Thesis REDUCCIÓN DE ORDEN PARA MODELOS DINÁMICOS DE PARQUES EÓLICOS
Date
2016
Authors
VELOSO RIVERA, SIMÓN PEDRO
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
La creciente penetracion de energía eólica esta incrementando dramaticamente elnumero de variables de estado en los estudios dinamicos de sistemas de potencia. Elpresente trabajo propone el uso de dos metodos de reduccion de orden, truncamientobalanceado (BT) y coincidencia de momentos, para aproximar los modelos dinamicosde los aerogeneradores por un lado, y por otro los modelos de parques eolicos completos,ambos con turbinas basadas en tecnologa generador de induccion doblemente alimentado.Se consideran esquemas de control actuales a nivel planta de supervision de tensiony participacion en regulacion de frecuencia.En el actual estado del arte se pueden encontrar tecnicas de reduccion utilizados porinvestigadores en distintas disciplinas como electronica digital, mecanica de estructurasy sistemas de potencia, entre otras. Los metodos de reduccion, de naturaleza linealo no lineal, logran su objetivo a traves de distintos enfoques como la conservacion decaractersticas intrínsecas al modelo de orden completo, ya sean retardos de tiempo maslargos, modos de oscilacion naturales, funciones de transferencia o valores singulares.Los metodos de analisis modal selectivo, truncamiento balanceado y coincidencia demomentos tienen su propia teora e implementacion, considerando la construccion decada algoritmo respectivo.El generador de induccion se controla mediante las corrientes inyectadas en el rotormediante anillos rozantes, de esa forma la velocidad rotorica no depende de la potenciaelectrica sino que se controla a discrecion logrando una curva optima de seguimiento depotencia eolica. Se conecta el otro lado del convertidor al estator y este a la red. Estacon guracion requiere de un convertidor con aproximadamente un tercio de la capacidaddel conjunto y permite a la unidad adaptarse a los nuevos requerimientos normativos.Se llevan a cabo simulaciones en el dominio del tiempo sobre el modelo de un parqueeolico real conectado al Sistema Interconectado Central (SIC) en Chile con diferentesensayos, entre ellos rechazo de carga del generador de mayor tamaño, cambio tipo escalón en la referencia del control supervisor de tension, y cambios en la velocidad deviento. La comparacion entre modelos reducidos y el modelo completo entregan positivosresultados respecto del seguimiento o precision de la señal aproximada, reteniendolos modos oscilatorios y estado estacionario. Adicionalmente, los modelos propuestosutilizan una fraccion del tiempo necesario para simular el modelo completo no lineal.Por todo ello, los modelos reducidos ofrecen un positivo balance entre esfuerzo computacionaly precision para las simulaciones en el dominio del tiempo.
Rising penetration of wind power is increasing the number of state variables in powersystem dynamic studies. This work proposes the use of two model order reduction(MOR) methods, Balanced Truncation (BT) and Moment-Matching, to approximatethe dynamic representation of both doubly-fed induction generators (DFIG) and DFIGbasedwind farms. Plant-level control schemes such as voltage supervisory control andfrequency regulation participation are considered in the analysis.Reduction techniques used by researchers in different disciplines such as digital electronics,mechanics of structures and power systems, among others, can be founded inthe current state-of-art. Reduction methods, of linear or nonlinear nature, achieve theirobjective through different approaches such as conservation of intrinsic features from thefull-order model, whether longer time delays, natural modes of oscillation or eigenvalues,transfer functions or Hankerl singular values. Methods sush as selective modal analysis,balanced truncation and moment-matching have their own theory and implementation,considering the construction of each respective algorithm.The induction generator is controlled by the currents injected into the rotor throughslip rings, thus the rotor speed does not depend on the electrical power but is controlledat discretion, achieving optimal tracking curve of wind power. The other side of theconverter is connected to the stator and to the network. This con guration requires aconverter with approximately one third of the capacity of the assembly and it allows tothe unit adapts itself to new regulatory requirements.Time-domain simulations for a real DFIG-based wind farm connected to the ChileanCentral Interconnected system were conducted for various conditions, among them alarge generator outage, a step change in the set-point of the plant-level voltage supervisorcontrol, and changes in wind speed. Simulation results comparing reduced models andthe full-order model showed that the reduced models were capable of tracing the signal,retaining the oscillations, and reaching the same nal values than the full-order model.Furthermore, the proposed reduced dynamic models could run in a fraction of the timeneeded to simulate the full non-linear wind farm models. Thus, the reduced models cano er a good compromise between computational effort and accuracy in time-domainsimulations.
Rising penetration of wind power is increasing the number of state variables in powersystem dynamic studies. This work proposes the use of two model order reduction(MOR) methods, Balanced Truncation (BT) and Moment-Matching, to approximatethe dynamic representation of both doubly-fed induction generators (DFIG) and DFIGbasedwind farms. Plant-level control schemes such as voltage supervisory control andfrequency regulation participation are considered in the analysis.Reduction techniques used by researchers in different disciplines such as digital electronics,mechanics of structures and power systems, among others, can be founded inthe current state-of-art. Reduction methods, of linear or nonlinear nature, achieve theirobjective through different approaches such as conservation of intrinsic features from thefull-order model, whether longer time delays, natural modes of oscillation or eigenvalues,transfer functions or Hankerl singular values. Methods sush as selective modal analysis,balanced truncation and moment-matching have their own theory and implementation,considering the construction of each respective algorithm.The induction generator is controlled by the currents injected into the rotor throughslip rings, thus the rotor speed does not depend on the electrical power but is controlledat discretion, achieving optimal tracking curve of wind power. The other side of theconverter is connected to the stator and to the network. This con guration requires aconverter with approximately one third of the capacity of the assembly and it allows tothe unit adapts itself to new regulatory requirements.Time-domain simulations for a real DFIG-based wind farm connected to the ChileanCentral Interconnected system were conducted for various conditions, among them alarge generator outage, a step change in the set-point of the plant-level voltage supervisorcontrol, and changes in wind speed. Simulation results comparing reduced models andthe full-order model showed that the reduced models were capable of tracing the signal,retaining the oscillations, and reaching the same nal values than the full-order model.Furthermore, the proposed reduced dynamic models could run in a fraction of the timeneeded to simulate the full non-linear wind farm models. Thus, the reduced models cano er a good compromise between computational effort and accuracy in time-domainsimulations.
Description
Catalogado desde la version PDF de la tesis.
Keywords
DFIG , Parque eólico , Reducción de orden