Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC45

2 - Renewable Scenario Generation using Adversarial Networks Baosen Zhang, University of Washington, Yize Chen, Wang Yishen, Daniel Kirschen Scenario generation is an important step in the operation and planning of power systems. In this talk, we present a data-driven approach for scenario generation using the popular generative adversarial networks, where to deep neural networks are used in tandem. Compared with existing methods that are often hard to scale or sample from, our method is easy to train, robust, and captures both spatial and temporal patterns in renewable generation. In addition, we show that different conditional information can be embedded in the framework. Because of the feedforward nature of the neural networks, scenarios can be generated extremely efficiently. 3 - Pricing Differentiated Services in Electric Vehicle Public Charging Station Networks Mahnoosh Alizadeh, University of California-Santa Barbara, Santa Barbara, CA, United States, Ahmadreza Moradipari We study the pricing problem of an electric vehicle charging network operator (CNO) offering differentiated access plans to its stations. We consider a scheme where users cannot directly choose which station to use when charging their vehicles. Instead, they are routed by the CNO to different stations with heterogeneous wait times and amenities based on their service plan. We design incentive compatible pricing-routing policies that take into account the heterogeneous energy needs of users, their value of time, travel plans, as well as locationally-variant prices of electricity and station capacities. 4 - Dynamic Power Distribution System Management with a Locally Connected Communication Network Hao Zhu, The University of Texas at Austin, 2501 Speedway, Austin, TX, 78712, United States, Kaiqing Zhang, Wei Shi Coordinated optimization of distributed energy resources (DERs) is a key distribution system management (DSM) problem. Two challenges exist therein: i) the possibly disconnected communication network; and ii) the system dynamics from the variable DERs/loads and measurement error. This talk will present the modeling and algorithm design for DSM by addressing these two concerns. First, a game-theoretic characterization is proposed to account for a locally connected communication network with Nash equilibrium analysis. Second, a projected- gradient based asynchronous DSM algorithm is developed for distributed equilibrium learning, with its convergence speed and tracking error analyzed. 5 - Car and Ride Sharing Platforms with Electric Vehicle Fleets Subhonmesh Bose, University of Illinois-Urbana Champaign, 306 N. Wright Street, 4058 ECE Building, Urbana, IL, 61801, United States Shared usage and electrification in urban transit systems are on the rise. Motivated by that trend, we analyze the business model of car sharing services such as Autolib’ and ride sharing services such as Uber with an electrified car fleet. Electric vehicles (EVs) impact the operation of such platforms in two ways. First, battery charging considerations affect waiting times for customers, driver’s willingness to accept rides, etc., and influence the revenues from transit services. Second, plugged-in EVs can garner revenues by utilizing their batteries to provide grid services. We employ a queuing theoretic framework to study the impact of these factors on pricing policies of such platforms. n WC45 North Bldg 228A Practice- Electrical Markets Contributed Session Chair: Srinivasa Prasanna, Electronics City, Opposite Infosys Technologies, Bangalore, 560100, India 1 - Using Python to Decompose Reduced Costs for a Capacity Expansion Model of the Electric Power Sector Kelly Eurek, National Renewable Energy Laboratory, 15013 Denver W. Pkwy, Golden, CO, 80401, United States The Regional Energy Deployment System (ReEDS) is a capacity expansion model that identifies least-cost solutions to build and operate the US electric power grid. ReEDS is formulated as a linear program and written in GAMS. To understand the decision making in ReEDS, we designed a Python tool to harvest data from the MPS and GAMS solution files to reconstruct the reduced costs of the decision variables. Examining reduced costs allows us to calculate the cost and value streams of supply-side technologies and compare which options recover costs versus those that do not. This Python tool helps to identify errors in the model, provides solution transparency, and can be applied to other models written in GAMS.

2 - Implement Real Time Pricing with Multiarmed Bandit Games Andrew Lu Liu, Associate Professor, Purdue University, 315 North Grant Street, School of Industrial Engineering, West Lafayette, IN, 47907, United States, Zibo Zhao The situation where price-responsive consumers determine what to do in the near future (such as when to charge their PEVs) forms a dynamic and incomplete-information game, in which the consumers’ collective actions will impact electricity prices, which in turn affect their payoffs. We propose a multiarmed bandit (MAB) game framework in which each consumer plays an MAB problem to minimize the cumulative regret, as opposed to naively responding to day-ahead prices. Numerical results show very fast convergence to a steady-state of the MAB game with much reduced price volatility and lower transmission congestion costs than the na ve-response case. 3 - Market Design and Competition in Short Term Electricity Markets - Lessons Learned from a Stochastic Multistage Energy System Model Frieder Borggrefe, German Aerospace Center (DLR), Pfaffenwald Ring 38-40, Stuttgart, 70563, Germany This paper provides results from an integrated model for short-term electricity markets, including day-ahead, intraday and balancing markets. Coping with volatile renewable feed-in is a key challenge to the future European electricity system. The paper shows first results from a stochastic linear commitment and dispatch model. Aim of the model is to analyze how trading in the short-term markets will change with increasing shares of renewable energies. The model results are part of the BEAM-ME project: This project developes speed-up methods and applies models to high performance computing (HPC). The paper discusses the next steps to expand the model and challenges when applying the model to HPC. 4 - Community Market Design for Unlocking Congested Distributed Energy Operations Jesus Nieto-Martin, Senior Research Fellow, London Business School, Sussex Place, Regent’s Park, London, NW1 4SA, United Kingdom, Derek W. Bunn This study is motivated by the Orkney Archipelago in Scotland, which presents a substantial amount of renewables connected to undersized infrastructure and is only linked to the Scottish mainland by two 33kV submarine cables. Currently, there is a curtailment mechanism based on the Last-In-First-Out (LIFO) principle. This study provides a market-based alternative to improve current operations by decreasing the curtailment of wind generators. This is achieved through participation into a Blockchain-based Community Local balancing market among islands, unlocking demand side response actions among them and increasing their trading opportunities. 5 - Handling Optimization in Large Scale Energy Management with Hierarchical Approach Srinivasa Prasanna GN, Professor, IIITB, Bangalore, India, Sunil K. Vuppala We present optimization methods of energy management in large scale smart grid systems. We handle violated coupling constraints in a hierarchical approach. We present three cases covering different possibilities of feasibility in the hierarchical approach and techniques to handle the infeasibility. We compare our results with non-hierarchical and a proposed All-or-None heuristic. n WC46 North Bldg 228B Sustainable Development of Food, Water, and Energy Resources Sponsored: Energy, Natural Res & the Environment/Energy Sponsored Session Chair: Yao Zhao, Rutgers University, Newark, NJ, 07102-1895, United States Co-Chair: Kwon Gi Mun, Fairleigh Dickinson University, Teaneck, NJ, United States 1 - Designing Wastewater Supply Chains for Riparian Developing States Designing wastewater supply chains for riparian states has become a highly controversial policy issue because of their asymmetric externality relationships and vulnerability positions. The problem is particularly pronounced in low- income countries facing tight public budgets and development uncertainties. We develop the concept of wastewater management for riparian states by proposing a spatially, temporally, and technologically explicit optimization model. A policy guidance for designing wastewater supply chains is provided for the case of Ganga river management in India. Jiyong Eom, KAIST College of Business, 85 Hoegiro, Dongdaemun-gu, Seoul, 130-722, Korea, Republic of, Kwon Gi Mun, Yao Zhao

484