基于UPF算法的涡旋压缩机压缩气体的温度预测研究

Research on Temperature Prediction of Compressed Gas in Scroll Compressor Based on UPF Algorithm

针对涡旋压缩机运行时腔体内气体温度变化的预测问题,提出一种基于粒子滤波(PF)算法的涡旋压缩腔内气体温度的预测方法。但是传统的PF算法存在粒子退化现象,使粒子集无法准确表示实际后验概率分布,导致预测精度降低。采用无迹卡尔曼滤波(UKF)算法改进PF算法的建议分布函数,抑制了粒子权重退化。通过推导计算压缩介质温度的数学模型,建立温度变化的状态方程,利用UKF算法、PF算法和改进后的无迹粒子滤波(UPF)算法对温度进行预测。预测结果表明UPF算法的预测估计误差平均差为2.74%,预测估计标准差为2.73%,其预测精度和稳定性远高于PF算法和UKF算法。

Aiming at the prediction of gas temperature change in scroll compressor chamber during operation, a prediction method based on particle filter(PF)algorithm was proposed.However, the traditional PF algorithm has particle degradation, which makes the particle set unable to accurately represent the actual posteriori probability distribution, resulting in reduced prediction accuracy.In this paper, the unscented Kalman filter(UKF)algorithm is used to improve the proposed distribution function of PF algorithm and suppress the particle weight degradation.By deducing the mathematical model to calculate the temperature of compression medium, the state equation of temperature change is established, and the temperature is predicted by UKF algorithm, PF algorithm and the improved undetected particle filter(UPF)algorithm.The prediction results show that the UPF algorithm has an average error of 2.74% and a standard deviation of 2.73%,and its prediction accuracy and stability are much higher than PF algorithm and UKF algorithm.