Traditional electrical load analysis methods have disadvantages on gaining the representative load profile of a single customer as well as evaluating similarity between the different load patterns. This paper proposes a new combinational load analysis method by using the density based clustering to gain the representative load profile of a single customer and clustering different customers into different clusters according to the cosine similarity function. After applying this method to a real power system’s data set, the result is compared with the one using traditional analysis method and the applications in demand side management (DSM) are discussed. The results show that this new method could effectively find the representative load profiles and recognize the different load profiles, and thus confirm the value of this method in DSM applications.