Please use this identifier to cite or link to this item:
|Title:||Advanced diagnosis of electrical systems||Authors:||Cao, Huimin||Keywords:||DRNTU::Engineering
|Issue Date:||2016||Abstract:||Modern industry has an increasing demand for high level power. In the same time, renewable energy technology develops rapidly. These two reasons explained why multilevel converter technology attracts more and more attentions. And many investigations have been done on that. Nowadays, the insulated gate bipolar transistor(IGBT) is the most popular fundamental device to achieve high electronic power conversion. A power cell is composed of 4 IGBTs and generates 3 level of output: E, 0 and –E, where E is the dc power supply. High voltage is achieved by connecting the power cell in series. Combing the multilevel converter technology and renewable energy technology together, a power system with no pollution and no fuel cost is capable to be realized. This project focus on the fault identification of multilevel converter circuit. With immediate and effective fault identification, the broken component can be replaced by a new one thus make sure the whole system maintains good performance all the time. Two kind of faults are investigated in this project: sensor measurement error and power cell fault. “l1 regularized least-squares optimization” method is employed for sensor fault identification. Cell fault identification is investigated by trial and error as well as dynamic programming. In trial and error approach, all 32 scenarios for “power_fivecells” circuit are enumerated for observation and test. A MATLAB function is developed to determine each cell broken or in good state independently. In order to detect power cell fault for large scale, dynamic programming is adopted to develop an algorithm to do the identification automatically. Information required is all power supply to fundamental cell, all control signals to each cell and the output. The main idea of this algorithm is to represent the problem into stages, and solving one stage optimization problem at a time. Analysis and results show that “l1 regularized least-squares optimization” method is not effective applied to sensor fault detection since it is costly and not applicable to physical model. And the power cell fault detection function works well on single cell fault as well as multi cells fault. The trial and error technique and dynamic programming approach needs further investigation and testing once a system is designed.||URI:||http://hdl.handle.net/10356/68236||Rights:||Nanyang Technological University||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Student Reports (FYP/IA/PA/PI)|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.