Design and implementation of greenhouse sensor fault self-identification system based on edge computing

Wireless sensor data provides decision basis for intelligent environment regulation and control, reasonable and accurate data is the prerequisite for correct decision making, and real-time detection of abnormal data is crucial. In response to the problems of low detection accuracy and efficiency of traditional algorithms used for static data anomaly detection, uploading data to the cloud computing center for analysis, increasing the transmission pressure of bandwidth and control decision feedback time, a new method based on edge computing and data fusion is proposed. The multi-modal sensing fusion algorithm was used to detect anomaly data, simulate and analyze the actual abnormal data sets of agricultural greenhouse such as humidity, temperatureand light. The sliding window method was used to deal with the infinite data flow, and calculate the variance of single sensor and multi-sensor data, and the correlation coefficient between multi-sensor data, so as to optimize the key structural parameters. The results show that the model can detect the abnormal data of the sensor, the fault recognition rate of single-node multi-sensor is 82. 5%, and the fault recognition rate of multi-node multi-sensor is 72. 5%. The aggregated data uploads can reduce the frequency of transmission, saving 30% of data traffic in a single pass, reducing server pressure and data transmission latency. And the results of this paper can provide useful references for solving the problem of abnormal greenhouse sensor data and the application of edge computing in greenhouse environmental equipment.