【摘要】 低温度场广泛存在于生物医疗、低温加工等过程中，热传导和传质等导致系统存在滞后特性。针对低温度场调控系统中时滞特性导致的系统超调、振荡等问题，在Smith预估结合PID控制基础上，引入了BP神经网络，实现了控制器增益的自适应调整。针对传统神经网络学习算法增益调整速度慢、结果不稳定等问题，在充分考虑系统的动态模型下，提出了基于遗传算法的神经网络权值优化方法，实现了控制器增益的快速稳定调整。系统仿真结果表明，较PID-Smith控制、NNPID-Smith控制等，在低温度场时滞系统调控中超调较小，调整时间短，有效改善了低温度场调控过程中的系统稳定性。更多还原

【Abstract】 Low temperature fields are widely present in processes such as biomedicine and low temperature processing, where heat conduction and mass transfer lead to system hysteresis.To address the issues of system overshoot and oscillation caused by the time delay characteristics in low temperature field control systems, a combination of Smith predictor and PID control is employed, along with the introduction of BP neural network for adaptive adjustment of controller gain.Considering the slow learning speed and unstable results of traditional neural network learning algorithms, a genetic algorithm-based neural network weight optimization method is proposed, which achieves rapid and stable adjustment of controller gain by fully considering the dynamic model of the system.Simulation results demonstrate that compared to PID-Smith control and NNPID-Smith control, this method exhibits smaller overshoot and shorter adjustment time in the control of low temperature field time-delay systems, effectively improving the system stability in the process of low temperature field control.更多还原