The relationship between households and transformers in low-voltage station areas is frequently changed because of routine maintenance, renovation and upgrading of the distribution network. Since it is difficult to maintain the household-transformer relationship in low-voltage station areas manually, a dynamic identification method for the relationship is proposed using Bayesian inference and spectral clustering technology. First, the Gaussian kernel function and IIR filter equation are used to calculate the voltage data similarity between nodes, and then Bayesian inference is used to construct the time series matrix of node similarity. Finally, the dynamic identification of the relationship can be implemented via the spectral clustering technology. The method can dynamically identify the household-transformer relationship using the latest voltage data along the time sequence direction, which not only can make a quick response to household changing the station area, but also can effectively improve the robustness of the identification in the case of missing data and node input or exit from the station areas. Taking four station areas in Huizhou City, Guangdong Province as an example, the results of the calculation example are good, and this verifies the feasibility and reliability of the proposed method for engineering practice.