STUDI PERANCANGAN PENGONTROL NEURO-FUZZY PADA PEMBANGKIT LISTRIK TENAGA SURYA

• Main Author: Witjakso, Ario
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/1563
• Institution: Institut Teknologi Bandung

The use of Neuro-Fuzzy Controller (NFC) is introduced to handle the limitation of on-off conventional controllers in reducing output oscillation generated by Solar Power Generator System. In the thesis, the simulation is carried out using one of the system applications known as Solar Home System SHS). NFC is also expected to solve the excessive solar energy problem that occurs due to limited conventional controller performance. Besides, the resulting NFC is still open for both better operation and further development. The proposed NFC can be trained and learned from data based on ANFIS Adaptive Network Base on Fuzzy Inference System). Training is intended to determine all the control parameters. The determination of these parameters does not only depend on learning algorithm but also data that is fed to the system. Error back propagation algorithm is used to train the controller using data that are generated from SHS model consisting of components such as Photovoltaic, Battery etc. Photovoltaic model having RMSE 0.0072, and Battery having 0.0865 (charging) and 0.0053 (discharging) are used to develop the NFC having 2 inputs, and 1 output. Each input has 5 membership generalized Bell functions. The simulation is also done for 3 and 4-membership functions, using 2000 epochs. The resulting errors are 0.689, 0.700 and 2.404 for 5, 4 and 3 membership function, respectively. The implementation of NFC on SHS model shows that the additional load is needed. However, the designed NFC can still reduce the number of oscillation and store more energy in the battery as much as 13% at charging condition 1 A.