针对船舶ORC系统因强非线性、强耦合特性导致系统难以实现多变量稳定控制的难题，提出一种基于深度确定性策略梯度参数优化的自抗扰解耦控制策略。以系统蒸发压力和过热度为被控量设计两输入-两输出的自抗扰控制器，构建虚拟控制量实现两个控制回路间的解耦，利用DDPG算法对控制器带宽参数进行自适应寻优。通过平台和数字仿真实验，在ORC系统的变工况运行条件下评估控制算法的有效性和可行性。结果表明，DDPG-LADRC控制器在设定值追踪和抗扰动测试中相较于LADRC控制器和PID控制器超调量和调节时间均最小。在解耦性能上，针对蒸发压力设定值阶跃引发的过热度波动，DDPG-LADRC的过热度IAE值相较于LADRC控制器和PID控制器分别降低了19.2%和60.6%，针对过热度设定值阶跃引发的蒸发压力波动，蒸发压力IAE值相较于LADRC和 PID 分别降低了51%和61.4%。

To address the challenges in achieving stable multivariable control for shipboard Organic Rankine Cycle (ORC) systems—arising from their strong nonlinearity and strong coupling characteristics—this paper proposes a decoupling control strategy based on Active Disturbance Rejection Control (ADRC) with parameter optimization via Deep Deterministic Policy Gradient (DDPG). A two-input two-output (TITO) ADRC controller is designed with evaporating pressure and superheat as the controlled variables. Virtual control inputs are constructed to decouple the two control loops, while the DDPG algorithm is employed to adaptively optimize the controller bandwidth parameters. The effectiveness and feasibility of the proposed control strategy are evaluated through numerical simulations and experimental platform tests under variable operating conditions of the ORC system. The results indicate that the DDPG-LADRC controller achieves the smallest overshoot and settling time compared to LADRC and PID controllers in both setpoint tracking and disturbance rejection tests. In terms of decoupling performance, for superheat fluctuations induced by a step change in the evaporation pressure setpoint, the superheat IAE of the DDPG-LADRC controller is reduced by 19.2% and 60.6% compared to the LADRC and PID controllers, respectively; similarly, for evaporation pressure fluctuations triggered by a step change in the superheat setpoint, the evaporation pressure IAE is reduced by 51% and 61.4% compared to the LADRC and PID controllers, respectively.

国家自然科学基金 (52371334)；湖北省自然科学基金和发展联合基金重点项目（2023AFD185）