A three-layer architecture with network clustering topology optimization for large-scale distributed wireless sensor networks is presented. By using the sensor sleep-scheduling scheme and considering many more clustering numbers, an optimal algorithm for multiple cluster heads based on minimally rigid graph is proposed. It is shown that the optimized clustering network can achieve balanced energy consumption while keeping good connectivity and robustness for the network. Simulation results show that, by adopting the proposed algorithm, the node death phenomenon can appear later than adopting some other related algorithms. It is more advantageous to realize the network load balancing as a whole. And by adopting the proposed algorithm, nodes in the network can survive longer, which shows that the proposed algorithm can prolong the network lifetime.