Addressing the degradation in control performance caused by the mismatch between simulated and real flight environments in aircraft models, this paper proposes a bandwidth parameter optimization method for extended state observer based on path integral (PI-ESO). We employ a path integral mechanism to online optimize the bandwidth parameters of an Extended State Observer (ESO), thereby dynamically adjusting observer performance in real-time. Multiple candidate bandwidth values are generated via a parallel exploration strategy. A reward function, designed by combining observation errors and parameter regularization, adaptively selects the optimal bandwidth to balance convergence speed against noise suppression capability. Then, we applied this method to the pitch channel of the aircraft and verified through observing the angular velocity of the pitch channel that this method can accurately estimate system disturbances.