Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA44

3 - Q-learning Applied to Multistage Generation Expansion Planning Vijay Kumar, PhD Candidate, Pennsylvania State University, State College, PA, 16801, United States, Mort Webster, Uday Shanbhag Generation expansion planning is a multi-stage decision about the technology, size and the timing for new electricity generators under uncertainty in future technical and economic uncertainties. The dynamic programming formulation for a time horizon of 10-50 years faces the Curse of Dimensionality and is computationally intractable. We demonstrate a novel adaptive sampling algorithm within a Q-Learning framework applied to multistage expansion under uncertainties in fuel prices and carbon limits. 4 - Solving Appointment Scheduling and Staffing Problem in Radiotherapy by Progressive Hedging Algorithm Cheng-Yu Rao, National Chiao Tung University, Hsinchu, Taiwan, Yu-Jyun Tang, Sheng-I Chen A two-stage integer stochastic programming model for appointment scheduling and staffing decisions in a radiotherapy department is developed. The treatment start and staff assignment are determined in the first stage, and the number of treatments and overtimes are decided in the second stage. A scenario decomposition scheme, along with the implementation of progressive hedging algorithm is presented. The computational experiment investigates different penalty settings for nonanticipative decisions. The convergence of the algorithm has been shown despite of oscillatory behavior is observed. We report the solution quality and runtime for general branch-and-bound and our proposed methods. 5 - Capacity Optimization for Innovating Firms Rita Pimentel, RISE SICS Vasteras, Vasteras, Sweden Rita Pimentel, Cemat, Instituto Superior Técnico, Lisboa, Portugal, Verena Hagspiel, Kuno Huisman, Peter M. Kort, Cláudia Nunes In case of a product innovation the firms start producing a new product. While doing so, they should decide what to do with their existing production process. They can choose between replacing the established production process by the new one, or keep on producing the established product so that they produce two products at the same time. Due to technological progress, the quality of the newest available technology increases over time. So, the firms face the tradeoff between innovating fast that enlarges the payoff soon but only by a small amount, or innovating later that leads to a larger payoff increase, the drawback being that they are stuck with producing the established product for a longer time. n WA44 North Bldg 227C Real-time Optimization in Power Systems Sponsored: Energy, Natural Res & the Environment/Electricity Sponsored Session Chair: Na Li, Harvard University, Cambridge, MA, 02138, United States Co-Chair: Hao Zhu, The University of Texas at Austin, Austin, TX, 78712, United States 1 - Real-time Resilient Operation of Multi-microgrid Networks Amin Gholami, Georgia Institute of Technology, Atlanta, GA, United States, Andy Sun Although microgrids pave the way for integrating various distributed energy resources, they raise many challenges. Emergency control of dangerous transients caused by the transition between the grid-connected and island modes is one of such challenges. To address this, we propose an optimization and real-time control framework for maintaining the frequency stability of multi-microgrid networks under an islanding event. We develop a strong MISOCP-based reformulation as well as a cutting plane algorithm for scalable computation. Our framework provides the optimal load shedding and optimal topology reconfiguration, while the frequency dynamics and AC power flow limitations are considered. 2 - Event Identification and Data Recovery from Highly Quantized Measurements Meng Wang, Rensselaer Polytechnic Institute (RPI), Troy, NY, 12180, United States, Ren Wang One can add noise and apply quantization to synchrophasor measurements to increase the data privacy and reduce the communication cost. This talk focuses on event identification and data recovery at the operators’ side from the highly quantized measurements. Event identification can be achieved by solving a data clustering problem that aims to group measurements affected by the same event together. The recovery and clustering are achieved simultaneously by solving a nonconvex constrained maximum likelihood problem. A fast algorithm with the performance guarantee is proposed to solve the nonconvex problem. The proposed method is evaluated numerically on recorded synchrophasor datasets.

3 - Online Localization of Forced Oscillations in the Power Grid Le Xie, Texas A&M University, College Station, TX, United States, Tong Huan, P.R. Kumar, Nickolaos M. Freris In this talk, we will present a data-driven method to pinpoint the source of a new emerging dynamical phenomenon in the power grid, referred to “forced oscillations in the difficult but highly risky case where there is a resonance phenomenon. By exploiting the low-rank and sparse properties of PMUmeasurements, the localization problem is formulated as a matrix decomposition problem, which can be efficiently solved by the ALM algorithm. The data-driven nature of the proposed method allows for a very efficient implementation. The proposed method may possibly be more broadly useful in other situations for identifying the source of forced oscillations in resonant systems. 4 - Online Optimization with Feedback for Power Systems Emiliano Dall’Anese, University of Colorado, Boulder, CO, United States This talk focuses on real-time incentive-based mechanisms to coordinate distributed energy resources (DERs) with both continuous and discrete control variables. A multi-period social welfare maximization problem is formulated and, based on its convex relaxation, a distributed stochastic dual gradient algorithm is proposed. Real-time implementations involve asynchronous updates and feedback being leveraged to account for nonlinear power flows as well as to reduce communication overhead. The resulting algorithm provides a general online stochastic optimization algorithm for coordinating networked DERs with discrete. 5 - Second-order Decomposition ACOPF Shenyinying Tu, Northwestern University, Evanston, IL, 60201, United States, Ermin Wei, Andreas Waechter This project proposes an algorithm to speed up solving large scale Alternating Current Optimal Power Flow (ACOPF) problem through decomposition and parallelization. In this project, we decompose the power flow network into a meshed network and a number of radial networks, and formulate the entire ACOPF problem as a bilevel optimization problem. In the upper level meshed network, we optimize the entire network and pass the information to the lower level tree networks. Then in each of the lower level problem, we solve it exactly via its relaxed SOCP formulation. The performance of the algorithm can be improved by parallel computing of the lower level problems. n WA45 North Bldg 228A Practice- Energy Policy and Planning I Contributed Session Chair: Yanyan Ding, North China Electric Power University, Beinong road, Changping, Beijing, 102206, China 1 - Resilient Based Optimal Interconnection Planning of Community Microgrids in Power Grids Masoud Barati, Assistant Professor, University of Pittsburgh, Patrick F. Taylor Hall, Room 3325, Benedum Hall, 3700 O’Hara Street, Pittsburgh, PA, 15261, United States We proposed a resilient planning problem of networked microgrids with DERs. We integrate the investment problem and two sets of operating problems associated with the grid-connected and islanding modes of microgrids into a two- stage stochastic model to capture randomness nature of hazard events and long-term load growth, as well as the islanding risk caused by external disturbances with a joint-chance constraint to prevent the risks. A SOCP formulation is presented to incorporate AC-OPF in short-term operation. Numerical results on distribution systems prove the effectiveness of the model. 2 - An Inter-hourly Methodology for Valuation of Windfarms Sergio Cabrales, Assitant Professor, Universidad de los Andes, Calle 19 A. no 1-37 Este, ML-325, Bogota, 111711, Colombia, Carlos Valencia, Daniela Moreno, Sebastian Toro This paper proposes a methodology for the financial valuation of wind power generation based on an hourly estimation approach, including inter-hour velocity and energy spot market. For this purpose, we propose that the energy generation is modelled through an autoregressive copula methodology for univariate series and the spot prices are estimated as the function of two components, a deterministic seasonal pattern and a Gaussian mean-reversion process. We applied the developed methodology to a case study in La Guajira, Colombia, reinforces the idea that an inter-hour approach improves significantly the precision of the financial indicators.

426