Abstract: The WSNs（Wireless Sensor Network） sensor technology is used to make the environmental parameters of the greenhouse, and it affects the normal working state of the terminal node sensor, which affects the precise decision of intelligent agricultural, as the sensor is distributed in large quantities and the disturbance of the data is caused by the disturbance of data redundancy and data error. In order to solve these problems, an abnormal data detection algorithm based on generalized regression neural network（GRNN） is proposed. The 300 groups of environmental measurements were carried out as a training parameter and 150 group parameters as experimental data. Comprehensive compare the results of the GRNN neural network, the PNN neural network,the traditional BP neural network in the performance indexes of the three lateral surface performance in the data prediction results, the accuracy and the running time. The experimental results show that the GRNN neural network algorithm is accurate and fast, and it is of great significance to the fine details of the crops, and has some effect on the further development of intelligent agriculture.