<p>Hydraulic systems play an essential role in many industrial applications. However, traditional control methods such as PID (proportional–integral–derivative) control have problems such as low accuracy, slow response, and poor adaptability. This paper proposes an intelligent control and optimization framework based on deep neural networks to solve the control difficulties encountered by hydraulic systems in practical applications. First, the working data of the hydraulic system are collected in real time through multiple sensors and preprocessed. Then, a deep neural network is used to model the hydraulic system nonlinearly to capture its complex input–output relationship and achieve precise prediction. Reinforcement learning is applied to optimize the control strategy, and a deep Q network model is used for adaptive control. Under variable working conditions, reinforcement learning continuously adjusts control parameters (such as pump speed and valve opening) through the interaction between the intelligent agent and the environment to achieve optimal system performance. In addition, a multi-objective optimization strategy is adopted to comprehensively consider control accuracy, response speed, and energy efficiency and balance them through a comprehensive reward function. Experimental results show that the proposed method significantly improves the hydraulic system’s control accuracy, response speed, and energy efficiency, which is better than traditional control methods. The response time from the change of the system input signal to the stable output of the system is 2.4&#xa0;s, and the system energy efficiency ratio is less than 0.7 under high load. The research in this paper provides new ideas for the intelligent control and optimization of hydraulic systems and has broad application prospects.</p>