Exploring Machine Learning Techniques for Smart Drainage System

With climate change and volatile weather patterns, smart drainage system becomes critical in flood management to prevent possible life and property lost in highly urbanized cities, like Singapore. In smart drainage system design, an innovative approach to integrating machine learning and Internet-of-Things (IOT) can provide data-driven solutions to mitigate the risk of flash flood. This study explores various machine learning techniques to model the nonlinear relationship between rainfall and drainage water level nearby and predict the water level with lead time up to 10 minutes in advance. Through the analysis, it is found that the Artificial Neural Network model is superior in modeling the nonlinear relationship during the training stage, while the Random Forest and SVM-radial models are robust and offer better accuracy in drainage water level prediction.