In　the　long-term　transmission　system　hardening,　critical　components　in　transmission　infrastructure　should　be　identified　and　hardened　to　reduce　the　load　loss　risk　in　the　presence　of　scheduled　and　unscheduled　outages.　Generally,　the　probabilistic　information　of　assets　in　power　systems　can　be　evaluated　via　historical　statistics.　On　the　basis　of　the　prior　probabilistic　information,　this　study　proposed　a　tri-level　optimisation　model　to　deal　with　the　problem　of　long-term　transmission　system　hardening,　in　which　the　risk　assessment　is　taken　into　account　in　the　objective　function.　Furthermore,　the　problem　is　formulated　as　a　two-stage　robust　optimisation　model.　To　address　the　non-linear　problem　in　the　inner　model,　logarithmic　transformation　and　piecewise　linearisation　are　utilised　to　exactly　linearise　the　non-linear　terms　in　the　model.　Finally,　the　standard　column-and-constraints　generation　algorithm　is　employed　to　solve　the　proposed　model　with　a　master-sub-problem　framework.　The　test　results　on　a　standard　IEEE　RTS-96　system　show　the　effectiveness　of　the　proposed　model.