A dynamic optimization strategy based on a hybrid of pseudospectral method and direct shooting method is proposed for the problem of fast optimization of the variable load control process of CHP units. Firstly, the dynamic optimization proposition of the optimal variable load control process of a CHP unit is constructed and discretized into a nonlinear programming problem. Then, the feasible solutions corresponding to the state variables and control variables are solved using the Gauss pseudospectral method, and the serial optimization strategy from feasible solutions to near-optimal solutions is used. Finally, the control variables are discretized at the nodes as initial values for the direct shooting method, and then the sequential quadratic programming (SQP) algorithm is used to solve the exact optimal solution. Simulations were conducted on three cases of a 300 MW CHP unit, and the results showed that the proposed method reduced the calculation time by approximately 32%, 36%, and 53%, respectively.