Polynomial Regression Calibration Method of Total Dissolved Solids Sensor for Hydroponic Systems

Abstract

Smart hydroponic systems have been introduced to allow farmers to monitor their hydroponic system conditions anywhere and anytime using Internet of Things (IoT) technology. Several sensors are installed on the system, such as Total Dissolved Solids (TDS), nutrient level, and temperature sensors. These sensors must be calibrated to ensure correct and accurate readings. Currently, calibration of a TDS sensor is only possible at one or a very small range of TDS values due to the very limited measurement range of the sensor. Because of this, we propose a TDS sensor calibration method called Sectioned-Polynomial Regression (Sec-PR). The main aim is to extend the measurement range of the TDS sensor and still provide a good accuracy of the sensor reading. Sec-PR computes the polynomial regression line that fits into the TDS sensor values. Then, it divides the regression line into several sections. Sec-PR calculates the average ratio between the polynomial regressed TDS sensor values and the TDS meter in each section. These average ratio values map the TDS sensor reading to the TDS meter. The performance of Sec-PR was determined using mathematical analysis and verified using experiments. The finding shows that Sec-PR provides a good calibration accuracy of about 91% when compared to the uncalibrated TDS sensor reading of just 78% with Mean Average Error (MAE) and Root Mean Square Error (RMSE) equal to 59.36 and 93.69 respectively. Sec-PR provides a comparable performance with Machine Learning and Multilayer Perception method.