Reliability Improvement of Rainwater-Drainage System Using IoT and AI

Japanese-government data on the degree of damage by flood type indicates that inland floods cause the highest recorded damage every year, without the occurrence of large-scale damage due to levee breaches. Therefore, research on improving rainwater-drainage systems is important for damage prevention for inland flooding A conventional rainwater-drainage operation scheme is called “water-level control,” where a pump is started-up/stopped when the drainage-water level of the pump-well reaches a given threshold. To prevent inland flooding caused by heavy rain, the drainage-pump operator sometimes manually manipulates the water-level control and starts-up the pump before the rainwater level reaches the threshold. Notably, conventional rainwater-drainage system pumps have never been operated by applying rain-forecast big data, which is one of the factors that make rainwater-drainage systems unreliable. Our research focuses on improving the reliability of an autonomous system using IoT and rain-forecast big data. This paper describes the design of our improved rainwater-drainage system and discusses the verification results of the system's inland flood-damage prevention through drainage experiments and inland-flood analysis. These verification results show that even an improved system is insufficient for reducing inland flood damage. With further consideration, we suggest that an advanced extreme-heavy-rain prediction system is required. New research is underway to develop an AI system for predicting extremely heavy rainfall. An overview and the plan of the development of this AI system are also discussed.