering, business analytics and speech recognition. The goal of the research is to design an automated algorithm for an agent to follow in order to find an optimal action according to his/her preferences.Typically, the criterion for selecting an optimal action/policy is a performance measure, determined jointly by the agent's preference and the random mechanism of the agent's surrounding environment. The random mechanism is reflected through a random variable of the outcomes attained by a given action, and the agent's preference is captured by a transformation on the potential outcomes from the set of possible actions. Many decision-making-under-uncertainty problems formulate the performance measure objective function and develop optimization schemes on that objective function. Although the idea on the high-level seems straightforward, there are many challenges, both conceptually and computationally, that arise in the process of finding the optimal action. The thesis studies a special class of performance measure defined based on Cumulative Prospect Theory (CPT), which has been used as an alternative to expected-utility based performance measure for evaluating human-centric systems. The first part of the thesis designs a simulation-based optimization framework on the CPT-based performance measure. The framework includes a sample-based estimator for the CPT-value and stochastic approximation algorithms for searching the optimal action/policy. We prove that, under reasonable assumptions, the CPT-value estimator is asymptotically consistent and our optimization algorithms are asymptotically converging to the optimal point. The second part of the thesis introduces an abstract dynamic programming framework whose transitional measure is defined through the CPT-value. We also provide sufficient conditions under which the CPT-driven dynamic programming would attain a unique optimal solution. Empirical experiments presented in the last part of thesis illustrate that the CPT-estimator is consistent and that the CPT-based performance measure may lead to an optimal policy very different from those obtained using traditional expected utility.Item Stochastic Systems with Cumulative Prospect Theory(2013) Lin, Kun; Marcus, Steven I.; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Stochastic control problems arise in many fields. Traditionally, the most widely used class of performance criteria in stochastic control problems is risk-neutral. More recent attempts at introducing risk-sensitivity into stochastic control problems include the application of utility functions. The decision theory community has long debated the merits of using expected utility for modeling human behaviors, as exemplified by the Allais paradox. Substantiated by strong experimental evidence, Cumulative Prospect Theory (CPT) based performance measures have been proposed as alternatives to expected utility based performance measures for evaluating human-centric systems. Our goal is to study stochastic control problems using performance measures derived from the cumulative prospect theory. The first part of this thesis solves the problem of evaluating Markov decision processes (MDPs) using CPT-based performance measures. A well-known method of solving MDPs is dynamic programming, which has traditionally been applied with an expected utility criterion. When the performance measure is CPT-inspired, several complications arise. Firstly, when solving a problem via dynamic programming, it is important that the performance criterion has a recursive structure, which is not true for all CPT-based criteria. Secondly, we need to prove the traditional optimality criteria for the updated problems (i.e., MDPs with CPT-based performance criteria). The theorems stated in this part of the thesis answer the question: what are the conditions required on a CPT-inspired criterion such that the corresponding MDP is solvable via dynamic programming? The second part of this thesis deals with stochastic global optimization problems. Using ideas from the cumulative prospect theory, we are able to introduce a novel model-based randomized optimization algorithm: Cumulative Weighting Optimization (CWO). The key contributions of our research are: 1) proving the convergence of the algorithm to an optimal solution given a mild assumption on the initial condition; 2) showing that the well-known cross-entropy optimization algorithm is a special case of CWO-based algorithms. To the best knowledge of the author, there is no previous convergence proof for the cross-entropy method. In practice, numerical experiments have demonstrated that a CWO-based algorithm can find a better solution than the cross-entropy method. Finally, in the future, we would like to apply some of the ideas from cumulative prospect theory to games. In this thesis, we present a numerical example where cumulative prospect theory has an unexpected effect on the equilibrium points of the classic prisoner's dilemma game.Item Sequential Search With Ordinal Ranks and Cardinal Values: An Infinite Discounted Secretary Problem(2009) Palley, Asa Benjamin; Cramton, Peter; Applied Mathematics and Scientific Computation; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)We consider an extension of the classical secretary problem where a decision maker observes only the relative ranks of a sequence of up to N applicants, whose true values are i.i.d. U[0,1] random variables. Applicants arrive according to a homogeneous Poisson Process, and the decision maker seeks to maximize the expected time-discounted value of the applicant who she ultimately selects. This provides a straightforward and natural objective while retaining the structure of limited information based on relative ranks. We derive the optimal policy in the sequential search, and show that the solution converges as N goes to infinity. We compare these results with a closely related full information problem in order to quantify these informational limitations.Item Internalizing Production Externalities: A Structural Estimation of Real Options in the Upstream Oil and Gas Industry(2009) Muehlenbachs, Lucija; Nerlove, Marc; Rust, John; Agricultural and Resource Economics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)There are hundreds of thousands of crude oil and natural gas wells across North America that are currently not producing oil or gas. Many of these wells have not been permanently decommissioned to meet environmental standards for permanent closure, but are in an inactive state that enables them to be more easily reactivated. Some of these wells have been in this inactive state for more than sixty years which begs the question of whether they will ever contribute to our energy supply, or whether they are being left inactive because the environmental remediation costs are prohibitively high. I estimate a structural model of optimal well operations over time and under uncertainty to determine what conditions or policies might push any of the inactive wells out of the hysteresis in which they reside. The model is further used to forecast production from existing wells and recoverable reserves from existing pools. The estimation uses data on production decisions from 84 thousand conventional oil and gas wells and estimates of the remaining reserves of 47 thousand pools. As the producer's decision depends on their subjective belief for how prices and recoverable reserves change over time, I also estimate the probability of changes in prices and recovery technology. I model increases and decreases in the estimated recoverable reserves to depend on price, and predict that natural gas reserves are more responsive to changes in price than conventional oil reserves. Under high prices there is potential for large increases in gas reserves, however this is not the case for oil reserves when the oil price is high. And likewise, under low prices, gas reserves decrease more than oil reserves. The dynamic programming model predicts that with only a drastic, arguably implausible, increase in prices and recovery rates will there be a significant increase in the number of inactive wells that are reactivated. If ideal conditions are not enough to induce well reactivation then this implies that typically wells are left inactive not because of the option to reactivate, but rather because the cost of environmental cleanup is too high. Should there be externalities from idling the wells (such as continued contamination of groundwater) that are not accounted for in the decision, then this behavior may not be socially optimal. The model predicts that a Pigouvian tax on inactive wells would have the added benefit of inciting the reactivation of oil and gas wells, however in the case of oil, a tax would incite more wells to be decommissioned than reactivated.Item ESSAYS IN EMPIRICAL INDUSTRIAL ORGANIZATION(2009) Chesnes, Matthew William; Rust, John; Jin, Ginger; Economics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Chapter 1: Capacity and Utilization Choice in the US Oil Refining Industry This paper presents a new dynamic model of the operating and investment decisions of US oil refiners. The model enables me to predict how shocks to crude oil prices and refinery shutdowns (e.g., in response to hurricanes) affect the price of gasoline, refinery profits, and overall welfare. There have been no new refineries built in the last 32 years, and although existing refineries have expanded their capacity by almost 13% since 1995, the demand for refinery products has grown even faster. As a result, capacity utilization rates are now near their maximum sustainable levels, and when combined with record high crude oil prices, this creates a volatile environment for energy markets. Shocks to the price of crude oil and even minor disruptions to refining capacity can have a large effect on the downstream prices of refined products. Due to the extraordinary dependence by other industries on petroleum products, this can have a large effect on the US economy as a whole. I use the generalized method of moments to estimate a dynamic model of capacity and utilization choice by oil refiners. Plants make short-run utilization rate choices to maximize their expected discounted profits and may make costly long-term investments in capacity to meet the growing demand and reduce the potential for breaking down. I show that the model fits the data well, in both in-sample and out-of-sample predictive tests, and I use the model to conduct a number of counterfactual experiments. My model predicts that a 20% increase in the price of crude oil is only partially passed on to consumers, resulting in higher gasoline prices, lower profits for the refinery, and a 45% decrease in total welfare. A disruption to refining capacity, such as the one caused by Hurricane Katrina in 2005, raises gasoline prices by almost 16% and has a small negative effect on overall welfare: the higher profits of refineries partially offsets the large reduction in consumer surplus. As the theory predicts, these shocks have a smaller effect on downstream prices when consumer demand is more elastic, resulting in a larger share of total welfare going to the consumer. Chapter 2: Consumer Search for Online Drug Information Consumers are increasingly turning to the internet and using search engines to find information on medicinal drugs. Between 2001 and 2007, the number of adults using the internet as an alternative source of health information doubled. At the same time, online and offline advertising spending by drug companies is growing rapidly. I seek to understand how consumers use search engines to find drug information and how this activity is influenced by direct to consumer advertising. I utilize a database of user click-through data from America Online to analyze the search behavior of consumers seeking drug information online. Compared with other searches, users submitting drug-related queries are more likely to click on more than one result in a search session, and when they do, they click more rapidly through the results and tend to migrate away from dot-com sites and toward those ending in dot-org and dot-net. Offline advertising on a drug serves to increase the frequency and intensity of these searches. Chapter 3: Drug Information via Online Search Engines This paper utilizes a database of organic and sponsored search results from four large search engines to analyze the supply of drug-related information available on the internet. I show that the information varies significantly across search engines, domain extensions, and between organic and sponsored results. Regression results reveal that websites with relatively more promotional content are pushed down in the search results while informational sites (including those ending in dot-gov and dot-org) are more likely to appear on page one of the results.Item Applications of Genetic Algorithms, Dynamic Programming, and Linear Programming to Combinatorial Optimization Problems(2008-10-16) Wang, Xia; Golden, Bruce L.; Mathematics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Combinatorial optimization problems are important in operations research and computer science. They include specific, well-known problems such as the bin packing problem, sequencing and scheduling problems, and location and network design problems. Each of these problems has a wide variety of real-world applications. In addition, most of these problems are inherently difficult to solve, as they are NP-hard. No polynomial-time algorithm currently exists for solving them to optimality. Therefore, we are interested in developing high-quality heuristics that find near-optimal solutions in a reasonable amount of computing time. In this dissertation, we focus on applications of genetic algorithms, dynamic programming, and linear programming to combinatorial optimization problems. We apply a genetic algorithm to solve the generalized orienteering problem. We use a combination of genetic algorithms and linear program to solve the concave cost supply scheduling problem, the controlled tabular adjustment problem, and the two-stage transportation problem. Our heuristics are simple in structure and produce high-quality solutions in a reasonable amount of computing time. Finally, we apply a dynamic programming-based heuristic to solve the shortest pickup planning tour problem with time windows.