李壮举, 王宁, 张晓宇,等.基于贝叶斯优化的二阶线性自抗扰控制参数整定[J].河南理工大学学报（自然科学版）,2026,45(3):85-92.

LI Z J, WANG N, ZHANG X Y,et al.Parameter tuning of second-order linear active disturbance rejection control based on Bayesian optimization[J].Journal of Henan Polytechnic University(Natural Science) ,2026,45(3):85-92.

基于贝叶斯优化的二阶线性自抗扰控制参数整定

李壮举1, 王宁1, 张晓宇1, 魏伟2

1.北京建筑大学 电气与信息工程学院，北京  102616；2.北京邮电大学 自动化学院，北京  100876

摘要:目的 线性自抗扰控制（linear active disturbance rejection controller，LADRC）是一种对模型依赖较小的控制方法，在工业过程中具有良好的应用前景，但其参数调整复杂，导致工业应用中试错成本较高，为此，提出一种可工程应用的LADRC参数整定方法。  方法 首先，通过分析二阶线性自抗扰控制器的等效结构，建立LADRC参数和PID参数之间的等效关系；然后，通过贝叶斯优化获得PID初始参数；最后，通过参数之间的等效结构得到LADRC初始参数。  结果 在Simulink中选取多类典型传递函数进行仿真实验，结果表明，对于多种类型的被控对象，二阶LADRC控制在稳定性和准确性等方面均优于PID控制。利用PLC温度控制系统进行实物系统实验，对被控对象的模型进行辨识，通过贝叶斯优化算法求得PID的初始参数，利用PLC程序实现LADRC控制器，然后分别进行PID控制和LADRC控制的实验，结果表明，在温度阶跃响应中，二阶LADRC控制系统调节时间约80 s，超调量约2.3%，稳态误差很小，ITAE指标好，而PID控制器调节时间约95 s，超调量约1.6%，ITAE指标一般。  结论 提出的基于贝叶斯优化的参数初始化方法能保障系统参数移植的有效性，减少调试工作量，为二阶LADRC广泛应用提供有力的参数调整策略。

关键词:参数整定;贝叶斯优化;线性自抗扰控制;PID控制;PLC

doi:10.16186/j.cnki.1673-9787.2024110009

基金项目:国家自然科学基金资助项目（62371032）

收稿日期:2025/02/17

修回日期:2025/04/07

出版日期:2026/04/28

Parameter tuning of second-order linear active disturbance rejection control based on Bayesian optimization

Li Zhuangju1, Wang Ning1, Zhang Xiaoyu1, Wei Wei2

1.School of Electrical and Information Engineering， Beijing University of Civil Engineering and Architecture，Beijing  102616， China；2.School of Automation， Beijing University of Posts and Telecommunications， Beijing  100876， China

Abstract: Objectives Linear Active Disturbance Rejection Control （LADRC） is a control method that has low dependence on the system model and shows promising application prospects in industrial processes. However， its parameter tuning is complicated， leading to high trial and error costs in industrial applications. To address this issue， an engineering oriented parameter tuning method for LADRC is proposed.  Methods First， by analyzing the equivalent structure of the second order LADRC， the equivalent relationship between LADRC parameters and PID parameters is established. Then， initial PID parameters are obtained through Bayesian optimization. Finally， the initial LADRC parameters are derived using the established equivalent relationship. Results Simulation experiments are carried out on several typical transfer functions in Simulink. The results show that for various types of controlled plants， the second order LADRC outperforms PID control in terms of stability and accuracy. Physical experiments are conducted using a PLC based temperature control system. The model of the controlled plant is identified， the initial PID parameters are obtained via Bayesian optimization， and the LADRC controller is implemented in the PLC program. Then， both PID and LADRC control experiments are performed. The results indicate that in the temperature step response， the second order LADRC system achieves a settling time of approximately 80 seconds， an overshoot of about 2.3%， very small steady state error， and a good ITAE index. In contrast， the PID controller yields a settling time of about 95 seconds， an overshoot of about 1.6%， and a moderate ITAE index.  Conclusions The proposed parameter initialization method based on Bayesian optimization ensures the effectiveness of parameter transfer between different platforms， reduces tuning effort， and provides a powerful parameter adjustment strategy for the widespread application of second order LADRC.

Key words:parameter tuning;Bayesian optimization;linear active disturbance rejection control;PID control;PLC