Using remote data analysis to estimate smart elec2 tricity meters (SMs) and detect SMs’ anomaly has aroused con3 siderable discussion in power industry, because its lower cost and 4 higher efficiency compared with the traditional field calibration. 5 However, the trouble of lacking topology and parameter of 6 the distribution grid and the ill-posed problem in SMs’ error 7 estimation and anomaly detection (AD) are not well-resolved in 8 most energy-conservation-theorem-based SM AD methods. This 9 article presents a sorted Top-N AD mechanism to generate a list 10 of suspicious anomalous SMs. The error estimation model (EEM) 11 only using SMs’ electricity consumption data is investigated. The 12 truncated singular value decomposition regularization with the 13 L-curve optimization (TSVD+L) method is proposed to address 14 the model’s ill-posedness. Three data processing modes, namely, 15 one-pot mode, segmentation mode, and sliding window technique 16 (SWT), are suggested to obtain multiple calculation results for 17 SMs’ error comprehensive evaluation. The top N% SMs in error 18 sequence are proposed for onsite calibration instead of full inspec19 tion. The effectiveness and practicality of the proposed method 20 are verified through both the simulation case and practical distri21 bution network application. The results show that the proposed 22 method has higher accuracy in SMs’ AD, compared with the 23 ordinary least-squares (OLS) method, the recursive least-squares 24 (RLS) method, and the Tikhonov regularization (Tik) method.