The invention belongs to the technical field of a wireless sensor network and provides a secondary clustering algorithm for the wireless sensor network based on a two-layer model system. According to the algorithm, a traditional secondary clustering algorithm is improved from two aspects of constructing prediction models and detecting near node data correlation, a new model construction and data correlation detection method is employed. The method mainly comprises the steps of constructing double prediction models through utilization of collected samples; and carrying out secondary clustering on intra-cluster nodes based on correlation of the double prediction models. The algorithm has the beneficial effects that compared with an existing duty cycle based data collection method, the algorithm has the advantage that the information of an area covered by a cluster is extracted more sufficiently, and compared with an existing partial data correlation based clustering method, the algorithm has the advantage that the life cycle of a whole network is prolonged on the premise of not reducing the collection precision.