1,720,969 research outputs found
Sensor Location For Network Flow And Origin-Destination Estimation With Multiple Vehicle Classes
Get PDFThe need for multi-class origin-destination (O-D) estimation and link volume estimation requires multi-class observations from sensors. This dissertation has established a new sensor location model that includes: 1) multiple vehicle classes; 2) a variety of data types from different types of sensors; and 3) a focus on both link-based and O-D based flow estimation. The model seeks a solution that maximizes the overall information content from sensors, subject to a budget constraint. An efficient twophase metaheuristic algorithm is developed to solve the problem. The model is based on a set of linear equations that connect O-D flows, link flows and sensor observations. Concepts from Kalman filtering are used to define the information content from a set of sensors as the trace of the posterior covariance matrix of flow estimates, and to create a linear update mechanism for the precision matrix as new sensors are added or deleted from the solution set. Sensor location decisions are nonlinearly related to information content because the precision matrix must be inverted to construct the covariance matrix which is the basis for measuring information. The resulting model is a nonlinear knapsack problem. The two-phase search algorithm proposed addresses this nonlinear, nonseparable integer sensor location problem. A greedy phase generates an initial solution, feeding into a Tabu Search phase which swaps sensors along the budget constraint. The neighbor generation in Tabu search is a combination of a fixed swapout strategy with a guided random swap-in strategy. Extensive computational experiments have been performed on a standard test network. These tests verify the effectiveness of the problem formulation and solution algorithm. A case study on Rockland County, NY demonstrates that the sensor location method developed in this dissertation can successfully allocate sensors in realistic networks, and thus has significant practical value
Multiclass Origin-Destination Estimation Using Multiple Data Types
Get PDFEstimating O-D tables for trucks is of substantial interest due to different emission characteristics, pavement damage, etc of trucks. This thesis proposes a bilevel optimization model and corresponding solution method for static multi-class O-D estimation using various types of data. Limited memory BFGS method with bounded constraints is used for solving the upper level optimization, which is used to derive O-D table entries by minimizing the sum of squared differences between observations from different data sources and the predictions of those values. A probit model is assumed in the lower-level stochastic user equilibrium problem for flow prediction. Extensive experiments have been performed on a test network with different types of link count sensors and turning movements. The tests verify the problem formulation and solution algorithm, and offer important insights into the multiclass O-D estimation process with different types of data available
The Analysis Of The Link Between Capital Flows And Macroeconomic-Financial-Spatial Income Distribution Indicators: Combining Financial Cge And Perception Models
Get PDFThis dissertation analyzes the intricate yet critical link between macroeconomic, financial, and social variables, including spatial income distribution, and how the income inequalities are affected by certain policies and external shocks. The first chapter shows the importance of including the financial sector in today's economic and policy analyses by demonstrating the difference between the computable general equilibrium (CGE) model and its extended version that incorporates financial sector, the financial computable general equilibrium (FCGE) model. The updated FCGE model in the second chapter is then employed to analyze the increased foreign capital inflows intermediated through the banking sector, reflecting the current phenomenon in Asian emerging countries. Based on the results of simulations, some policies are proposed. The upsides and the downsides of each are analyzed in great detail in Chapter 3 by using the analytic hierarchy process (AHP) and analytic network process (ANP). Chapter 1 analyzes the difference between CGE and FCGE models, from which we conclude that serious erroneous implications and inaccuracies arise from CGE's neglect of the financial sector's role. By simulating both models in scenarios of increased government spending, depending on whether government spending is financed through taxes or government bonds, the results clearly show how the negative impacts on the social indicators generated by the CGE model can be underestimated. Examples of this underestimation are the macroeconomic impact of increased government spending and the social impact of financing the spending through taxes. I also found that the CGE results underestimated these negative impacts of increased capital flows in the same fashion. The analysis in Chapter 2 highlights the negative impact of risky financial investment behaviors of the banking sector resulting from the increased capital inflow on the economy. This chapter, in particular, stresses that one must consider not only its macroeconomic impact but also its negative repercussions on spatial income distribution and poverty conditions. Chapter 2 also shows that the risks of a boom and bust cycle where the bank-led flows are reversed from inflows to outflows and the impact of the change in banks' behavior from risk-taking to risk-averse. While riskaverse behavior can produce more favorable macroeconomic and social outcomes, there is no reason to expect that such behavior will be maintained by banks when capital inflows increase. It is therefore suggested that some measures should be taken to limit the size of bank-led flows. Chapter 3 focuses on the policy analysis based on the results of model simulations in Chapter 2. From three alternative policies - i.e., aggressive monetary policy, assigning a levy on non-core liabilities, and encouraging capital outflows - it is suggested that policymakers seriously consider imposing some sort of levy on bank- led flows. Such a conclusion is derived after taking into account the benefits, opportunities, costs, and risks of bank-led inflow based on the priority ranking of the policies, the components (criteria) and strategic goals that include macroeconomic, financial and social considerations. A series of sensitivity analyses confirm that the results are robust. In the context of the present situation in many countries, the suggested policy is part of what is known as macroprudential policy. Finally, directions for future research are suggested. The analysis in Chapter 2 could be extended upon by incorporating more financial instruments and other social indicators, or by improving the accuracy of the parameters involved in the model. For example, rather than calibrating all of the parameters, one could estimate some of the parameters by utilizing econometric equations with time series data. Furthermore, for Chapter 3, one could conduct the analysis by using direct interviewing with the same approach and model. Respondents could include experts or policymakers who would express their perceptions regarding the relations among variables in the model. In this way, the resulting priority ranking from the model simulation can be compared with, or tested against, policymakers' perceptions, from which new insights may emerge
A Study Of Electricity Planning In Thailand: An Integrated Top-Down And Bottom-Up Computable General Equilibrium (Cge) Modeling Analysis
Get PDFThis dissertation examines the potential impacts of three electricity policies on the economy of Thailand in terms of macroeconomic performance, income distribution, and unemployment rate. The three considered policies feature responses to potential disruption of imported natural gas used in electricity generation, alternative combinations (portfolios) of fuel feedstock for electricity generation, and increases in investment and local electricity consumption. The evaluation employs Computable General Equilibrium (CGE) approach with the extension of electricity generation and transmission module to simulate the counterfactual scenario for each policy. The dissertation consists of five chapters. Chapter one begins with a discussion of Thailand's economic condition and is followed by a discussion of the current state of electricity generation and consumption and current issues in power generation. The security of imported natural gas in power generation is then briefly discussed. The persistence of imported natural gas disruption has always caused trouble to the country, however, the economic consequences of this disruption have not yet been evaluated. The current portfolio of power generation and the concerns it raises are then presented. The current portfolio of power generation is heavily reliant upon natural gas and so needs to be diversified. Lastly, the anticipated increase in investment and electricity consumption as a consequence of regional integration is discussed. Chapter two introduces the CGE model, its background and limitations. Chapter three reviews relevant literature of the CGE method and its application in electricity policies. In addition, the submodule characterizing the network of electricity generation and distribution and the method of its integration with the CGE model are explained. Chapter four presents the findings of the policy simulations. The first simulation illustrates the consequences of responses to disruptions in natural gas imports. The results indicate that the induced response to a complete reduction in natural gas imports would cause RGDP to drop by almost 0.1%. The second set of simulations examines alternative portfolios of power generation. Simulation results indicate that promoting hydro power would be the most economical solution; although the associated mix of power generation would have some adverse effects on RGDP. Consequently, the second best alternative, in which domestic natural gas dominates the portfolio, is recommended. The last simulation suggests that two power plants, South Bangkok and Siam Energy, should be upgraded to cope with an expected 30% spike in power consumption due to an anticipated increase in regional trade and domestic investment. Chapter five concludes the dissertation and suggests possibilities for future research
Optimal Recovery From Disruptions In Water Distributions Networks
Get PDF: In this study, we develop a bi-level optimization model for recovery of a disrupted water distribution system. The model minimizes the total cost of recovery of the system. The cost includes the repair cost of the system and systemic impact of disruption on the system during the repair process. The systemic cost is calculated in terms of the unmet demand while the system is still damaged. The upper level problem is to schedule the repair tasks, and the lower level problem is to optimize the supply of water (optimal mitigation) given the schedule from upper level problem. The upper level problem is solved using Simulated Annealing, and the lower level problem uses a Generalized Reduced Gradient algorithm. We first apply and validate the model on a small water distribution system with only three elements damaged due to disruption. In this case, the recovery process requires tasks that can be performed in only one mode. We later apply the model to a larger and more complex water distribution system with eight elements damaged due to disruption. We perform three different experiments on this system . In the first experiment, limited resources are available at each time period. In the second experiment, the resources are increased by 50%, and in third experiment, some tasks are provided with an additional mode. The results show that the availability of resources has a significant impact on total cost of recovery of systems. Adding modes to a few tasks can help in reducing the total cost of disruption on the system
Bayesian Ranking And Selection Models For Discrete Network Design Problems With Uncertainties And Multiple Environmental Objectives
Get PDFIn this dissertation we develop a comprehensive Bayesian Ranking and Selection (R&S) modeling framework for single and multi-objective Network Design Problem with Uncertainty (NDPU). NPDU is a classical problem in transportation sciences and engineering. Due to the complex bi-level nature, NDPU is usually solved with heuristic algorithms where the objective value of many candidate solutions are "simulated" for evaluation. As the size of the transportation network can be large, the evaluation of objective values often become the computational bottleneck and should be kept to minimum numbers. On the other hand, most current formulations for (NDPU) characterize uncertainty as a discrete scenario set and tend not to fully explore the inherent correlations among alternatives. Therefore, we feel there is room for improving the efficiency of NDPU solution algorithms with a more rigorous statistical learning model. In Chapter 2, we formulate the NDPU problem as a Constrained Bayesian Ranking and Selection (R&S ) problem with exact correlated beliefs. In this formulation, each solution to the NDPU problem represents an "alternative" and the corresponding objective value represents a "reward" we want to maximize. Uncertainties in the objective values are modeled by normal distributions of the rewards and constraints of the NDPU problem are utilized for pre-eliminating infeasible solutions. At each sampling iteration, we update our belief about the distribution of all alternative performances and use the cumulative sampling history to make the next sampling decision. We use a customized version of the Knowledge Gradient policy with Correlated Beliefs (KGCB) to account for constraints and unknown variances of the rewards. Case studies are conducted on transportation networks of different sizes, using popular heuristics such as Genetic Algorithm and Simulation Annealing as comparisons. Results show that the Bayesian R&S model generally provide better accuracy and convergence rate, particularly in scenarios with uncertainty and larger networks. In Chapter 3, we build upon our model Bayesian R&S model in Chapter 2 to improve its performance under large number of projects/alternatives. The new model features 1) a recursively updated linear approximation of the upper-level objective function using Gaussian-binary basis functions, and 2) A surrogateassisted knowledge gradient sampling policy which utilizes the optimal solution of the approximated surrogate objective function to constraint the scale of the expensive knowledge gradient calculation. With the two features the computational complexity of our algorithm is reduced to only a low degree (typically [LESS-THAN OR EQUAL TO] 2) polynomial of the number of projects. Case studies are conducted on the Sioux Fall network and Anaheim network with as many as 20 projects and over 1,000,000 possible network configurations. Results showed that this parametric Bayesian R&S model is able to identify highly optimal solutions in only around 100 iterations, significantly outpacing our bench-marking Genetic Algorithm and Simulated Annealing Algorithm in both convergence speed and computational cost. Our new method provides a highly scalable framework for discrete NDPU without sacrificing much of the performance advantage of Bayesian R&S models. It also extends the Bayesian R&S model and the knowledge gradient sampling policies to generic large-scale discrete optimization problems, which provides valuable insights for a large class of similar optimization and learning problems. In Chapter 4, we further extend the Bayesian R&S model to the MultiObjective discrete Network Design Problem with Uncertainty (MONDPU), an emerging area in transportation planning due to the need for sustainable transportation systems. In this formulation, we put independent parametric beliefs on the expected reward of each objective function like we did in Chapter 3 and update them in parallel through sequential samples. We define a multi-objective version of the Knowledge Gradient policy with Correlated Beliefs which use a crowding distance metric to ensure the diversity of the Pareto optimal front. Case studies are conducted on the Sioux Fall network and Anaheim network. Results showed that our multi-objective Bayesian R&S model is able to identify a very diverse set of highly optimal solutions under very limited budget, significantly out-performing the bench-marking NSGA-II algorithm in both solution quality and practicality. Our model is also the first to extend the Bayesian R&S model and the knowledge gradient sampling policies to generic multi-objective problems. In summary, the Bayesian R&S formulation is well-suited for NDPU and MONDPU due to its uncertainty management capabilities and the sampling efficiency of knowledge-gradient related policies. The models provide an innovative statistical learning perspective to NDPU, which has mainly been studied as an optimization problem. The new formulation is intuitive to understand and easily applicable to similar discrete optimization problems such as the Optimal Sensor Location problem, Uncapcitated Fixed Charge Facility Location problem, etc. The global Bayesian belief structure and the sequential valueof-information sampling policies make the model especially efficient for black- box, gradient free optimization problems where the evaluation of each objective value take up the majority of the computational burden. We believe the models themselves as well as this unique statistical perspective is of great interest and value for transportation network modelers and simulation optimization practitioners
Parallel Real Asset Management With Environmental Regulation: Integer Programming And Approximate Dynamic Programming Approaches
Get PDFThis dissertation presents a pair of models designed to assist in the management of multiple deteriorating real assets, given financial and environmental concerns. Whether the assets are buildings or vehicles or machines, their purchase and upkeep can be costly, making optimal management policies valuable. The models presented build upon a strong literature. They incorporate numerous factors which have been modeled previously, though generally not together. These include technological change, linked decisions for multiple assets, and non-steady-state demand. They stand out from previous literature due to their ability to model retrofits, as well as repairs and replacements. These retrofits can have initial as well as ongoing costs, and can impact externalities, making them relatively general. The integer program model is fast and well suited to analysis requiring large numbers of runs, such as the comparison of a wide range of regulatory alternatives. The approximate dynamic program, while slower, is able to handle stochastic asset failures and repair costs for large asset portfolios, something which previous models have struggled to accomplish without strong simplifying assumptions. A customized value iteration approach produces good solutions within a few hours for sample problems involving a fleet of well over a thousand vehicles subject to clean diesel regulation
A Macroeconometric Model For Thailand With Welfare Linkages: Analyses Of Selected Policies In Response To Energy Price Increases
Get PDFThis dissertation uses the data from 1993 to 2004 to demonstrate the effects of the increase in petroleum price on the general economy of Thailand and their linkage with aspects of welfare. It also evaluates the effectiveness of the counterfactual policy responses. The evaluation is done using a macroeconometric model in which the data are first incorporated into a system of simultaneous equations. Then a policy or a combination of policies is simulated for each scenario. Charts of the results of the variables simulated in each scenario are analyzed. There are seven chapters in this dissertation. Chapter One begins with an introduction that discusses why an increase in the price of crude oil concerns not only an individual country but also the world as a whole, and then turns to the counterfactual policies proposed by this dissertation. The important scenarios of price surge in the past and the recent price situation are briefly discussed. Because the recent cases of price increase have different causes from those of the past, the shifts in the demand and the supply curves of both cost-push and demand-pull inflation as well as the dynamic movements of aggregate demand and supply are explained. This discussion is followed by an explanation of how the mechanism of an increase in the price of oil can lead to a change in welfare. Thailand's recent economic situation, including the years before, during, and after the era of the Asian financial crisis (1985-2006) is presented. The chapter also briefly discusses the data of the nation's gross domestic product, consumer price index (CPI), and current account balance during the 1970s price surge. (This brief discussion reflects a suspicion that changes in these variables had been affected by changes in crude oil prices.) The chapter concludes with the discussions of Thailand's oil price structure and its oil fund program, an influential tool that the government has been using as an immediate policy response. Chapter Two surveys the literature in two main categories: The first category is the macroeconometric models of various countries, and the second is the impact of changes in oil prices as determined by different modeling methods. Chapter Three explains the details of the macroeconometric framework that serves as the core model of the dissertation. This chapter also discusses the construction and the history of the core model, which is based on a supply and demand concept, as well as the advantages and disadvantages of the model. Chapter Four presents the general economy block that is composed of the blocks for aggregate demand (C,I), trade (X, M), production (total output), and price (PGDP and CPI), all of which served as a core model. Chapter Five presents the energy, fiscal, and welfare blocks. Chapters Four and Five together explain the fundamental theories in building the overall structure and also present each dependent variable as a function of other variables. Finally, the relationships among the variables within a system or a block are demonstrated by a flowchart. With a predefined set of explanatory variables for each dependent variable, each equation and ex-post simulation was calibrated using EVIEWS 6.0. The results of the coefficients, the fitted graphs, and the mean absolute percentage error (MAPE) are shown in Appendix C. The results of the baseline simulations, which are the attempts to match the model with actual data, can be found in Appendix D. Chapter Six presents the results of the seven simulated scenarios. These scenarios include the impact of the world oil price increase (Scenario 1); the use of the oil fund as a counterfactual policy response when the world crude oil price increased by 50% (Scenario 2); the use of the oil fund when the world crude oil price increased by 200% (Scenario 3); the use of a tax reduction when the world crude oil price increased by 50% (Scenario 4); the impact of a reduction in the sales of automobiles (Scenario 5); the impact of the monetary policy response (Scenario 6); and the impact of the fiscal policy response in addition to the monetary policy response (Scenario 7). The results show that the world crude oil price increase is followed by a decline in almost every variable, among which investment presents the greatest decline. When the price of the world crude oil increases by 50%, a 1.2 bath/liter subsidy from the oil fund or a 35% tax cut is needed to stabilize the economy. However, when the crude oil price increases by 200%, a 2.5 bath/liter subsidy from the oil fund is needed. A reduction in automobile sales shows only a few percent reduction in the usage of diesel as well as a very small reduction in the total number of automobiles. Finally, a rising interest rate in response to the rising price level indeed worsens the overall economy, and increasing government expenditures significantly helps only some variables such as unemployment. Chapter Seven concludes the dissertation
Efficient Design Of Inbound Logistics Networks
Get PDFLogistics is a vitally important part of the economy, and it is now a $1.45 trillion industry in the United States representing 8.3 percent of GDP. Efficient design of routes and schedules for moving materials into manufacturing or assembly plants is a central part of inbound logistics operations. This dissertation builds on elements of traditional vehicle routing as well as broader elements of logistics planning. At the core of the process is a mathematical optimization termed capacitated clustering. Two major categories of suppliers are analyzed in this research. The first supplier category includes suppliers with small quantities of materials, so daily pickups may not be required. A new approach is proposed that considers pick-up frequency and spatial design as joint decisions to minimize total logistics (transportation plus inventory) cost. The clustering-based optimization uses an approximation to the actual cost of a routing solution without actual route construction. The problem is shown to be analogous to a single-source fixed-charge facility location problem, and near-optimal solutions can be found using an efficient heuristic algorithm. Computational experiments show the effectiveness of how this model is formulated and a case study demonstrates that substantial total cost savings can be achieved in realistic applications. A second category of suppliers ships moderately large volumes to a single plant but not enough to fill a truck themselves. One commonly used process is to have plant- based collection routes on a daily basis that stop at multiple suppliers and return to the plant. The model developed here is formulated as a two-stage stochastic program, which includes uncertainty in the load quantities at suppliers and controls (either penalties or constraints) designed to improve the "regularity" of service to individual suppliers. Two adaptive decomposition heuristics are explored for solving the stochastic program in large scale, integer L-shaped method (ILSM) and progressive hedging (PH). An application to logistics operations in the automotive industry is used to demonstrate the effectiveness of the model and the PH solution method
Bilateral Interactions In Two-Sided Networks – A Perspective From Matching Theory
No full textTraditional methods such as gravity models and general equilibrium theory for two-sided network analysis focus mainly on characterizing the aggregate and macro-level outcomes of twosided interactions that commonly occur in a wide range of applications such as trade market, transportation, and migration. As customer-oriented service and human-centered design become more feasible in the information age, theories and models that capture and represent individual behaviors are crucial and essential for the studies on two-sided networks, i.e. in understanding the observations of two-sided interactions, forecasting future activities, and designing policies, platforms, markets and mechanisms to achieve desirable outcomes. For example, in international trade analysis, we need advanced agent-based theories and models to explain two-sided phenomena observed in trade, forecast trade levels, and design rules and platforms to promote fair and efficient market operations. Matching theory, one of the most exciting intellectual endeavors of human minds, promises, in the author's view, suitable methodologies and powerful analytical tools for the study of how the agents in a network or market make decisions and interact, hence how to formulate matching mechanisms for desirable outcomes. This dissertation aims to contribute to the matching literature by proposing and studying generalized matching, which expands the existing matching theory to multi-unit many-to-many matching with quota constraints. This is a more general and realistic framework for matching that happens in real world. First, models for two-sided and one-sided matching with newly defined preference relationships and solution concepts are developed to pave the theoretical foundation for analyzing multi-unit and multipartner matching with quota constraints. The corresponding new matching mechanisms are then designed to produce stable and favorable matching outcomes. Second, a hybrid model for generalized matching is established to encompass both one-sided and two-sided matching under the generalized framework. Again, the corresponding hybrid matching mechanism with desired properties is proposed and discussed. Next, linking the newly proposed theoretical work to empirical application, a novel bi-level estimation model is proposed for generalized matching to make inferences of agents' matching behaviors/decisions. Last but not least, the dissertation also points out the remaining challenges and offers opinions on directions and topics for future research
- …
