High performance proportional valves are commonly utilized for precise control of aircraft actuators to ensure flight safety. Fault diagnosis play a crucial role in maintaining equipment reliability. However, traditional diagnostic methods using vibration signals face challenges such as inability to directly measure failure points, fault characteristics being influenced by sensor position, and large data processing volume, limiting engineering application scope. This paper introduces a valve spool wear fault diagnostic method based on energy loss model and data hybrid drive, leveraging throttling loss characteristics. An energy loss failure mechanism model, incorporating differential pressure and flow rate of valve port, was established, and experimentally verified. This method effectively addresses the limitations of vibration sensors in diagnosing aviation hydraulic systems. By combining particle swarm optimization algorithm with deep extreme learning machine, the number of neurons, weight distribution, and data set ratio are rapidly optimized, reducing data processing complexity and enhancing diagnostic efficiency. Comparative tests demonstrate a significant increase in average diagnostic accuracy rate, with more than 8% improvement after integrating energy loss information and particle swarm optimization, reaching over 98%. The proposed method exhibits superior performance in average diagnostic accuracy rate and stability compared to other methods and can be served as a valuable reference for predicting faults in aviation hydraulic valves.