This work shows an ongoing activity with the main goal of developing a Digital Twin (DT) model to be applied to a wind farm optimal control strategy. Since Wake Steering has shown very promising results for ‘freeze’ layouts, the aim of this study is to give wind farm designers a compact and reliable tool able to predict the control strategy for the yaw of each turbine in a wind farm with a given wind history. In order to choose the proper simulation tool, the first step has been doing an exhaustive numerical validation of the wind tunnel campaign for the EU Project performed by Politecnico di Milano, CL-Windcon. The tools used for the validation have been FLORIS, FAST.Farm and FLORIDyn; the last one has been chosen as a foundation for the new DT, thanks to its low computational cost and ability to handle dynamic simulations to describe the behavior of the wake. The methodology for this new tool, called FLOWARESC, leads to the construction of a dynamic database for the Wake Steering technique, allowing the implementation of a very basic algorithm to extract the optimal combination of yaw misalignment of all the turbines that maximizes the power output of the farm. Even if the algorithm is not fully optimized, the control technique from FLOWARESC improves the greedy technique (each turbine uses an individual yaw control to have zero yaw misalignment for each wind direction), but also a static look-up table from FLORIS, used in the EU Project. Future research will aim to solve the inefficient algorithm for big clusters, overcoming the exponential complexity by introducing alternatives such as genetic algorithms or surrogate models with Deep Reinforcement Learning techniques, and to implement an interface between FLORIS and FLOWARESC to achieve a pre-initialization in order to obtain the optimal layout and then start the loop of the new tool presented in this work.
Questo lavoro presenta una ricerca in corso che ha l'obiettivo principale di sviluppare un modello di Gemello Digitale (GD) per una strategia di controllo ottimale di un parco eolico. Poiché il Wake Steering ha mostrato risultati molto promettenti per layout "congelati", lo scopo di questo studio è fornire ai progettisti uno strumento compatto e affidabile, capace di prevedere la strategia di controllo dello yaw di ciascuna turbina in un parco eolico con una determinata storia del vento. Per scegliere lo strumento di simulazione più adatto, il primo passo è stata un'esaustiva validazione numerica della campagna in galleria del vento del progetto europeo CL-Windcon, condotta dal Politecnico di Milano. Gli strumenti utilizzati per la validazione sono stati FLORIS, FAST.Farm e FLORIDyn; l’ultimo è stato scelto come base per il nuovo GD, grazie al suo basso costo computazionale e alla capacità di gestire simulazioni dinamiche per descrivere il comportamento della scia. La metodologia del nuovo strumento, denominato FLOWARESC, si basa sulla costruzione di un database dinamico per la tecnica di Wake Steering, permettendo l'implementazione di un algoritmo elementare per ricavare la combinazione ottimale del disallineamento di yaw delle turbine che massimizza l’output di potenza del parco. Sebbene non sia completamente ottimizzata, la strategia di FLOWARESC migliora sia la tecnica "greedy" (in cui ogni turbina utilizza un controllo individuale dello yaw per avere zero disallineamento per ogni direzione del vento) sia una look-up table da FLORIS, usata nel progetto europeo. Le ricerche future punteranno a risolvere l'inefficienza dell'algoritmo per grandi cluster, superando la complessità esponenziale con alternative come gli algoritmi genetici o modelli surrogati tramite tecniche di Deep Learning Technique, e a implementare un'interfaccia tra FLORIS e FLOWARESC per ottenere una pre-inizializzazione al fine di ottimizzare il layout e poi avviare il ciclo dello strumento qui presentato.
Towards a digital twin for optimal control strategies in wind farms
Firrincieli, Lorenzo
2023/2024
Abstract
This work shows an ongoing activity with the main goal of developing a Digital Twin (DT) model to be applied to a wind farm optimal control strategy. Since Wake Steering has shown very promising results for ‘freeze’ layouts, the aim of this study is to give wind farm designers a compact and reliable tool able to predict the control strategy for the yaw of each turbine in a wind farm with a given wind history. In order to choose the proper simulation tool, the first step has been doing an exhaustive numerical validation of the wind tunnel campaign for the EU Project performed by Politecnico di Milano, CL-Windcon. The tools used for the validation have been FLORIS, FAST.Farm and FLORIDyn; the last one has been chosen as a foundation for the new DT, thanks to its low computational cost and ability to handle dynamic simulations to describe the behavior of the wake. The methodology for this new tool, called FLOWARESC, leads to the construction of a dynamic database for the Wake Steering technique, allowing the implementation of a very basic algorithm to extract the optimal combination of yaw misalignment of all the turbines that maximizes the power output of the farm. Even if the algorithm is not fully optimized, the control technique from FLOWARESC improves the greedy technique (each turbine uses an individual yaw control to have zero yaw misalignment for each wind direction), but also a static look-up table from FLORIS, used in the EU Project. Future research will aim to solve the inefficient algorithm for big clusters, overcoming the exponential complexity by introducing alternatives such as genetic algorithms or surrogate models with Deep Reinforcement Learning techniques, and to implement an interface between FLORIS and FLOWARESC to achieve a pre-initialization in order to obtain the optimal layout and then start the loop of the new tool presented in this work.File | Dimensione | Formato | |
---|---|---|---|
2025_04_Firrincieli_Executive Summary.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
1.3 MB
Formato
Adobe PDF
|
1.3 MB | Adobe PDF | Visualizza/Apri |
2025_04_Firrincieli_Tesi.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Testo della Tesi
Dimensione
6.32 MB
Formato
Adobe PDF
|
6.32 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/10589/235037