********************************************************************
Tuesday Oct 3, 2017 at 17:15, room INF 017
Title: Feedback-Based Real-Time Optimization of Multiphase Distribution Networks
Speaker: Dr. Andrey Bernstein,NREL, USA
Abstract: We outline an algorithmic framework for real-time optimization and control of distribution grids that leverages a feedback-based online optimization methodology. The goal of the online algorithm is to maximize operational objectives of distribution-level distributed energy resources (DERs), while satisfying the network-wide constraints and adjusting the aggregate power generated (or consumed) in response to services requested by grid operators. The design of the online algorithm is based on a primal-dual method, suitably modified to accommodate appropriate measurements from the distribution network and the DERs. By virtue of this approach, the resultant algorithm can cope with inaccuracies in the representation of the AC power flows, it avoids pervasive metering to gather the state of noncontrollable resources, and it naturally lends itself to a distributed implementation. Optimality claims are established in terms of tracking of the solution of a well-posed time-varying convex optimization problem.
*******************************************************************************************************************************
Thursday August 31 at 17:00 INF017
Title: Distributed optimization of stochastic energy systems
Speaker: Dr. Roman Legoff-Latimier, ENS Rennes
Abstract: The multiplication of renewable energy sources and flexible loads pave the way for a collaborative management in order to make the most of their complementarities. The case of an electric vehicle EV fleet which is associated to a PV plant will here be considered as an example. The management of the EV fleet will first be presented from an aggregated point of view and solved using Stochastic Dynamic Programming. Then the particular situation of each vehicle will be taken into account thanks to a distributed resolution based on ADMM. Finally, this distributed resolution enables to differentiate the users behaviours and preferences. Their charging schedule will then correspond to their own choices.
********************************************************************************************************************************
DESL-LCA2 Smart Grid Workshop
Nov 28, 2016, INF017
9 :30 Introduction / Session 1
Andreas kettner: Improved L-index for microgrid stability assessment
Asja Derviskadic: New class M PMU
Cong Wang: Existence and Uniqueness of Load-Flow Solutions
Eleni Stai: Dispatching of Stochastic Heterogeneous Resources Accounting for Grid Losses and Imperfect Batteries
10:45 Break
11 :00 Session 2
Enrica Scolari: Model-driven forecasting of PV
Fabrizio Sossan: Dispatchable feeder
Lorenzo Reyes: Experimental validation of the COMMELEC
Lorenzo Zanni: PECE method
Tech Tesfay: PMU patching
12:15 Break (standing lunch in INF017)
12 :45 Session 3
Marco Pignati: RTSE applications to locate faults
Mokhtar Bozorg: Impact of dispatchable feeders on the system reliability and reserve
Mostafa Nick: AR-OPF
Nadia Christakou: Inference of admittance matrix from measurements
13 :45 Break
14 :00 Session 4
Roman Rudnik: Handling Large Power Steps in Real-Time Microgrid Control
Sergio Barreto: Undetectable timing-attacks on linear state estimators
Maaz Mohiuddin: Tolerating delay faults in the Commelec controller
Wajeb Saab: Reliability of Control Point
15 :00 Conclusion
*****************************************************************************************************************************
* Wednesday 5 Aug 2015 14:00 Laboratory for Electronics, MIT Title: Commelec: Real-Time Control of Electrical Grids Speaker: Maaz Mohiuddin, PhD Student, LCA2, EPFL
Slides from the talk.
*****************************************************************************************************************************
* Friday 7 May 2015 10:00 ELG 116 Title: Some Applications of Sparse Estimation in Power Grids Speaker: Professor Ali Abur, Northeastern University, Boston, MA
Sparse estimation problems show up in many engineering projects where the measurements are few compared to the unknown variables to be estimated with the additional information that majority of the unknowns are expected to be exactly zero. Such problems can be shown to exist in power grid operation by reformulating the problems to transform the unknown vector into a sparse array by exploiting the special structure of the power network equations. In this talk we will describe two such problems. The first one considers identification of branch outages for grids which are not fully observable. Possible cases include identification of outages in external systems from which a very limited number of real-time measurements are typically available. The problem of line outage identification will be formulated as a sparse selection problem and will be solved by a mixed integer programming algorithm. The second application considers fault identification and location in power grids where synchronized voltage measurements are sparsely available at a limited number of buses. Fault location and identification is first formulated as a sparse regression problem and then solved using the least absolute selection and shrinkage operator (lasso) algorithm. The approach appears to work irrespective of the fault type and branch configuration. These and other similar applications are expected to motivate further work in effective utilization of synchrophasors which are quickly populating power grids all around the globe.
******************************************************************************************************************************
* Friday 13 March 2015 17:00 BC 329 Title: Distributed Non-Convex Optimization, Application to the Optimal AC Power Flow Speaker: Jean-Hubert Hours, STI, EPFL
Many engineering problems, such as the optimal AC power flow, can be cast as nonlinear programs involving network constraints. For solving such problems, distributed optimization strategies are becoming increasingly popular. However, for such methods, convergence guarantees generally hinge upon convexity, which is not satisfied in a significant number of practical cases. In this talk, we present a novel decomposition algorithm for solving distributed non-convex programs with convergence guarantees to a local minimum. We further discuss the applicability of our approach to solve parametric nonlinear programs, as they arise in real-time optimization, and present theoretical results about stability of the optimality tracking scheme. Moreover, we show how the local convergence rate of our algorithm can be improved by merging it with an inexact Newton method, while preserving a distributed algorithmic structure. The proposed algorithm is successfully applied to compute local solutions to optimal AC power flow problems over transmission and distribution networks.