The present invention discloses a universal non-invasive blood glucose estimation method based on time series analysis, which comprises 5 steps, i.e., data input and preprocessing, features screening, establishment of single-feature model based on time series analysis, multi-feature fusion, and non-invasive blood glucose estimation. During non-invasive blood glucose estimation, new non-invasive test data is inputted, and the blood glucose estimation series is calculated using related feature information obtained through modeling, single-feature model, and multi-feature fusion model. The estimation method provided in the present invention is easy to execute, and can overcome the delay between changes of physiological parameters of human body and changes of blood glucose, and thereby can obtain more accurate non-invasive blood glucose test results. The estimation method is universal and is applicable to different non-invasive blood glucose monitoring methods.