Horizontal Axis Wind Turbines Modeling for Smart Grids Applications
Primary Faculty Mentor’s Name
Dr. Adel El Shahat
Proposal Track
Student
Session Format
Poster
Abstract
This research proposes Horizontal Axis Wind Turbines (HWT) modeling for smart grid systems. The HWT is modelled by the use of actual data. Artificial Neural Network (ANN) numerical technique is used to simulate and evaluate the designed proposed work. In order to simulate and predict the characteristics of different types of wind turbines; a lot of real data are taken from the manufacture manual of each type. It is proposed that by identifying the output power from the turbine unit, the design limits would be calculated. The design limits are summarized as follows: Starting wind speed (m/s), Average wind speed (m/s), Hub height (m), Fin length (m), Rotor diameter (m), Rotor speed (r.p.m), and Unit cost ($). MATLAB tool box is used to predict the characteristics correlations based on a non - linear and Artificial Neural Network (ANN) techniques. The data were obtained and checked from about fifty different companies working in the field of manufacturing of wind turbines. The data points are fitted by two methods. The first method is done by the curve fitting tool box, and the second is done by the Neural Network technique with feed-forward back-propagation. Based on the available data units; the hidden layer would be a nine neurons and the output layer would be six neurons. The input is one parameter (Power) and the outputs are six parameters. The configuration here is a general approximator to any function with a log-sigmoid function in the hidden layer and pure- line for output layer. Number of neurons in hidden layer is selected by inspection or by try and error until reaching the desired performance goal, accuracy, and minimum error with little time for training and with low number of neurons as possible. The ANN regression function for each unit is introduced to be used directly.
Keywords
Electrical engineering, Grid application, Renewable energy, Power
Award Consideration
1
Location
Concourse/Atrium
Presentation Year
2014
Start Date
11-15-2014 9:40 AM
End Date
11-15-2014 10:55 AM
Publication Type and Release Option
Presentation (Open Access)
Recommended Citation
Nwobodo, Ekene Micheal, "Horizontal Axis Wind Turbines Modeling for Smart Grids Applications" (2014). Georgia Undergraduate Research Conference (2014-2015). 33.
https://digitalcommons.georgiasouthern.edu/gurc/2014/2014/33
Horizontal Axis Wind Turbines Modeling for Smart Grids Applications
Concourse/Atrium
This research proposes Horizontal Axis Wind Turbines (HWT) modeling for smart grid systems. The HWT is modelled by the use of actual data. Artificial Neural Network (ANN) numerical technique is used to simulate and evaluate the designed proposed work. In order to simulate and predict the characteristics of different types of wind turbines; a lot of real data are taken from the manufacture manual of each type. It is proposed that by identifying the output power from the turbine unit, the design limits would be calculated. The design limits are summarized as follows: Starting wind speed (m/s), Average wind speed (m/s), Hub height (m), Fin length (m), Rotor diameter (m), Rotor speed (r.p.m), and Unit cost ($). MATLAB tool box is used to predict the characteristics correlations based on a non - linear and Artificial Neural Network (ANN) techniques. The data were obtained and checked from about fifty different companies working in the field of manufacturing of wind turbines. The data points are fitted by two methods. The first method is done by the curve fitting tool box, and the second is done by the Neural Network technique with feed-forward back-propagation. Based on the available data units; the hidden layer would be a nine neurons and the output layer would be six neurons. The input is one parameter (Power) and the outputs are six parameters. The configuration here is a general approximator to any function with a log-sigmoid function in the hidden layer and pure- line for output layer. Number of neurons in hidden layer is selected by inspection or by try and error until reaching the desired performance goal, accuracy, and minimum error with little time for training and with low number of neurons as possible. The ANN regression function for each unit is introduced to be used directly.