Classification Algorithms for Pump Control as well as Multi-Objective Optimization for Pump Scheduling

Over or undersupply of water will negatively impact crop health. A Smart Irrigation system is the need of the hour in agriculture technology which can optimize water use and enhance crop health for sustainable and precision farming promoting higher yields. IoT and AI-driven systems are more efficient than a simple automated system by reducing water and power wastage. Our system leverages IoT devices such as capacitive soil moisture sensors and DHT-11 sensors. Our system predicts pump activation status (binary classification: pump on/off) based on environmental features. It Optimizes water usage by accurately determining the necessary conditions for pump activation, thereby reducing over-irrigation and conserving resources. We used DNN with 3 layers with a feed-forward neural network. The result from Deep Neural Network shows an accuracy of 99% with an F1 score of 99.5% while a precision of 99.7% with a recall of 99.3% which predicts the optimal parameters for pump scheduling. Similarly, the LSTM model achieved 99.5% accuracy with an F1 score of 99.5% while precision is 99.6% and recall was 99.3% and the CNN model resulted with 86% accuracy and an F1 score of 84.4% with a precision of 97.4% and recall of 74.5%. Similarly, the RNN model resulted in 86% accuracy with an F1 score of 85% precision of 92%, and recall of 80% whereas the CNN + LSTM hybrid model resulted in 87% accuracy and an F1 score of 86.7% with 90.6% of precision and recall of 83%.