Special medical supplies such as blood and vaccines have strict temperature requirements in medical cold storage. In order to improve the stability of the temperature monitoring system in the medical cold storage and solve the problem of low measurement accuracy caused by node failure or abnormal data interference, we propose an algorithm based on classification adaptive estimation (CAEWF). The algorithm first monitors and classifies the data collected by the nodes, and uses the mutual support matrix between the data to filter out the validity abnormal data and transmit them to the cluster head. Then use the classification adaptive estimation weighted fusion algorithm to estimate the weighted fusion of the data. The simulation results show that the accuracy of the CAEWF algorithm is better than the arithmetic mean and batch estimation fusion algorithm, which can meet the temperature accuracy requirements of medical cold storage.