The invention discloses a power grid load recovery robustness optimization method based on an information gap decision theory. A determinacy load recovery optimization model is converted into a robustness optimization model, and an artificial bee colony algorithm is adopted for carrying out solving, so a load recovery scheme related to indeterminacy is acquired. The method comprises steps of 1, establishing a determinacy load recovery optimization model in a power grid recovery process; 2, adopting the artificial bee colony algorithm to solve a determinacy model to obtain an optimal solution of the determinacy load recovery optimization model; 3, according to an optimal solution of an original model, determining the acceptable smallest recovery quantity of the load recovery; 4, based on the information gap decision theory, establishing a robustness optimization model considering indeterminacy of the load recovery; and 5, adopting the information gap decision theory to solve the robustness model so as to obtain a load recovery scheme capable of achieving an expected recovery target. According to the invention, load fluctuations in a certain range can be borne by the obtained load recovery scheme; safety of a power grid recovery process can be ensured; and the method has certain theoretical and engineering value.