To assist the operators in detecting the fault component and identifying the fault developing process，a prototype system of power system fault diagnosis based on temporal Bayesian knowledge bases（TBKB）is studied.The causal relationship and temporal constraint relationship among the fault components and protection operations，protection operations and related breakers tripping are utilized to build TBKB models.The consistency checking is made to find the temporal rationality of the protections and breakers.For the information missing of states in the nodes，the state assumption is adopted to create hypothetical state combinations.Bayesian reasoning is made to detect the fault component and identify the mal-function and rejection of protections and breakers.A prototype system based on the theory is designed，including the modules of data acquisition，TBKB diagnosis model development，fault diagnosis reasoning and layered cause-effect graph display.The key techniques are described，such as the automatic generation method of TBKB fault diagnosis models，implementation of diagnosis algorithm and building and display of layered cause-effect graph.The implementing of prototype system provides a reference for the engineering application of power system fault diagnosis.