Aim of the present work is to develop an engineering tool, based on a Double Multiple Stream-Tube method able to reproduce the performance of a small scale H-shaped vertical-axis wind turbine. Low order simulation models are indeed frequently preferred in the wind turbine sector, because of their inherently lower computational requirements. In order to increase prediction accuracy, the base model was integrated with specific state-of-the-art corrections, to account for higher order aerodynamic phenomena: flow curvature effect, finite blade length, aerodynamic interaction of spokes and tower and dynamic stall. The results, validated with respect to an experimental campaign performed in the Wind Tunnel of Politecnico di Milano, provide good approximation of real turbine performance, once the higher-order corrections are implemented. Moreover importance of extended airfoil polar curves was proved. Dynamic stall correction has emerged to have a non-negligible contribution and to be the most uncertain and of recent development. The phenomenon was investigated by means of implementation of different available corrective models, classified mainly in two classes: empirical-based (Gormont's theory and various adaptations) and physical-based (Sheng's model, derived from Beddoes-Leishman theory). The different predictions are compared globally and locally, showing different level of accuracy and complexity: with reference to the available experiments, even though both categories are able to predict mean power coefficient, only the physical-based formulation can reproduce instantaneous local torque, exchanged between the flow and the rotor. To asses which is the uncertainty linked to such models, advanced Uncertainty Quantification tools were used. By applying proper sampling techniques and metamodels based on Kriging/polynomial chaos expansions, and then performing Monte Carlo analysis on such surrogates, it was possible to evaluate the probability distribution of the output performance once an input probability is assigned on the model empirical constants, and to finally provide the engineering tool estimations with an uncertainty band, with a selected confidence.
Obiettivo di questo lavoro di tesi è lo sviluppo di un modello ingegneristico capace di predire con sufficiente accuratezza le prestazioni di una turbina eolica ad asse verticale, di piccola taglia, di tipo H-Darrieus. I modelli ingegneristici sono infatti, spesso, la scelta predefinita nel settore eolico, grazie alla minore richiesta di risorse computazionali. Il modello sviluppato è basato sul concetto di doppio disco attuatore e discretizzazione a tubi di flussi multipli, a cui sono state aggiunte delle specifiche correzioni per integrare effetti aerodinamici di ordine superiore: curvatura del flusso, lunghezza finita delle pale, interazione delle strutture di supporto (razze e albero) e stallo dinamico. I risultati, validati rispetto a una campagna sperimentale presso la Galleria del Vento del Politecnico di Milano, mostrano che queste correzioni sono necessarie per garantire una buona approssimazione delle prestazioni reali della turbina. Si è dimostrata, inoltre, l'importanza dell'utilizzo di curve polari estese ad alti angoli di attacco. Lo stallo dinamico è risultato essere una correzione dall'effetto non trascurabile e caratterizzata da incertezza, poichè di più recente sviluppo. Il fenomeno fisico è stato studiato implementando diversi modelli correttivi disponibili, classificati principalmente in due classi: modelli empirical-based (di Gormont e i vari adattamenti) e physical-based (di Sheng, derivato dalla teoria di Beddoes-Leishman). I risultati sono stati confrontati sia globalmente che localmente, individuando diversi livelli di accuratezza e complessità: rispetto ai dati sperimentali, nonostante entrambi siano in grado di stimare l'effetto sulla produzione media di potenza, solamente la correzione physical-based riesce ad approssimare il momento locale istantaneo, scambiato tra il rotore e il flusso. Per stimare l'incertezza introdotta dal modello di stallo dinamico scelto, sono stati utilizzati gli strumenti della Uncertainty Quantification: scienza che permette di quantificare le incertezze, nei risultati di una simulazione numerica, a partire dalla distribuzione di probabilità degli input, che in questo caso risultano essere i coefficienti empirici delle varie correzioni. Applicando efficienti metodi di campionamento e metamodelli basati su Kriging e espansione in caos polinomiale, e poi svolgendo un'analisi Monte Carlo su questi surrogati, è stato possibile valutare delle bande di incertezza relative al modello ingegneristico sviluppato.
Impact of dynamic stall on VAWT performance : implementation into a double multiple stream-tube tool and quantification of model uncertainty
GALETTA, MANUELA
2017/2018
Abstract
Aim of the present work is to develop an engineering tool, based on a Double Multiple Stream-Tube method able to reproduce the performance of a small scale H-shaped vertical-axis wind turbine. Low order simulation models are indeed frequently preferred in the wind turbine sector, because of their inherently lower computational requirements. In order to increase prediction accuracy, the base model was integrated with specific state-of-the-art corrections, to account for higher order aerodynamic phenomena: flow curvature effect, finite blade length, aerodynamic interaction of spokes and tower and dynamic stall. The results, validated with respect to an experimental campaign performed in the Wind Tunnel of Politecnico di Milano, provide good approximation of real turbine performance, once the higher-order corrections are implemented. Moreover importance of extended airfoil polar curves was proved. Dynamic stall correction has emerged to have a non-negligible contribution and to be the most uncertain and of recent development. The phenomenon was investigated by means of implementation of different available corrective models, classified mainly in two classes: empirical-based (Gormont's theory and various adaptations) and physical-based (Sheng's model, derived from Beddoes-Leishman theory). The different predictions are compared globally and locally, showing different level of accuracy and complexity: with reference to the available experiments, even though both categories are able to predict mean power coefficient, only the physical-based formulation can reproduce instantaneous local torque, exchanged between the flow and the rotor. To asses which is the uncertainty linked to such models, advanced Uncertainty Quantification tools were used. By applying proper sampling techniques and metamodels based on Kriging/polynomial chaos expansions, and then performing Monte Carlo analysis on such surrogates, it was possible to evaluate the probability distribution of the output performance once an input probability is assigned on the model empirical constants, and to finally provide the engineering tool estimations with an uncertainty band, with a selected confidence.File | Dimensione | Formato | |
---|---|---|---|
2019_04_Galetta.pdf
accessibile in internet per tutti
Descrizione: Testo completo della tesi
Dimensione
16.26 MB
Formato
Adobe PDF
|
16.26 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/146402