Abstract: In view of the need of accurately grasping the characteristics of the user’s requirement for power consumption in the process of the demand side management（DSM）,these characteristics are finely mined with respect to the dimensions of time,genre and response.The power consumption characteristics are finely analyzed from multiple dimensions by combining the advantages of the system clustering and fuzzy clustering,introducing and improving the secondary clustering algorithm,with the corresponding load characteristic vectors extracted for the clustering analysis.The results show that the improved algorithm is simple and effective,and capable of mining more effective information for the demand side as well as selecting potential users appropriate for DSM.It will provide specific information,such as outage capacity,outage time,and assist to estimate risks and benefits for the DSM subject.