Archives

Time-Series Forecasting and Hyperparameter Tuningwith LSTM Approach for Root Zone Temperature of Nutrient Solution In Indoor DWC Hydroponic


Srivani P, Yamuna Devi C R, Manjula S H,Venugopal K R
Abstract

Environmental and growth parameters make an impact on the plant’s growth in the hydroponic system. Machine learning models can be applied to analyse these parameters and control them intelligently. In a hydroponic system, the Root-zone temperature (RZT) is an essential parameter that affects plant growth, which changes with pH and other environmental parameters. This study forecasts the RZT using Long Short Term Memory (LSTM) model grown in a DWC system for Amaranthus Dibius crop. The network architecture was implemented with three input features and the performance was evaluated by the Root Mean Square Error (RMSE) value. With a time-series data set, the model was able to predict the RZT with a test accuracy score () greater than 0.8 with two-layered stacked LSTM. The model was tuned considering the different combination of hyperparameters on multi-layer LSTM model which showed better performance with higher epochs and increasing the neurons

Volume 12 | 06-Special Issue

Pages: 952-960

DOI: 10.5373/JARDCS/V12SP6/SP20201114