Abstract: Aiming at the nonlinear problem of the core power model caused by steady-state neutron density at different power levels, generalized predictive control (GPC) is applied to the core power control to realize the automatic control of the core power under variable working conditions. This paper firstly establishes a core power model based on zero-power core model and temperature feedback model. The prediction time domain is designed based on the order of the model, and the model parameters at different power levels are identified online by the least square method with forgetting factor in the GPC correction link according to the input and output data of the system. In order to verify the robustness of the controller,  the reactive disturbance is added at full power smooth operation. The performance of the controller is verified by simulation based on MATLAB platform, and the results show that the GPC designed in this paper can quickly and accurately track the set value when the core is in variable working condition, and can identify the core model parameters of different power levels on-line, and has certain anti-interference ability.