过程控制常用连续模型的直接辨识法及应用

针对工业过程中最常用的一阶加滞后、二阶加滞后、二阶加零点、二阶加零点及滞后、积分惯性加滞后等环节，给出了基于阶跃响应的连续模型参数直接辨识算法。由传递函数的拉普拉斯逆变换式和对象阶跃响应的采样数据构成模型参数回归表达式，用最小二乘法或辅助变量法直接辨识对象的连续时间传递函数模型参数。仿真与实际应用结果表明，该算法提高了模型辨识精度，减小了对过程的扰动，并且对输出测量噪声不敏感，鲁棒性强，容易编程实现，可提高实际PID控制器参数整定质量。

Direct Identification Algorithm of Continuous Models and Its Application to Process Control

A direct algorithm is presented based on step response for identification of linear continuous time processes that are commonly used in process control, such as first-order plus dead-time, second-order plus dead-time, second-order plus zeros and dead-time, integral plus delay. The regression equations of model parameters are derived from Inverse Laplace transform of transfer function and sampled data of step response. The ordinary least squares or the instrumental variable method are applied to direct identification of model parameters. The simulation results show that the model identification is improved and the resulting parameter estimation is robust in the face of measurement noise in the output. The method is easy to realize and can be applied to improve tunning of PID controllers.