Design and Implementation of Deep Learning Based Model Predictive Controller to Automatically Adjust Nutrient of Solution for Hydroponic Crop

Abstract

Smart farming is the future of agriculture sector and brings a new era in agriculture; it enables farmers to increase the production and quality of crops with minimal use of resources. In current scenario land availability decreases enormously, hence soilless hydroponic cultivation is considered as the fastest growing sector of agriculture. However, in hydroponic system it is a very challenging task to manage nutrient for crop. To solve these issues this study was conducted which could control robustly EC and pH of hydroponic solution with the help of deep learning model long short-term memory (LSTM). A model predictive controller (MPC) using LSTM was designed and simulated to control EC and pH in hydroponic farm. The predicted outcome of LSTM was operating time of pH buffer solution pump (T_on_pH) and nutrient solution pump (T_on_EC).  The proposed MPC adjust these operating times to control EC and pH with an RMSE of 0.24 and 0.27s, respectively. Furthermore, proposed system improves the predicting accuracy of T_on_pH and T_on_EC of 77% and 61%, respectively, as compared to fuzzy logic controller. This study provides a smart and efficient way to predict and estimate the optimum value for robustly manage the nutrient as per crop requirements.