For the external perturbation and obstacle avoidance problems of the electric power inspection wheeled mobile robot, a model predictive control strategy based on the dilated state observer is proposed. The trajectory tracking control is executed using nonlinear model predictive control (NMPC). Real-time uncertainty estimation is achieved by implementing an extended state observer, estimating system and external disturbances, and incorporating the disturbance term into the NMPC prediction model. Obstacle avoidance constraints are also integrated into the trajectory tracking process to eliminate the need for local path planning, thereby enhancing efficiency of the inspection robot system. Additionally, Kalman filtering is employed to a priori estimate the speeds of dynamic obstacles, improving the robustness of the obstacle avoidance strategy. Simulation results demonstrate that this approach effectively manages input constraints, unknown disturbances, and model uncertainty, while the extended state observer design offers reliable uncertainty estimation, leading to improved trajectory tracking effectiveness and obstacle avoidance performance.