Abstract: Aiming at the problem that the number of load nodes is large in large-scale distribution network， it will add decision variables of the distribution network planning model and increase planning difficulties when considering all load nodes at a time. This paper puts forward a planning method based on clustering blocks of load nodes. Firstly， using load weight to improve basic k-means clustering algorithm and then load nodes are clustered by the improved algorithm. Secondly， a connection matrix which can represent the upstream and downstream relations of load nodes is established， to speed up current calculation and analysis process about the influence of tie lines location on the downstream load transfer of a fault line. Then the improved minimum spanning tree algorithm is used to solve the main network planning model， which minimizes annual investment cost and operation cost while taking geographical factors into consideration. Finally， tie lines between load nodes blocks are set up， aiming at minimum annual investment cost and outage cost， meanwhile according to different load transfer conditions， different methods to calculate the outage cost are adopted. The example results support the validity and effectiveness of the method proposed in this paper.

Key words: load weight, clustering blocks of load nodes, connection matrix, main network planning, tie line planning