چكيده به لاتين
Multilevel converters are appropriate choices of power supply in high-power electrical systems. However, a large number of elements required in the structure of multilevel converters, increases the probability of failure, decreases the trust of their use in critical processes, and increases the time of their repair. As a result, it is essential to propose the methods for detecting the exact location and type of faults.
This report provides three methods of fault detection of semiconductor switches and capacitors in the flying capacitor converters. Multilevel converters are formed from numerous semiconductor switches, direct voltage current sources, capacitors and diodes; thus providing a mathematical model for these systems is very complex and time consuming.
Intelligent methods are perfect choices for modeling fault detection system in multilevel converters, as they can model complex and non-linear systems properly.
The first and second methods use two layer neural network and decision tree algorithm, respectively in order to detect faults in the flying capacitor converters. The harmonic components of output voltage are used in both methods. The first method accurately models the fault detection system as it uses non-linear functions. So, this method can detect the location and type of fault in less than 25 milliseconds. The second method is proposed to reduce computations and simplify the fault detection system. This method can be implemented easily as it uses a few comparators; and also, it can detect the location and type of error in less than 30 milliseconds.
The third method is proposed to reduce the time of fault detection. This method uses the current voltage of flying capacitors, so it is good at rapid detection of location and type of fault. This method can detect the location and type of fault in less than 4 milliseconds in the flying capacitor converters.
The practical implementation of flying capacitor converter and the second fault detection method show the accuracy of proposed method to detect the exact location and type of faults.
