<p>In order to address the insufficient robustness performance of traditional centralized PI controllers within the full flight envelope of aero-engines, this paper proposes a centralized PI composite control method based on composite model-reference adaptive compensation. The turboprop engine affected by external disturbances and uncertainties is taken as the research object, the parameters of the centralized PI controller are first designed based on frequency domain approximation method. Then the reference model and the regression filter are introduced to construct the tracking error dynamics equation and the residual signal, and an adaptive rate parameter estimation term is introduced on the basis of the traditional composite model-reference adaptive law, so as to propose a composite model-reference adaptive control law based on covariance adaptive parameter estimation. It is proved through Lyapunov stability theory that when the auxiliary matrix satisfies the full-rank condition, both the dynamic tracking error and parameter estimation error of the system can achieve exponential convergence characteristics. Finally, simulation verification was conducted on turboprop engine, and the results showed that the method proposed in this paper has more significant advantages under non-design conditions, compared with the traditional composite model-reference adaptive law, this method reduces the control performance indicators IAE, ISE, ITAE, and TV of the gas turbine speed <i>N</i>_g control loop by 16.21%, 21.90%, 15.24%, and 1.14%, respectively; and the control performance indicators IAE, ISE, ITAE, and TV of the power turbine speed <i>N</i>_p control circuit decreased by 60.37%, 82.69%, 62.70%, and 51.28%, respectively.</p>