This paper considers a stochastic job allocation problem inspired by a real-life electric power grid operator. The power grid requires regular maintenance to ensure the sustainability of the network. To mitigate undesired power breakdown from the maintenance operations, the objective of the problem is to minimize a stochastic risk function measuring the risk of power grid maintenance schedule. The problem can be formulated as a Mixed Integer Programming (MIP) problem involving many integer variables, which is very challenging to solve. A hybrid mathematical programming-heuristic algorithm is developed to solve large-sized problems, consisting of two phases: The first phase utilizes a simplified MIP formulation with objective perturbation to find feasible solutions quickly. The second phase adapts the simulated-annealing algorithm to improve the solutions. A computation study on the real-life instances shows that the proposed algorithm outperforms Cplex. Especially, the proposed algorithm could find good feasible solutions for all tested instances, while Cplex failed to obtain feasible solutions within three hours.