New approach to monitoring a wastewater irrigation system controlled by the artificial neural network (ANN)

Abstract

In this paper, a cost-effective and scalable wireless sensor network (WSN) system has been developed, targeting monitoring and observation applications. The system specifically emphasises wastewater control and management in the context of irrigation practises. An artificial neural network (ANN) technique, incorporating the Well Node approach, has been employed to process a one-year dataset collected from sensor nodes in an actual wastewater treatment station in Morocco. Thus, the acquired data include a range of parameters, such as electrical conductivity (EC), total dissolved solids (TDS), chlorine levels (Cl), nitrogen content (N), hydrogen carbonate concentrations (HCO3), and pH values. The proposed hybrid system incorporates the Kohonen self-regulating characteristic map for data validation and reconstruction, along with multilayer perceptrons for sample prediction. The selected ANN demonstrates impressive performance, with recognition rates exceeding 95% and an F1 score of approximately 96.81% in terms of wastewater reusability. In particular, the system exhibits robustness in handling uncertainties, deviations, and low information content in input data, allowing for informed decisions on usage or reprocessing through its self-learning capabilities. The study highlights the adaptability of the WSN system to different farming requirements and geographical areas, contributing to improved resource management, enhanced water utilization efficiency, and protection of plants, soil, and the environment. The research findings advance the understanding of ANN architecture and its unique modifications for effective wastewater control. The results also emphasise the effectiveness of the system during the adaptation and implementation phases, offering valuable insights for algorithmic improvements.