- Carbon Trading
- Renewable Energy
- Waste Management
- All Categories
In the wake of a devastating dam overflow in August 2020 that incurred damages exceeding USD 76 million, researchers at Pohang University of Science and Technology (POSTECH) in Korea have pioneered a groundbreaking study on how AI is aiding in the smooth running of dams, enhancing dam management and preventing such catastrophic incidents.
Led by Professor Jonghun Kam and PhD candidate Eunmi Lee from the Division of Environmental Science & Engineering at POSTECH, the research team recently published their findings in the prestigious Journal of Hydrology. The study addresses Korea’s dam challenges, particularly in the Seomjin River basin, which experienced a damaging overflow due to prolonged drought and intense rainfall.
Korea, heavily reliant on dams for water management during the summer precipitation peak, has seen increased climate-related uncertainties, such as typhoons and droughts. To overcome these challenges, the research team turned to artificial intelligence, employing deep learning techniques to analyze and predict dam operation patterns.
The team focused on developing an AI model capable of predicting operational patterns for dams in the Seomjin River basin, explicitly emphasising the Seomjin River Dam, Juam Dam, and Juam Control Dam. Their aim was to predict dam water levels and understand the AI models’ decision-making processes.
Utilizing the Gated Recurrent Unit (GRU) model, a deep learning algorithm, the researchers trained their AI model on extensive big data from 2002 to 2021, incorporating variables such as precipitation, inflow, and outflow data as inputs and hourly dam levels as outputs. The resulting analysis demonstrated remarkable accuracy, boasting an efficiency index exceeding 0.9. The groundbreaking study showcases how AI aids in the smooth running of dams by accurately predicting operational patterns and enhancing overall efficiency.
In addition to predictive capabilities, the team focused on creating explainable scenarios to understand how the trained GRU model responded to input changes. The researchers gained insights into the model’s decision-making processes by manipulating input variables, such as precipitation, inflow, and outflow.
The study, aimed at understanding how AI is aiding in the smooth running of dams, unveiled that changes in precipitation had a negligible impact on dam water levels. At the same time, variations in inflow significantly influenced water levels. Notably, the identical change in outflow resulted in different water levels at distinct dams, highlighting the AI model’s remarkable ability to grasp the unique operational nuances of each dam.
Professor Jonghun Kam emphasized, “Our examination delved beyond predicting the patterns of dam operations to scrutinize their effectiveness using AI models. We introduced a methodology to indirectly understand the decision-making process of AI-based black box models determining dam water levels.” The groundbreaking research signifies a significant leap forward in leveraging technology, where AI aids in dams’ smooth running, enhancing their overall efficiency and resilience.
He added, “Our aspiration is that this insight will contribute to a deeper understanding of dam operations and enhance their efficiency in the future.” The pioneering study opens new avenues for integrating AI in critical infrastructure management, showcasing its potential to mitigate risks and enhance resilience in the face of evolving climate challenges.