The electrical customer clustering is very important for DSM, for electricity price evaluation. The behavior of one electricity customer is considered in load curve, participation in forming the system peak load and the coincidence of customer peak load and system peak load. The clustering the customer basing on these features will be carried out. The calculation of the distance between two customers is the key problem. The Pulsa algorithm will be applied for clustering process. This algorithm has many stages. Each stage consists of some iterations and the radius of clustering will be changed due to the number of objects falling to the cluster. The application is for the customers of HochiMinh city.