Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2410.09530 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914970527072256 |
|---|---|
| author | Khoa, Tran Dang |
| author_facet | Khoa, Tran Dang |
| contents | This study presents two models to optimize pressure management in water distribution networks. The first model forecasts pressure at distribution points and compares predictions with actual data to detect anomalies such as leaks and blockages. Early detection allows for timely interventions, minimizing economic losses and ensuring system sustainability. The second model estimates the necessary inlet pressure based on the influence of various distribution points, ensuring consistent water supply while reducing waste and optimizing resource management. Both models utilize modern machine learning algorithms to enhance the prediction process. The methodology includes the CNN-EMD model, which analyzes historical data collected every 15 minutes over two months to predict future pressures. The Empirical Mode Decomposition (EMD) method identifies fluctuations and anomalies, improving prediction accuracy. The second model combines CNN, EMD, and LSTM techniques to forecast required inlet pressure, emphasizing the impact of distribution points. Results show that the CNN-EMD and CNN-EMD-LSTM models enhance pressure management capabilities, with the first model achieving an anomaly detection accuracy of 85% to 95% and the second model predicting inlet pressure with an average accuracy of 93%. This enables flexible system adjustments and identifies critical factors affecting inlet pressure. In conclusion, advanced machine learning models like CNN-EMD and LSTM significantly improve pressure management in water distribution networks, facilitating early issue identification, ensuring efficient water supply, and optimizing resource management for future generations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_09530 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Anomaly Detection and Inlet Pressure Prediction in Water Distribution Systems Using Machine Learning Khoa, Tran Dang Systems and Control 91B30 H.2.4; I.2.3; I.6.8 This study presents two models to optimize pressure management in water distribution networks. The first model forecasts pressure at distribution points and compares predictions with actual data to detect anomalies such as leaks and blockages. Early detection allows for timely interventions, minimizing economic losses and ensuring system sustainability. The second model estimates the necessary inlet pressure based on the influence of various distribution points, ensuring consistent water supply while reducing waste and optimizing resource management. Both models utilize modern machine learning algorithms to enhance the prediction process. The methodology includes the CNN-EMD model, which analyzes historical data collected every 15 minutes over two months to predict future pressures. The Empirical Mode Decomposition (EMD) method identifies fluctuations and anomalies, improving prediction accuracy. The second model combines CNN, EMD, and LSTM techniques to forecast required inlet pressure, emphasizing the impact of distribution points. Results show that the CNN-EMD and CNN-EMD-LSTM models enhance pressure management capabilities, with the first model achieving an anomaly detection accuracy of 85% to 95% and the second model predicting inlet pressure with an average accuracy of 93%. This enables flexible system adjustments and identifies critical factors affecting inlet pressure. In conclusion, advanced machine learning models like CNN-EMD and LSTM significantly improve pressure management in water distribution networks, facilitating early issue identification, ensuring efficient water supply, and optimizing resource management for future generations. |
| title | Anomaly Detection and Inlet Pressure Prediction in Water Distribution Systems Using Machine Learning |
| topic | Systems and Control 91B30 H.2.4; I.2.3; I.6.8 |
| url | https://arxiv.org/abs/2410.09530 |