Self-tuning PID controller based on generalized minimum variance evaluation

The present work proposes a self tuning PID regulatory control method based on generalized minimum variance evaluation. The proposed one is categorized by implicit self-tuning control which directly tunes the PID gains without identifying the plant model. Most of implicit self-tuning control approach updates control parameters by evaluating the tracking error. To the contrary, the proposed one employs the cost function which evaluates the variance of the generalized output which consists of sum of weighted input and output measurements. In addition, the proposed one realizes self-tuning PID gains in regulatory control by routine operation data generated by stochastic disturbance while keeping the reference signal a constant value. In the proposed method, the PID gains are updated at every sampling time so that the cost function consisting of on-line input and output measurements is minimized. The paper introduces the recursive least square (RLS) method with a forgetting factor to the cost function evaluating the variance of generalized outputs. The normalized signal for the self-tuning algorithm is employed in order to improve the stability of the self-tuning control systems. The efficiency of the proposed method is demonstrated through a numerical simulation.