Abstract: A household phase recognition technology based on improved genetic algorithm was studied. This method first mines the data of the distribution station area through data processing for algorithm implementation. Secondly, adding detection and judgment steps on the basis of genetic algorithm to improve the algorithm and increase its speed. Finally, establish the correct objective function and conduct modeling tests based on the actual substation structure. The results show that the improved genetic algorithm based household phase recognition technology can effectively distinguish household correlation systems and provide convenient conditions for meter reading systems.