Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MD45

2 - Offline Framework to Optimize Current Energy Converter Simulation with Multiple Correlated Functional Response Sterling Stewart Olson, Sandia National Labs, Albuquerque, NM, 87185, United States, Chris Chartrand, Jesse Roberts, Humberto Silva, Cheping ‘Jack’ Su Current energy converters are an important component of sustainable energy. However experimental measurements are prohibitively expensive making modeling a critical component to increasing the technology readiness level of devices. This research sought to optimize a Gaussian process regression meta- model relating four turbulence parameter inputs to two outputs (velocity and turbulent intensity) related by partial differential equations as a function of distance for a long wall clock simulation. The meta-model optimized parameters showed acceptable results for a ten-fold decrease in total simulation time and provides many opportunities for future research. 3 - Rolling Horizon Optimization for a H2Electrolyzer Station Used to Service H2 Vehicles Ross Guttromson, PhD, Sandia National Labs, Albuquerque, NM, 87185, United States The integration of hydrogen vehicles into the US transportation system relies upon coordinated access to various energy supply systems. The diversity of these energy supplies helps to manage cost and resilience, but will also add complexity with regards to economic and other operating decisions. A multi-period, rolling- horizon optimization was developed to provide reliable and sound economic, real-time decisions for a hydrogen electrolyzer system used to fulfil potential H2 vehicle demand. The optimization considers decisions associated with the purchase of wind and solar renewable energy, storage of H2, the purchase of electricity on a day-ahead basis, the sale of excess electricity on a real-time market, fulfillment of station service electrical demand, and the sale of a zero-net- energy regulation product, similar to the PJM RegD market. This model was also used to economically size the infrastructure based on demand and price forecasts. 4 - Design of Community-Focused Resilience Metrics for Effective Infrastructure Investment Robert Fredric Jeffers, PhD, Sandia National Labs, Albuquerque, NM, 87185, United States Abstract not Available 5 - Power System Resiliency Reliability Cooptimization Bryan Arguello, R&D Computer Science, Sandia National Labs, 1515 Eubank Blvd. SE, NM, Albuquerque, NM, 87123, United States, Brian Pierre, Shabbir Ahmed, Emma Johnson The purpose of our work is to co-optimize power system reliability and resiliency investment decisions. Our working definition of resiliency is resistance to load shed during low-probability, high-impact events. Similarly, we define reliability as resistance to load shed during high-probability, low-impact events. We have developed both a distribution-level reliability investment model driven by historical outage data, and a transmission-level stochastic resiliency investment model driven by extreme weather event scenarios. Through our co-optimization technique, we explore the tradeoff between transmission resiliency and distribution reliability when the models share a budget. Joint Session ENRE/Practice Curated: Optimization and Machine Learning for Electric Power Systems Sponsored: Energy, Natural Res & the Environment/Electricity Sponsored Session Chair: Yu Zhang, PhD, UC Santa Cruz, 1156 High St SOE2, Santa Cruz, CA, 95064, United States 1 - Application of Machine Learning to Distribution Synchrophasors Alireza Shahsavari, University of California, Riverside, Riverside, CA, United States, Hamed Mohsenian-Rad The recent development of distribution-level phasor measurement units, a.k.a. micro-PMUs, has been an important step towards achieving situational awareness in power distribution networks. A typical micro-PMU is connected to single- or three-phase distribution circuits to measure GPS time-referenced magnitudes and phase angles of voltage and current phasors at 120 readings per second. The challenge in using micro-PMUs is to transform the large amount of data that is it generates to actionable information and match the said information to use cases with practical value to distribution system operators. In this talk, we adopt a big- data approach to address this open problem. We introduce a novel data-driven event detection technique to pick out valuable portion of data from extremely large raw micro-PMU dataset. We then develop a data-driven event classifier to effectively classify power quality events. Importantly, we use field expert knowledge and utility records to conduct an extensive data-driven event labeling. Certain aspects from event detection analysis are adopted as additional features to be fed into the classifier model. n MD44 North Bldg 227C

2 - Joint Structure and Parameter Estimation in Power Distribution Under Limited Observability Deepjyoti Deka, Los Alamos National Lab Efficient operation of distribution grids in the smart-grid era is hindered by the limited presence of real-time meters. This paper studies the problems of topology and parameter estimation in the limited observability regime where measurements are restricted to only the terminal nodes of the grid and all intermediate nodes are unobserved/hidden. To this end, we propose two algorithms for exact topology (and impedances) estimation. We discuss the computational and sample complexity of our proposed algorithms and demonstrate that topology (and impedance) estimation by our algorithms are Conventional power system unit commitment problems are deterministic and assume piece wise constant behavior of net-load, while in reality net-load is uncertain and inter-period generator ramps are not well defined. We will discuss a stochastic continuous time problem formulation, where we try to give a more realistic generator ramp trajectories, as well as incorporating a multi-stage stochastic decision framework, where dispatch decisions are revised over the modeling horizon. 4 - Wind Power Forecasting via Deep Neural Networks Yu Zhang, University of California, Santa Cruz, Jiao Hao Miao Accurate forecasting of renewable generation is a challenging task due to its inherent intermittency and volatility. In this context, this talk deals with a new machine learning approach for predicting wind power generation with various meteorological factors including wind speed, direction, humidity, etc. We first utilize data visualization techniques for the feature selection, and then develop a deep neural network to predict the wind power outputs. Numerical results show that our proposed approach outperforms existing methods including persistence, support vector regression and ARMA. Equilibria in Low-Carbon Electricity Markets Sponsored: Energy, Natural Res & the Environment/Electricity Sponsored Session Chair: Rodrigo Moreno, University of Chile, Santiago, Chile Co-Chair: Francisco Munoz 1 - Investigating the Role and Value of Flexible Technologies in Low Carbon Deregulated Electricity Markets Dimitrios Papadaskalopoulos, Imperial College-London, London, United Kingdom The challenges associated with the decarbonization and the deregulation of electricity markets have attracted significant interest around the role and value of new flexible technologies such as flexible demand and energy storage. The author will present a new multi-period equilibrium programming model of the electricity market, capable of capturing for the first time both the strategic behavior of independent market players and the complex time-coupling characteristics of new flexible technologies. Results from case studies on a UK-based test market will demonstrate the impacts of these technologies on the business case of different market players and the overall efficiency of the market. 2 - Equilibrium Analysis of a Tax on Carbon Emissions with Pass-through Restrictions and Side-payment Rules Francisco David Munoz, Universidad Adolfo Ibáñez, Santiago, Chile, Gabriel D az, Rodrigo Moreno We develop an equilibrium model with endogenous investments in generation capacity to quantify the economic inefficiencies of an emissions policy with pass- through restrictions and side-payment rules. We provide general results that show that social costs and long-term electricity prices are always lower under a standard carbon tax than under an emissions policy with such restrictions. We explore the effects of the policy through numerical experiments and find that its effect on carbon emissions is ambiguous. Tax revenues can be much higher under such emissions policy than under a standard carbon tax, which gives the regulator weak incentives to modify the regulation. optimal with respect to number of nodal observability required. 3 - Stochastic Continuous Time Unit Commitment Anna Scaglione, Arizona State University, Tempe, AZ, United States, Kari Hreinsson, Bita Analui n MD45 North Bldg 228A

229