XLPE insulation aging affects the operation of the power system. Based on the insulation state detection project, this paper proposes a PLS aging time prediction model based on multiple feature detection quantities. Aiming at the small data collected and the multi-collinearity problem in the model, the least squares support vector machine (LSSVR) is introduced to optimize the model principal component score vector. Then, the LSSVR-PLS aging time model is established utilizing the new score vector. Finally, the nonlinear processing ability is compared and tested by a T test and the 110 kV XLPE cable samples in a certain area of Hangzhou is considered. It is shown that the improved model is suitable for the processing of small sample data of cable detection, which can eliminate the multi-collinearity problem existing in the original model and achieve a higher prediction accuracy. The proposed research provides an important guiding significance for the cable operation and maintenance and the transformation of power grid.