With the large-scale access of new energy to the power grid, in order to effectively solve the uncertainty of load frequency control caused by the energy’s randomness and fluctuation, and to realize the optimal operation of the Load Frequency Control (LFC) system with multiple constraints, an LFC polytopicl model with wind turbines is established to reduce the influence of model parameter uncertainty on the control system. The Tube-Robust Model Predictive Control (Tube-RMPC) strategy based on the Primal-Dual Neural Network (PDNN) is designed. Combining a nominal model predictive controller with an auxiliary feedback controller, the nominal model predictive controller is solved by PDNN in real time to ensure an optimal state trajectory for the LFC system. The auxiliary feedback controller is designed to counteract external disturbances so as to control the state of the actual system to be maintained in the tube with the nominal trajectory as the center. Finally, the simulation results of a three-area system with wind power shows that the proposed Tube-RMPC control strategy can not only effectively improve control accuracy, but also enhance the robustness of the system and improve the efficiency of real-time optimization. This work is supported by National Natural Science Foundation of China (No. 51777027) and Key Research and Development of Technology Plan of Jilin Province (No. 20180201010GX).