摘 要：组合式空调机组运行过程中，具有大迟延和大惯性的特性，这意味着单一控制系统对控制信号的响应存在明显的延迟和滞后。因此，提出了基于神经网络PID控制器的组合式空调机组温度控制方法。分析组合式空调机组温度阶跃响应变化，以分析结果为基础，借助动态矩阵分析历史温度数据，预测未来时刻空调机组输出温度，为温度控制提供依据。结合BP神经网络和PID调节器，搭建具有参数自适应整定性能的神经网络PID控制器。通过神经网络的自学习不断调整控制参数，最终实现高质量的空调机组温度控制。实验结果表明：该方法温度控制结果表现出的超调量总是不大于0.6%，极大提高了温度控制过程的稳定性。

关键词：BP神经网络；PID控制器；组合式空调机组；温度控制；参数整定

中图分类号：TP183文献标识码：A

A Temperature Control Method for Combined Air Conditioning

Units Based on Neural Network PID Controller

ZHANG Yinzhou

（Beijing Yongxinjiacheng Engineering Technology Co.， Ltd.，Beijing 100043，China）

Abstract：During the operation of a modular air conditioning unit， it has the characteristics of large delay and inertia， which means that there is a significant delay and lag in the response of a single control system to control signals. Therefore， a combined air conditioning unit temperature control method based on neural network PID controller is proposed. Analyze the temperature step response changes of the combined air conditioning unit. Based on the analysis results， using dynamic matrix to analyze historical temperature data， predict the output temperature of the air conditioning unit at future times， and provide a basis for temperature control. Build a neural network PID controller with parameter adaptive tuning performance by combining BP neural network and PID controller. By continuously adjusting control parameters through self learning of neural networks， high quality temperature control of air conditioning units is ultimately achieved. The experimental results show that the overshoot exhibited by this method in temperature control is always not greaterthan 0.6%， greatly improving the stability of the temperature control process.

Key words：BP neural network; PID controller; combination air conditioning unit; temperature control; parameter tuning

组合式空调机组近年来在大型商业建筑、办公楼、医院及数据中心等场所得到广泛应用，实现了对室内环境舒适度与空气质量的维持。但组合式空调机组温度控制是一个复杂的系统工程，涉及多变量、强耦合、非线性和时变等特性，其在高效运行过程中，需要具备精准的温度控制能力，以确保输出温度符合预期要求。尤其在人们对室内环境要求不断提高的背景下，组合式空调机组的温度控制方法成为一项重要研究课题。邹宇航等提出基于模型预测控制的温度控制方法[1]，依托于等效电路法预测机组未来温度变化，并将其看作目标函数输入基于遗传算法的温度控制模型，通过粒子群优化算法求解最佳控制参数，完成空调温度优化调整。……