Water quality prediction using LSTM with combined normalizer for efficient water management

Predicting water quality is a significant area of study in the field of smart water technology, since it may provide valuable assistance in managing and mitigating water pollution. Due to the increasing global population and the need for effective methods of agriculture and irrigation, there is a continuous increase in the demand for water, which lead to a scarcity of water resources. Consequently, smart water management systems have been created with the objective of enhancing the effectiveness of water management. Nevertheless, conventional water quality prediction models mostly use data-driven approaches and only depend on diverse sensor data. In recent research, deep learning algorithms have been extensively used for water quality prediction due to their robust ability to map highly nonlinear connections while maintaining acceptable computational efficiency. Therefore, the LSTM-CN model presented in this paper integrates the benefits of three normalisation calculation methods: z-score, Interval, and Max. This allows for adaptive processing of multi-factor data while preserving the data's inherent characteristics. Ultimately, the model collaborates with the codec to learn the data's characteristics and generate accurate prediction results. When compared with existing water quality prediction methods in terms of various parameters the proposed LSTM-CN methods achieves 99.3% of accuracy,95% of precision, 93.6% of recall, 18% of MSE and 11.45% of RMSE.