The invention discloses a label clustering technology-based power consumption behavior analysis method for large power customers, which comprehensively considers customer power consumption characteristics and influence factors on the basis of massive customer archives, loads and electric quantity data, establishes a customer power consumption behavior label library, and adopts k-means algorithm tocarry out Label clustering, and achieve power utilization behavior portraits of different types of power customers. The selected power utilization behavior tags are reasonable and effective, the adopted clustering algorithm is remarkable in effect, and powerful data support can be provided for power companies to grasp the power utilization habits of customers, mine the demands of the customers and improve the service level.