Abstract: To deal with the double uncertainty of the source and load during the load restoration process, a robust optimization model is established. The linear-programming approximation of AC power flow (LPAC) is applied to linearize the original model. Then, based on the decoupling idea, the original model is decomposed into a main problem in the prediction scenario and a sub-problem in the error scenarios. The main problem determines the optimal restoration scheme to maximize the weighted load restoration amount. By introducing the slack variables, the scheme verification sub-problem considering the uncertainty of the prediction errors is transformed into an optimization model with a max-min structure. The load forecast error and the wind power forecast error are described by the adjustable box set and the linear polyhedron set respectively. In the solving process, the strong duality theory is used to transform the sub-problem with the max-min structure into the dual model, and the big M method is introduced to linearize the dual model obtained. And then, the column-and-constraint generation algorithm is utilized to efficiently solve the main and sub-problems iteratively. The numerical results of case system verify the effectiveness and feasibility of the proposed model and method.