Detection and Correction of Defective Relative Humidity Data Collected from the Greenhouse Environment Using Nested Kalman Filters with Standard Deviation Analysis

This study proposes a method for detecting and correcting defective data read from drifted humidity sensors used in the greenhouse, leveraging a nested Kalman filter approach with two components: an inner Kalman filter and an outer Kalman filter. The inner Kalman filter is built using data from neighboring sensors. It serves as the prediction for the outer Kalman filter, where its measurement is the data from the sensor of interest (in a defective data case) or from interpolation of neighbor sensors (in a severely defective data case). Analysis based on the standard deviation value of sensor data is used to classify the data into three cases: normal, defect, and severe defect. In the first case, the sensor data are considered trustable. In the last two cases, the data are corrected accordingly using the nested Kalman filters. Experimental results demonstrated significant improvement in accuracy. The mean absolute deviation (MAD) and the root-mean-square error (RMSE) between the corrected data and ground truth reduced when applying the proposed method. The experiments were conducted in two scenarios. The first scenario involves a sensor of interest that exhibits all three data cases: normal, defective, and severely defective data. The second scenario focuses on a sensor that mainly experiences two cases: normal and defective. The MAD decreased by approximately 75.71% and 50.93% in the former and the latter, respectively. Likewise, the RMSE dropped by about 66.41% and 41.80%, respectively. Therefore, the proposed method provides an effective and reliable technique for defective data detection and correction in humidity sensors where accurate and stable humidity measurement is critical.