Abstract: To address the poor search capability and tendency of Sand Cat Swarm Optimization (SCSO) to fall into local optima, an improved algorithm named SSCSO based on Quadratic Interpolation Optimization is proposed. A good point set is employed for population initialization to enhance diversity, while an adaptive tangent flight strategy is introduced in the attack stage to improve global search ability and accelerate convergence. In the search stage, the Quadratic Interpolation Optimization (QIO) strategy is integrated to generate candidate solutions and expand the exploration space. In addition, a sparrow alert mechanism is incorporated to further improve convergence performance. The effectiveness of the proposed algorithm is verified using CEC2017 benchmark functions through convergence curves, box plots, and Wilcoxon rank-sum tests. Finally, SSCSO is combined with a hybrid kernel extreme learning machine for pantograph-catenary fault detection, achieving an accuracy of over 98%.