Informs Annual Meeting 2017

SD64

INFORMS Houston – 2017

SD63

SD64

370D Energy and Climate 4 Invited: Energy and Climate Invited Session Chair: Peter Larsen, Lawrence Berkeley National Laboratory, PHLarsen@lbl.gov 1 - Sequencing Distributed and Central Power Supply in Systems with Low Electricity Access and Low Reliability: An Application of the GAP Model Juan Pablo Carvallo, University of California-Berkeley, CA, United States, jpcarvallo@lbl.gov, Daniel Kammen, Duncan Callaway Over 1.1 billion people do not have access to electricity, 600 million in Africa alone. We develop a novel and spatially explicit approach to jointly assess the sequencing and pacing of centralized, distributed, and off-grid electrification strategies. We develop and employ the linear program and network optimizer Grid Access and Planning (GAP) model. Contrary to existing results, we find that optimal electrification strategies involve co-existence of grid and off-grid technologies. We assess the impact of several variables in defining the strategy and suggest policy frameworks for their implementation. 2 - Impact of High Variable Renewable Energy Penetrations on Demand and Supply Side Decisions Andrew Mills, Lawrence Berkeley National Laboratory, CA, United States, admills@lbl.gov Many long-lasting decisions for supply- and demand side electricity infrastructure and programs are made based on historical observations or assuming a business- as-usual (BAU) future. As the share of variable renewable energy (VRE) increases, however, fundamental characteristics of the power system will change. Decisions related to the supply of or demand for electricity made based on an assumed BAU future may not achieve their intended objective in a high VRE future. We highlight how the potential shift to high VRE futures can impact wholesale prices in ways that should be considered in decisions related to long- lasting supply- and demand-side electricity infrastructure and programs. 3 - Scenarios with and with out Overshoot for End of Century Climate Goals Felipe A. Feijoo, Pacific Northwest National Laboratory, 5825 University Research Court, Suite 3500, College Park, MD, 20740, United States, felipe.feijoo@pnnl.gov Over 3000 scenarios for global temperature over the 21st century were generated by a single integrated assessment model (GCAM) to explore pathways that either do not exceed a global average temperature target or reach that target in 2100. All scenarios reached above 1.8 C in 2050. Temperature pathways from present to 2045 were found to be less sensitive to climate policy assumptions (carbon price in GCAM) than uncertainty in Earth system modeling. Following 2045 pathway diverge to reach in 2100 from just 0.75 C to 3.5 C (reference case) in 2100 (1.5 to 3.5 C when CCS is excluded). 4 - Quantifying the Contribution of Water and Environmental Management to the Economies of Developing Countries Brent Boehlert, Industrial Economics, Inc., Cambridge, MA, United States, bboehlert@indecon.com Brent Boehlert, Massachusetts Institute of Technology, Cambridge, MA, United States, bboehlert@indecon.com, Kenneth M. Strzepek, Kenneth M. Strzepek, James Thurlow Most developing countries face increasing challenges to management of water and environmental resources. To ensure these resources are sustainably managed, it is important to clearly articulate their contribution to economic growth and development. We present a modeling framework that evaluates the macroeconomic implications of investments in water and environmental resources through a series of impact channels that link biophysical model outputs to an economy-wide model. For illustration, we focus on the case of the Ugandan economy, and find that enhanced investment in management of these resources can increase per capita GDP by 9% in 2040.

370E Joint session DM/QSR: Process Monitoring for Diverse Types of Data Sponsored: Data Mining Sponsored Session Chair: Youngseon Jeong, China, youngseonjeong@gmail.com Co-Chair: Behnam Tavakkol, Rutgers University, 5200 BPO Way, Piscataway, NJ, 08854, United States, btavakkol66@gmail.com 1 - Detection Method of Abnormal Condition in Fabrication Process using Wafer Gradient Vector Based on Semiconductor EDS Chip Data Dongkyu Jeon, Samsung Electronics, 1, Samsungjeonja-ro, Hwaseong-si, 18448, Korea, Republic of, jdkclub85@gmail.com, BokYoung Kang, ByongHun Cho, Jung-Chan Ahn, ChiYong An, Kil Soo Kim, Seung Hoon Tong In general, abnormal condition detection in a Fabrication process uses wafer-level representative statistics summarizing EDS (Electrical Die Sorting) data that exist in each semiconductor chip unit. However, this conventional method has a limitation in that the numerical discrimination power of the wafer defects in a specific area decreases in the statistical summarization process. In this research, we propose a methodology to detect abnormal condition in fabrication process through Density based Clustering Algorithm using Wafer Gradient Vector based on EDS chip data. In addition, we verified the proposed algorithm through real- life cases. 2 - Online Bayesian Learning Process Monitoring and Control Sangahn Kim, Rutgers University, 1050 George St. Apt. 6B, New Brunswick, NJ, 08901, United States, sk1389@scarletmail.rutgers.edu, Myong Kee Jeong, Elsayed Elsayed Conventional statistical process control and monitoring methodologies in Phase II commonly assume that the parameters obtained in Phase I are fixed over time and sufficient to test whether the process is normal or out of control. However, as new observations become available in Phase II, they are incorporated in monitoring as prior knowledge through a Bayesian framework and the parameters obtained in Phase I are continuously updated. In this presentation, we introduce a new process control scheme in which we update the prior distribution at every sampling point. A Bayesian hierarchical model enables us to apply the proposed method in monitoring of high-dimensional processes with sparse mean shifts. 3 - Clustering Validity Indices for Uncertain Data Objects Behnam Tavakkol, Rutgers University, 5200 BPO.Way, Piscataway, NJ, 08854, United States, btavakkol66@gmail.com, Myong Kee Jeong, Susan L.Albin In cluster analysis, clustering validity indices are the main tools of evaluating the quality of the formed clusters. In this study, we develop a few clustering validity indices for a type of data objects known as uncertain data. Uncertain data objects can be considered as probability density functions or samples of observations. To our best knowledge, in the existing literature, there is not any clustering validity index for uncertain data. 4 - A Low-Rank Multivariate General Linear Model for Multi-Subject fMRI Data and a Non-Convex Optimization Algorithm for Brain Response Comparison Minh Pham, University of Virginia, 1555 Montessori Terrace, Charlottesville, VA, 22911, United States, mtp9b@eservices.virginia.edu, Tingting Zhang The focus of this talk is on identifying brain regions with different responses using stimulus-evoked functional magnetic resonance imaging (fMRI) data. To jointly model many brain voxels’ responses to designed stimuli, we present a new low- rank multivariate general linear model for stimulus-evoked fMRI data. We show that the proposed method can outperform several existing voxel-wise methods by achieving both high sensitivity and specificity. We apply the proposed method to the fMRI data and identify anterior dACC to have different responses to a designed threat and control stimuli. 5 - Group Classification Methods for SAR Imagery Data Mariya Naumova, Rutgers University, 640 Bartholomes Road, Piscataway, NJ, 08854, United States, mnaumova@rci.rutgers.edu Given a finite number of learning samples from several populations (groups) and a collection of samples from the union of these populations, we classify the entire collection (not a single sample) to one of the groups. Such problems often arise in medical, chemical, biological and technical diagnostics, classification of signals, etc. We consider different parametric statistical methods of solving the problem and make comparison of their quality based on numerical results. We give an illustrative example with SAR imagery data to demonstrate the effectiveness of the classification methods.

123