In the relentless pursuit of sustainability and efficiency, the water industry is turning to artificial intelligence (AI) to tackle one of its most persistent challenges: water leakages from supply networks. A groundbreaking study led by V. I. Bazhenov of JSC “Water and Wastewater” sheds light on how machine learning (ML) algorithms can revolutionize the way we predict and prevent pipeline failures, offering significant commercial benefits for the energy sector.
Water leakages are more than just a nuisance; they represent a substantial financial and environmental burden. According to Bazhenov’s research, leaks and unauthorized connections to centralized water supply networks (CWN) lead to considerable losses of drinking water, driving up costs and reducing the accuracy of water resource metering. “The relevance is associated with solving practical AI problems based on the latest innovations—forecasting and preventing accidents at CWN with optimal planning of repair work and timely maintenance,” Bazhenov emphasized.
The study, published in the journal Stroitel’stvo: Nauka i Obrazovanie (Construction: Science and Education), delves into the application of various ML models to diagnose and predict water leaks. Bazhenov and his team identified 18 different ML algorithms that can be employed to address leakage issues in CWN. Among these, the Kohonen neural network (KNN) stands out, particularly in the Russian context, where it is the only neural network software available in Russian, known as STATISTICA Automated Neural Networks.
The research highlights the importance of reliable and continuous data collection for making informed decisions based on AI predictions. Databases should include crucial parameters such as pipe diameter, length of the section, age of the pipe, pressure, and type of soil. “The pressure itself (or difference) in the network is not a sign of an accident. This parameter should be considered together with the number of network failures (accidents) in the sections,” Bazhenov noted, underscoring the complexity of interpreting data.
One of the most promising developments mentioned in the study is the use of acoustic and ultrasonic methods for monitoring the condition of underground pipeline networks. These methods, which detect the propagation of volumetric and directional waves (noise), are rapidly gaining traction and could significantly enhance the accuracy of leak detection.
The implications of this research are far-reaching. For the energy sector, which often relies on water for cooling and other processes, reducing water losses can lead to substantial cost savings and improved operational efficiency. Moreover, the integration of AI and ML in water management can pave the way for smarter, more resilient infrastructure, capable of adapting to changing environmental conditions and demand patterns.
As we look to the future, the role of AI in water management is set to become even more pivotal. Bazhenov’s work provides a solid foundation for further innovation, encouraging water utilities and energy companies to embrace these technologies. By doing so, they can not only mitigate the risks associated with water leakages but also contribute to the broader goals of sustainability and resource efficiency. The journey towards smarter water management has begun, and the insights from this research are a significant step forward.
