In order to reduce the data transmission and energy consumption of sensor nodes and improve the broadband utilization and life cycle in real-time cold chain logistics monitoring system for agri-food based on wireless sensor networks, a clustering fusion method based on arithmetic mean and batch estimation for the cold chain temperature monitoring of agricultural products was proposed, Firstly, the mistake errors of the collected data was eliminated, and then the data which were sent from the cluster member nodes were merged by the mean and the batch estimation method. Secondly, the cluster head node used the adaptive weighting algorithm to further analyze the fusion data of the member nodes. The experimental results showed that the network lifetime of the cold chain monitoring system based on the data fusion method was 34.2% higher than that of the traditional method, and the stability period was 11.4% higher than that of the traditional low power adaptive cluster clustering protocol. Compared with the traditional arithmetic average method, the accuracy of data fusion was improved by 7.6%, the system energy consumption was decreased by about 32.5% per round, which can not only reduce the influence of redundancy and less reliable data on the measurement results, but also reduce the unnecessary data transmission loss, as well as reduce the cost of cold chain logistics and improve the degree of informatization of cold chain logistics to a certain extent.