Implementing sensorless control is necessary to reduce the system volume of linear oscillatory machines (LOM) used in linear compressors and achieve efficient and reliable operation. The existing piston stroke observers have low observation accuracy and are susceptible to DC components, resulting in a decrease in system compression performance or cylinder collision risk. Therefore, this paper designs an improved high-precision piston stroke observer for linear oscillation machines based on a high-order generalized integrator (HOGI).

Firstly, a theoretical analysis is conducted on traditional back electromotive force integration, low-pass filter (LPF), and second-order generalized integrator (SOGI), elucidating the existence of integral saturation problems in back electromotive force integration, amplitude attenuation, and phase shift problems in LPF. SOGI performs slightly better than the previous two but still cannot eliminate the DC component. When operating at low resonant frequencies or in systems with large DC components, SOGI is no longer applicable. Secondly, in response to the shortcomings of traditional integrators, this paper adopts HOGI as a piston stroke observer. This method can eliminate the DC component, and no DC bias exists in the observed stroke signal. The paper also uses the forward Euler method to derive the digital implementation method of HOGI. Finally, experiments are conducted to compare SOGI and HOGI. The experimental results show that the piston stroke observed by HOGI is more accurate than SOGI without additional DC bias. Furthermore, when an additional 0.2 A DC bias is added, the piston stroke average offset observed by SOGI at the given value of 5 mm, 6 mm, and 8 mm is 1.367 5 mm, 1.365 mm, and 1.351 5 mm, respectively. The piston stroke observed by HOGI is unaffected by DC bias. Therefore, the piston stroke observer with HOGI is suitable for occasions with serious DC disturbance.

The contributions of this paper are as follows. (1) Based on traditional SOGI, an improved HOGI piston stroke observation structure is designed. Multiple filtering feedback characteristics are used to eliminate the influence of DC components on stroke observation results, improving the accuracy of the piston stroke observation. (2) The complex frequency domain method is used to analyze the pure integrator, LPF, SOGI, and HOGI. The superiority of HOGI is theoretically proven. (3) Based on the forward Euler method for discretization and digital implementation of HOGI, this method has the advantages of simple calculation and easy implementation.