Home

DPI2008-02153

printable page

Treball final de màster sobre mètodes duals aplicats a la l'optimització de problemes estocàstics de mercats elèctrics.

Wed, 03/30/2011 - 11:00 — admin

El passat dimecres 16 de març es va presenta a la Facultat de Matemàtiques i estadística el treball de final de màster titulat Optimización de modelos estocásticos de mercado eléctrico múltiple mediante métodos duales realitzat per l'alumne Unai Aldasoro, del Màster d'Estadística i Investigació Operativa UPC-UB, sota la meva direcció. En aquest treball s'estudia l'aplicació del mètode d'optimització dual conegut com a proximal bundle method, descrit a [1] a la resolució del problema estocàstic d'optimització de l'oferta a mercats elèctrics múltiples desenvolupat a [2].

Aquest treball, que forma part del projecte de recerca del MICINN DPI2008-02153 i va ser sel·leccionat en la 4a convocatòria d'ajuts CERMET de la FME a la realització de treballs finals de màster, li ha estat concedida la menció "Matrícula d'Honor" per la Comissió de d'Avaluació  de Treballs Fí de Màster del MEIO, a proposta del tribunal que el va jutjar.

 
[1]  J. B. Hiriart-Urruty, C. Lemaréchal, Convex Analysis and Minimization Algorithms II – Advanced Theory and Bundle Methods. Springer-Verlag, 1993.

[2] Cristina Corchero, F.-Javier Heredia, Optimal Day-Ahead Bidding in the MIBEL's Multimarket Energy Production System, Proceedings of the 7th Conference on European Energy Market EEM10, Madrid, IEEE, pp. 1 - 6 , DOI: 10.1109/EEM.2010.5558714

Optimización de modelos estocásticos de mercado eléctrico múltiple mediante métodos duales

Wed, 03/30/2011 - 09:47 — admin
Publication TypeTesis de Grau i Màster // BSc and MSc Thesis
Year of Publication2011
AuthorsUnai Aldasoro Marcellan
DirectorF. Javier Heredia
Tipus de tesiMSc Thesis
TitulacióMàster in Statistics and Operations Research
CentreFacultat de Matemàtiques i Estadística, departament d'Estadística i Investigació Operativa, UPC
Data defensa16/03/2011
Nota // markMatrícula d'Honor (10/10)
Key Wordsteaching; research; dual methods; electricity markets; DPI2008-02153; mixed integer nonlinear programming; proximal bundle method; optimal day-ahead bid; electricity multimarket; MSc Thesis
AbstractEl presente trabajo plantea la resolución computacional de un modelo de optimización de la oferta de generación eléctrica para compañías eléctricas que participan en el mercado eléctrico liberalizado MIBEL. Dicho mercado se circunscribe a España y Portugal y se compone de una serie de subastas energéticas consecutivas donde el operador de mercado realiza para cada una de ellas la casación entre la oferta y demanda. Así, el objetivo de la compañía generadora será maximizar los beneficios obtenidos en la participación del conjunto de mercados teniendo en cuenta el cumplimiento de las obligaciones contractuales ya establecidas. El modelo matemático propuesto para su caracterización corresponde a un modelo de programación estocástica multietapa cuyo equivalente determinista es un problema de optimización cuadrática con variable binaria. Con el objetivo de aprovechar la estructura del problema se procede a plantear la dualización de un grupo de restricciones que producen que el problema original pueda ser dividido en subproblemas. Para su resolución se deberá estudiar la idoneidad de diversos métodos duales (subgradiente, Bundle Methods, ACCPM) y seleccionar el más conveniente para el caso abordado. La decisión finalmente adoptada ha consistido en elegir como método de resolución el algoritmo Proximal Bundle Method descrito en [18] y adaptado satisfactoriamente a problemas de coordinación de la generación hidro-térmica [17]. El análisis de método Proximal Bundle Method corresponderá a su compresión e interpretación gráfica, a la resolución de un ejemplo de pequeña escala de manera analítica y a su resolución computacional. El objetivo de la fase de resolución será valorar el proceso iterativo y la convergencia del Proximal Bundle Method aplicado al problema multimercado de oferta óptima y la comparación de resultados respecto a otro método dual como el método del subgradiente. La implementación computacional se realizará mediante el lenguaje C++, específicamente se utilizará el metalenguaje Concert Techonolgy creado por IBM para el enlace entre el código C++ y el solver CPLEX. Se comprueba que dicho lenguaje tiene como ventajas principales su simplicidad estructural y el compacto código que produce. No obstante la implementación del Proximal Bundle Method manifiesta una serie de limitaciones prácticas de Concert Technology en cuanto al almacenado y actualización de problemas de optimización. Se propone como línea de futuro el análisis de lenguajes alternativos. En todo caso, los resultados obtenidos desprenden que el Proximal Bundle Method se adapta satisfactoriamente al problema multimercado de oferta óptima, además se concluye que en la aplicación numérica considerada un tamaño de Bundle ilimitado produce los mejores resultados. Además en trabajo propone una serie de líneas de investigación futuras en las que destacan la paralelización de la resolución de los subproblemas, y la definición del subproblema asociado a cada térmica como un problema de caminos mínimos
DOI / handlehttp://hdl.handle.net/2099.1/13917
URLClick Here
ExportTagged XML BibTex

Dr. Marcos J. Rider and Dr. Marina Lavorato visits GNOM

Fri, 02/04/2011 - 08:39 — admin
marina, Marcos and myself (left to right)Dr. Marcos J. Rider and Dr. Marina Lavorato, both professors of the dept. of electrical engineering of the Univ. Estadual Paulista, UNESP, (Brasil), have been visiting the GNOM site in Barcelona during January as a part of their participation on the research project DPI2008-02153. We are collaborating in the development of  new stochastic programming models for the optimal expansion of the transmision network in electricity markets. They already visited our group on 2008, and we expect to continue this collaboration in the future reinforcing the relationship between GNOM and the UNESP, both in research and postgraduate tuition.
 
PD: Dr. Rider won last week a permanent position of associate lecturer in the UNESP. Congratulations Marcos!!

Congratulations Dr. Cristina Corchero!

Thu, 02/03/2011 - 16:45 — admin

Dr. Heredia (left) and Corchero (right).Yesterday February 2 my doctorate student, and friend, Cristina Corchero presented her Ph. D. dissertation on optimal bids in electricity markets. Dr. Corchero developed stochastic programming optimization models to find the optimal bid curve that integrates bilateral and future contractes in the sequence of short-term MIBEL's electricity markets.

The members of the examination panel were prof. Stein-Erik Fleten , from the Norwegian University of Science and Technology, prof. Andres Ramos from the Universidad Pontificia de Comillas and prof. Jordi Castro, from the  Universitat Politècnica de Catalunya. After a very interesting discussion they agreed to give the maximum qualification, Cum Laude, to the research work of Ms. Corchero. I would like to thank the members of the examination committe for their  interesting comments and analysis of the thesis contributions. I'm also very grateful with prof. Pilar Muñoz and  prof. Marcos J. Rider for their collaboration in the supervision of the thesis.

And, of course, my most sinceres congratulations to Dr. Cristina Corchero.

Short Term Bidding Strategies for a Generation Company in the Iberian Electricity Market

Thu, 01/27/2011 - 15:01 — admin
Publication TypeThesis
Year of Publication2011
AuthorsCristina Corchero
Academic DepartmentDept. of Statistics and Operations Research. Prof. F.-Javier Heredia, advisor.
Number of Pages166
UniversityUniversitat Politècnica de Catalunya
CityBarcelona
DegreePhD Thesis
Key Wordsresearch; teaching; DPI2008-02153; electricity markets; stochastic programming; MIBEL
AbstractThe start-up of the Iberian Electricity Market introduced a set of new mechanisms in the Spanish electricity sector that forced the agents participating in the market to change their management policies. This situation created a great opportunity for studying the bidding strategies of the generation companies in this new framework. This thesis focuses on the short-term bidding strategies of a price-taker generation company that bids daily in the Iberian Electricity Market. We will center our bidding strategies on the day-ahead market because 80% of the electricity that is consumed daily in Spain is negotiated there and also because it is the market where the new mechanisms are integrated. One of the main contributions of this thesis has been the study the Spanish electricity price time series and its modeling by means of factor models. In this thesis, the new mechanism introduced by the Iberian Market that a fects the physical operation of the units is described. In particular, it considers in great detail the inclusion of the physical futures contracts and the bilateral contracts into the day-ahead market bid of the generation companies. The rules of the market operator have been explicitly taken into account within the mathematical models, along with all the classical operational constraints that a fect the thermal and combined cycle units. The expression of the optimal bidding functions are derived and proved. Once these main objectives were full filed, we improved the previous models with an approach to the modeling of the influence that the sequence of very short markets have on optimal day-ahead bidding. These markets are cleared just before and during the day in which the electricity will be consumed and the opportunity to obtain benefi t from them changes the optimal day-ahead bidding strategies of the generation company, as it has been shown in this thesis. The entire models presented in this work have been tested using real data from a generation company and Spanish electricity prices. Suitable results have been obtained and discussed.
URLClick Here
ExportTagged XML BibTex

Anounce of Cristina Corchero's PhD Thesis defense

Thu, 01/27/2011 - 14:52 — admin
It is my pleasure to announce that next wednesday February 2, Cristina Corchero, member of the GNOM research group , will defense the PhD. Thesis dissertation
 
 
which I proudly supervised. All of you are welcome to the dissertation defense that will take place on February 2 at 11.00 am in the Sala d'Actes of the Facultat de Matemàtiques i Estadística, UPC, Barcelona.

A stochastic programming model for the optimal electricity market bid problem with bilateral contracts for thermal and combined cycle units

Sun, 01/16/2011 - 23:55 — admin
Publication TypeJournal Article
Year of Publication2012
AuthorsF.-Javier Heredia; Marcos J. Rider; C. Corchero
Journal TitleAnnals of Operations Research
Volume193
Issue1
Pages107-127
Start Page107
Journal Date2012
PublisherSpringer
ISSN Number0254-5330
Key Wordsresearch; paper; stochastic programming; day-ahead market; combined cycle; bilateral contracts; modeling; DPI2008-02154
AbstractThis paper develops a stochastic programming model that integrates the most recent regulation rules of the Spanish peninsular system for bilateral contracts in the dayahead optimal bid problem. Our model allows a price-taker generation company to decide the unit commitment of the thermal and combined cycle programming units, the economic dispatch of the bilateral contract between all the programming units and the optimal sale bid by observing the Spanish peninsular regulation. The model was solved using real data of a typical generation company and a set of scenarios for the Spanish market price. The results are reported and analyzed. The main contributions of this paper include: (a) a new model for the optimal bid function and matched energy for thermal and CC units, (b) a new and detailed mixed-integer formulation of the operation rules of the CC units and (c) the joint optimization of all the above-mentioned factors together with the BC duties. The model was tested with real data of market prices and programming units of a GenCo operating in the Spanish electricity market.
URLClick Here
DOI10.1007/s10479-011-0847-x
Preprinthttp://hdl.handle.net/2117/2282
ExportTagged XML BibTex

A Stochastic Programming Model for the Thermal Optimal Day-Ahead Bid Problem with Physical Futures Contracts

Sun, 01/16/2011 - 23:51 — admin
Publication TypeJournal Article
Year of Publication2011
AuthorsCristina Corchero; F.-Javier Heredia
Journal TitleComputers & Operations Research
Volume38
Issue11
Pages1501-1512
Start Page1501
Journal Date2011
PublisherElsevier
ISSN Number0305-0548
Key Wordsresearch; paper; stochastic programming; optimal bod; day-ahead market; MIBEL; DPI2008-02154; modeling
URLClick Here
DOI10.1016/j.cor.2011.01.008
ExportTagged XML BibTex

New paper accepted for publication in Annals of Operations Research

Wed, 01/12/2011 - 16:47 — admin

The work A stochastic programming model for the optimal electricity market bid problem with bilateral contracts for thermal and combined cycle units of F.-Javier Heredia, Marcos J. Rider and C. Corchero has been accepted for publication in the journal Annals of Operations Research. A preliminary version of the manuscript is available at E-Prints UPC http://hdl.handle.net/2117/2282. This study, which was developed as a part of the research project DPI2008-02153,  allows a price-taker generation company to decide the unit commitment of the thermal and combined cycle programming units, the economic dispatch of the bilateral contract between all the programming units and the optimal sale bid by observing the Spanish peninsular regulation.

New paper accepted for publication in Computers & Operations Research

Mon, 01/10/2011 - 17:52 — admin
The work A Stochastic Programming Model for the Thermal Optimal Day-Ahead Bid Problem with Physical Futures Contracts of C. Corchero and F.-Javier Heredia, has been accepted for publication in the journal Computers & Operations Research (DOI:10.1016/j.cor.2011.01.008). A preprint version of the manuscript is available at http://hdl.handle.net/2117/2795. The goal of this work, which was developed as a part of the research project DPI2008-02153,  is to optimize coordination between physical futures contracts and the day-ahead bidding which follow the MIBEL's regulation. The authors propose a stochastic quadratic mixed-integer programming model which maximizes the expected profits, taking into account futures contracts settlement.
« first‹ previous123456next ›last »
Syndicate content