Load estimation for distribution systems using clustering techniques

In electrical distribution systems, except the usual measurements from stations, there are few information about the state of network. As a result, there is at any moment a generalized uncertainty about the power demand conditions and therefore about the network loading, voltage level, power losses, etc. Therefore, it is important to establish accurate the load of nodes from distribution system. An alternative to the settlement based on metered demand is load profile based settlement. In this manner, each node (distribution substation), without appropriate meters, is assigned a representative load profile. For that reason it is necessary to determine typical load profiles. In the first part of the paper, the K-means clustering method is used for typical load profiles determination. In the second part, an improved approach based on the obtained typical load profiles for the load estimation of the electric distribution systems is proposed. To demonstrate the effectiveness of load simulation of the distribution system based on the typical load profiles of the nodes, a practical example is presented.