Wave Energy Converter Cone-Cylinder Buoy Dimension Optimization using Backpropagation Neural Network and Genetic Algorithm Method

Dwitama, Nathanael (2024) Wave Energy Converter Cone-Cylinder Buoy Dimension Optimization using Backpropagation Neural Network and Genetic Algorithm Method. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5007201074-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (5MB) | Request a copy

Abstract

As population grows, an increase in human need for electrical energy is inevitable. Renewable energy-based power generation technology is necessary considering the dwindling availability of fossil fuels and coal. Wave energy converter (WEC) is a technology that can convert waves into electrical energy. Heaving point absorber is the most used type of WEC because it has shape that easily adapts to wave conditions. The unreliability in extreme weather and dependence on buoy size caused WEC technology to be underused. Research to utilize shape changes is needed to reduce WEC dependence on buoy size. The purpose of this research is to design a WEC buoy showing maximum power absorbed with the minimum value of volume and comparing it with the existing buoy shape such as the basic cone-cylinder and bullet shape. The research begins by determining the variant of dimensions for the cone-cylinder buoy head cylinder height with the value of 400, 550, 700, 850, 1000 and 1150 in millimeters and body concavity diameter with the value of 3200, 3900, 4600, 5300, 6000, 6700, 7400 and 8100 in millimeters. The 48 buoy parameters will become a 3D model using Fusion 360 software to be simulated. The Simulation process will use Ansys AQWA software with heaving motion values as output to calculate the power absorbed. The optimization process uses the Backpropagation Neural Network (BPNN) method followed by the Genetic Algorithm (GA) on MATLAB software. The optimized parameters will be simulated to validate the prediction of GA. For the buoy volume BPNN training, the better setting is 2 hidden layers, 5 nodes, Log-Sigmoid activation function, 70% training ratio, 15% validating ratio, and 15% testing ratio with 4.59E-09 mean squared error (MSE). For the buoy power absorbed BPNN training, the better setting is using 4 hidden layers, 5 nodes, Tan-Sigmoid activation function, 70% training ratio, 15% validating ratio, and 15% testing ratio with 5.95E-25 MSE. For GA the better setting is to generate 500 chromosomes with using weight 0.5. The result of GA is 1150 mm of cylinder height and 7146.1 mm of body concavity diameter, simulation shows 12.8503 m3 of volume and 3379 Watt of power absorbed. Compared to basic cone-cylinder buoy, optimized buoy manages to increase power absorbed by 1.70% with only 0.17% increase in volume. Compared to bullet buoy, optimized buoy manages to decrease 5.72% buoy volume and increase 2.53% in power absorbed.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Renewable Energy, Wave Energy Converter, Buoy Optimization, Power Output
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > Q Science (General) > Q325.78 Back propagation T Technology > TJ Mechanical engineering and machinery T Technology > TJ Mechanical engineering and machinery > TJ808 Renewable energy sources. Energy harvesting.
Divisions:	Faculty of Industrial Technology > Mechanical Engineering > 21201-(S1) Undergraduate Thesis
Depositing User:	Nathanael Dwitama
Date Deposited:	14 Aug 2024 05:02
Last Modified:	14 Aug 2024 05:02
URI:	http://repository.its.ac.id/id/eprint/114733

Actions (login required)

View Item