Aiming for the low accuracy and difficulty of predicting solar cell life by using soft failure mode, a new method is proposed to obtain solar panel spectrum by using terahertz spectrometer. Based on the cuckoo algorithm, the study predicts the cell's remaining life by applying Particle Swarm Optimization-Support Vector Regression(PSO-SVR) algorithm and finally employs the ultraviolet acceleration test to verify the prediction results. It turns out that the method is applicable to predict the remaining life of solar cells with different levels of loss. Compared with other algorithms, the technique works better on the life prediction of traditional silicon solar panels and GaAs solar cells, and the accuracies are up to 98.92% and 92.86% respectively.