With the widespread deployment of smart meters and the increasing complexity of electricity usage patterns, massive volumes of load data are now available for analysis. This paper proposes a novel method for representing load curves using least squares B-spline (BS), coupled with a robust K-means clustering algorithm RKMOR to extract typical daily load patterns. The BS method reduces the dimensionality of raw load data by fitting smooth spline curves with a small number of control points, preserving the shape structure of load profiles while achieving effective compression. To improve robustness against anomalous load behaviors, we employ the BS_RKMOR algorithm that incorporates outlier removal and median-based center updates. Experiments on the real electricity load dataset of UCI demonstrate that the proposed method achieves better clustering quality than traditional the K-means model.