AI-Enhanced Gross Pollutant Traps: A Smart Approach to River Health and Pollution Control

Abstract

Flooding and river pollution pose significant challenges in Malaysia, exacerbated by the inefficiencies of Gross Pollutant Traps (GPTs), which rely on manual monthly cleaning processes. These conventional methods are inadequate for addressing the dynamic influx of pollutants, particularly during adverse weather conditions. This research proposes an innovative AI-powered framework that integrates logistic regression for weather prediction and Convolutional Neural Networks (CNNs) for real-time garbage classification. By predicting weather patterns and classifying pollutants, this system optimizes GPT maintenance, enhancing its effectiveness and efficiency. The proposed system leverages real-time data from sensors, cameras, and weather forecasts, enabling authorities to implement proactive maintenance strategies based on accurate weather predictions and pollutant types. Logistic regression models forecast adverse weather conditions, while CNNs accurately classify garbage types, allowing targeted GPT cleaning during periods of increased pollutant buildup. The logistic regression model achieved an accuracy of 86.41%, and the CNN model attained a classification accuracy of 79.37%, showcasing strong performance in predicting weather conditions and categorizing pollutants. The integration of AI technologies in GPT maintenance significantly enhances environmental planning, mitigates flooding risks, and improves the accuracy of pollution monitoring. This solution provides valuable insights for decision-makers, helping them allocate resources effectively and maintain sustainable water management practices. In conclusion, the AI-driven system offers a robust and efficient approach to optimizing GPT operations, contributing to better environmental protection and urban sustainability.