The coordinated control system of ultra-supercritical (USC) units often faces multi-source uncertain disturbances, such as coal quality variations and load fluctuations. Those factors weaken the system’s response and stability. In addition, the boiler’s thermal inertia and the turbine’s mechanical inertia add complexities to the control task. To address it, a cascade adaptive phase optimized active disturbance rejection control (CAPOADRC) is proposed. An adaptive phase optimized extended state observer (APOESO) is constructed to enhance the disturbance estimation. To overcome the bandwidth limitations of ESO, a cascade structure is introduced, the first layer estimates partial disturbances and filters the system output, while the second layer utilizes the estimation information from the first layer to get the remaining disturbances. The CAPOADRC can estimate the total disturbance rapidly and accurately. Numerical results validate the proposed method in suppressing system disturbances and improving response speed. It may provide a feasible solution for the coordinated control of USC units.