The uncertainties of the predicted load demand and N-K contingencies are very significant aspects to composite generation and transmission expansion planning (CGTEP). In this paper, multi-contingency constrained CGTEP with load uncertainty was analyzed from stringent mathematical view and formulated as a tri-level optimization model. To effectively solve the tri-level optimization, the entire problem is formulated as two problems using Benders' decomposition: master problem with expansion planning and the sub-problem with the worst case load shedding. The sub-problem is a bi-level optimization problem which can be solved mathematically using strong duality theory and linearization method. CGTEP with the tri-level optimization can endure the disturbances of interval load and N-K contingencies. A benchmark test system is simulated to validate the effectiveness of the proposed approach. Furthermore, for Bender's decomposition with many sub-problems of worst load shedding, the numerically comparable results of a special case demonstrate that all sub-problems of composite contingencies must be validated at each iteration even if certain contingency meets the standard of load shedding at the previous iteration.