In order to explore the impact of different environmental features (temperature, humidity, etc.) on the operation error of smart meters, it is necessary to cluster smart meters operating in different areas according to environmental features. In the existing clustering algorithm research of intelligent smart meters, it is based on operation data or load curve, but it is lack of using environmental factors to cluster. Therefore, this paper proposes the principle of environmental information extraction, which can effectively reduce the data dimension and improve the calculation efficiency. The initial value optimization K-means algorithm is proposed, which improves the traditional K-means algorithm on initial value selection and clustering center moving rules, making it more suitable for clustering intelligent energy meters based on environmental features. Finally, the simulation results show that the accuracy of this method is 17.7% higher than other algorithms, and the calculation time is reduced by 0.16 seconds on average.