闭式循环柴油机配氧及其控制策略实验研究

Experimental Study on Neural Control Strategy of Closed Cycle Diesel Oxygen Replenishment

在闭式循环柴油机配氧反馈控制的基础上采用了神经网络前馈控制策略。通过使反馈偏差最小化在线训练神经网络,以网络前馈输出逐渐取代原有的反馈控制,从而在系统运行时以自适应方式始终保证氧浓度偏差最小。该前馈补偿不受配氧压力等时变因素的影响。建立了实验台架,神经网络前馈复合控制由监控计算机和现场反馈控制器共同完成。通过实验研究了神经网络前馈控制的训练过程并验证了该方法的有效性,配氧控制的精度和稳定性得到较大提高,最大动态偏差由原来的14.2%减小到3.1%。

Based on the feedback control, the neural feedforward compensation was applied in oxygen control of closed cycle diesel. The neural network was trained online by minimizing the feedback-error, and feedforward output is used to replace the former feedback control. This makes the self-adaptive controlling maintaining the minimized deviation of oxygen concentration. This control has the advantages due to the less influence from the time variables such as oxygen supplying pressure. The test-rig was set up and the neural feedforward-feedback combined controlling was achieved by using the industrial computer and field PID controller. The network training process was experimental investigated and the effectiveness of the method was evaluated. The accuracy and stability of the oxygen control is improved, for the maximum transient deviation of oxygen control can be reduced from 14.2 percent to 3.1 percent.