Search

In the wind industry, power prediction of wind farm is commonly implemented by analytical wake models, which is low-cost but insufficient in accuracy for high-turbulent wake modelling. In this study, a novel machine-learning-based wake model is developed to improve the power prediction of wind farms. The presented model can reproduce the velocity and turbulence fields in turbine wakes commensurate to the high-fidelity Computational Fluid Dynamics (CFD) simulations while achieving good computational efficiency. Driven by massive CFD simulation dataset, the implicit relationship between inflows and wake flows is established using the Artificial Neural Networks (ANN) technique based on back-propagation algorithm. The reduced-order method Actuator Disk Model with Rotation (ADM-R) and modified k − ε turbulence model are implemented into RANS simulations to save the computational costs dramatically in producing the big-data of wake flows. The ANN wake model is deployed in the Horn Rev wind farm, and validated against LES, onsite measurement, and analytical wake models. The conclusions show that the ANN model can appreciably improve the power predictions compared with the existing analytical models and match the LES and measurement data well. The validated model is also adopted to investigate the influence of wind direction and turbine layout on power production of wind farms. • The machine learning-based wake model is competitive in both efficiency and accuracy for wind farm power prediction. • Turbulence modeling dominates the recovery speed of turbine wake, which is critical for accurate power prediction. • Comprehensive predictions of power output and losses inside Horns Rev wind farm under various wind directions are discussed. • In the Horns Rev wind farm, aligned layout performs better than staggered layout due to the local dominated wind directions. [ABSTRACT FROM AUTHOR]

This paper presents a data-driven wind farm layout optimization framework that uses a machine learning wake model that considers physical control stages. The machine learning wake model is trained using well-validated Computational Fluid Dynamics (CFD) simulation data, and consists of thousands of sub-models, each of which is an artificial neural network (ANN) wake model. The ANN wake models are trained separately for low-speed and high-speed inflows to ensure high accuracy of the predictions, with less than 2% error compared to CFD simulations. The accuracy and efficiency of the framework are validated, and the results show better agreement with CFD simulation than an analytical wake model developed in recent years. A parametric study on the number of wind turbines in the Horns Rev wind farm demonstrates that more wind turbines can be added with a minor decrease in average power, with more even and staggered layouts. Under full-wake uniform inflow, the selected analytical wake model exhibits a power prediction error of 5%–8%, while the differences between optimal layouts searched by different wake models range from 2% to 8%. When introducing a wider range of inflow direction sectors, the discrepancy between optimal layout performances decreases. [ABSTRACT FROM AUTHOR]

A wind turbine operating in the wake of another turbine and has a reduced power production because of a lower wind speed after rotor. The flow field in the wake behind the first row turbines is characterized by a significant deficit in wind velocity and increased levels of turbulence intensity. To maximize the wind farm net profit, the number of turbines installed in the wind farm should be different in depend on wind farm project investment parameters. Therefore modeling wake effect is necessary because it has a great influence on the actual energy output of a wind farm. In this paper, the extreme learning machine (ELM) coupled with wavelet transform (ELM-WAVELET) is used for the prediction of wind turbine wake effect in wind far. Estimation and prediction results of ELM-WAVELET model are compared with the ELM, genetic programming (GP), support vector machine (SVM) and artificial neural network (ANN) models. The following error and correlation functions are applied to evaluate the proposed models: Root Mean Square Error (RMSE), Coefficient of Determination ( R 2 ) and Pearson coefficient ( r ). The experimental results show that an improvement in predictive accuracy and capability of generalization can be achieved by ELM-WAVELET approach (RMSE = 0.269) in comparison with the ELM (RMSE = 0.27), SVM (RMSE = 0.432), ANN (RMSE = 0.432) and GP model (RMSE = 0.433). [ABSTRACT FROM AUTHOR]

Production losses and increased turbine loadings are observed in wind farms, when wind turbines interact with each other. If a wind turbine is located in the wake of another one, its incoming flow is disturbed, slowed down, and its potential wind power is decreased. It is therefore necessary to study the wind turbine wakes and their interactions. It is important to consider these wake effects in the design of a wind farm in order to maximize the energy output and lifetime of the machines. The exact modeling of the wind speed distribution within a wind park is a fairly complicated task and many of the necessary parameters are not routinely available. A large number of studies have been established concerning the calculation of wake effect. Even though a number of mathematical functions have been proposed, there are still disadvantages of the models like very demanding in terms of calculation time. Artificial neural networks (ANN) can be used as alternative to analytical approach as ANN offers advantages such as no required knowledge of internal system parameters, compact solution for multi-variable problems and fast calculation. In this investigation adaptive neuro-fuzzy inference system (ANFIS), which is a specific type of the ANN family, was used to predict the wake power deficit. Neural network in ANFIS adjusts parameters of membership function in the fuzzy logic of the fuzzy inference system (FIS). This intelligent algorithm is implemented using Matlab/Simulink and the performances are investigated. The simulation results presented in this paper show the effectiveness of the developed method. [ABSTRACT FROM AUTHOR]