Abstract: With the design of multiple reactor modules coupling with one steam turbine generator, the high temperature gas-cooled reactor (HTR) ‘unit’ contains multiple reactor modules, which means HTR will endure evident extra impacts of correlations under seismic external events. Therefore, it requires view of the whole HTR unit to fully integrate the correlations from different reactor modules with the overall seismic probabilistic safety analysis (PSA) for multi-module HTR nuclear power plant seismic risks. An almost comprehensive single-module seismic PSA of high temperature gas-cooled reactor pebble bed module (HTR-PM) has already been accomplished. Based on the single-module seismic PSA, the key technologies, which consist of the multi-module event sequence modeling and seismic failure correlation evaluation, of seismic PSA for multi-module HTR are recognized and researched. With the purpose of a reasonable representation of system correlations among several HTR reactor modules as well as the omission of unnecessary function events, the key technology multi-module event sequence modeling needs to implement the combination of common function events in pending single-module event trees. A process is developed by coupling the modeling methods of International Atomic Energy Agency (IAEA) with the site-specific characteristics of multi-module HTR-PM: single-module event tree function events preprocessing; all function events rearranging in order; multi-module modeling based on the first fundamental event tree. It should be noted that an important part of the simplification of multi-module event tree branches is the consequence preprocessing which mainly depends on the definition of the safety goal and the amount of HTR modules. The other key technology seismic failure correlation evaluation reconsidered for the assumption of seismic failure correlation in the single-module seismic PSA may lead to unacceptable results in the multi-module seismic PSA. In this paper, four suggested seismic failure correlation evaluation methods by NRC were reviewed. The most appropriate ‘separation of independent and common variables approach’ was illustrated in detail and furthermore application process of the approach in multi-module HTR was given: 1) modeling and quantifying the seismic failure correlation with typical ‘100% or 0 dependency’ assumption; 2) re-modeling and re-quantifying it for partial dependence if the acceptable standard value is badly unsatisfied or the risk insight is beyond existing knowledge. Then considering the double-module HTR-PM under conditions of seismic-induced loss-of-offsite-power, the multi-module seismic PSA is modeled and the quantification results show a release frequency far less than the probability safety goal. The double-module HTR-PM seismic-induced loss-of-offsite-power case also demonstrates the feasibility of the event sequence modeling strategies and the seismic failure correlation evaluation methodologies of multi-module HTRs.