Dynamic fault-detection in shipboard electric load sharing.
MetadataShow full item record
- Institutt for marin teknikk 
The main contribution of this thesis is an investigation of dierent fault detection systems that canbe used for marine diesel electric power plants.The motivation of this thesis is to develop fault detection systems that are capable of detectinggovernor and AVR actuator faults for a marine diesel electric power plant. Faster detection ofthese types of faults reduces the risk of blackout, and might also enable more ecient operation ofmarine diesel electric power systems.The studied system is a diesel electric power system consisting of three parallel connected gensets.The gensets share a common load sharing system and are run in droop mode. The gensets produceelectric power, to supply the consumers on the vessel. A failure in the Governor or AVR, mightcause all gensets to disconnect and hence cause a system blackout. Since diesel electric powersystems are common for DP vessels, a blackout is a severe situation. If a failure occurs in onegenset, this will inuence the operation of the other connected gensets. The failure source mighttherefore be dicult to detect. Better fault detection systems, might therefore enable faster andmore reliable protection of the power system.In this thesis three fault detection systems are investigated. One of the fault detection systemsis commonly used for marine power systems, and is based on the governor and AVR droop-curves.The two other fault detection systems are based on a simplied model of a marine diesel electricpower system. The rst is based on an analytical redundancy relation derived from the systemmodel. The other fault detection system, is based on a nonlinear Luenberger-type observer.To verify the performance of the fault detection systems, they are implemented in Matlab/Simulink.For this purpose, a more advanced power plant model is developed and implemented in Simulink.Simulations are run for several normal and faulty simulation modes. The fault detection systemsperform well for all simulation modes, and are capable to detect the faults within reasonable time.iii