Abstract: In multilevel heterogeneous wireless sensor networks, the initial energy of nodes are random distributed in a certain range, load balancing and energy efficiency are the significant challenges of clustering algorithm for energy heterogeneous networks. Current distributed clustering algorithm is mainly designed for homogeneous or two-level heterogeneous networks, and it is hard to implement load balancing when the nodes energy represents multilevel heterogeneity. So a load balance clustering algorithm LBCA (load balance clustering algorithm) for multilevel energy heterogeneous sensor networks is proposed. The algorithm select cluster head nodes and implements load balance according to the condition of energy distributing, and could prolong the stability period. In the process of cluster head selecting, when the energy is balanced in local area, the nodes which have the lower average communication cost are prior to be the cluster-head nodes, and it is propitious to decrease the total energy cost of local area. When the energy is imbalanced in local area, the high residual energy nodes are prior to be the cluster-head nodes, and it is propitious to implement load balancing. LBCA is compared with primary distributed clustering approach. The simulation results show that in multilevel energy heterogeneous networks, LBCA could better implement load balance and prolong the stability period.