1. Background & Purpose

Vu Gia - Thu Bon River basin is the most extensive river system of Quang Nam province and Danang City, with an area of 10,350 km² (1). The river contributes to the development of the country's economy, society, and water security with  2.5% of the total water supply; this river basin contributes 1.5% to the national GDP. However, this river basin is changing for the worse because the flow of the Vu Gia - Thu Bon river basin is also declining, which leads to empty downstream of the river basin and saltwater intrusion (2). The intrusion of saltwater caused by withdrawals of freshwater from the groundwater system means there is insufficient water to supply for daily life and production of the people (agriculture, aquaculture,…) around the area. Solving saltwater intrusion and water scarcity in Da Nang City, Quang Nam Province, is urgent. 

Fig 1. Location of Vu Gia - Thu Bon River basin in Quang Nam - Da Nang and Vietnam

There are many solutions used to predict salt intrusion, such MIKE 11 method can, predict many scenarios of salt intrusion via the effects of climate change and sea level rise in the future. Le Hung (2016) estimated salt intrusion in the Vu Gia – Thu Bon River basin up to 2100 due to climate change (3), however, some limitations are found, such as the significant variance of accuracy, prediction time being too long, and accuracy being affected by many factors. 

Fig 2. Thousands of hectares of crops have been damaged by saltwater in the Mekong Delta Vietnam

This research project proposes an intelligent technique for predicting the early salt intrusion process to address the problems above. The proposed model will obtain higher accuracy and earlier prediction than previous methods. Also, it can be developed and embedded into the LoRa-based Internet of Things to monitor water quality in real time in the future. The findings prove a new approach to Artificial Intelligence (AI) and Machine Learning (ML). 

The consequences of saltwater intrusion in the Vu Gia River basin are severe for communities’ health, agriculture, environment, and economy. Therefore, to gain an overview of the theoretical to practical basis for solving this problem, team members include artificial intelligence experts, medical doctors, public health experts, and environmental experts.

Our purpose to develop a practical AI model with control chart technique to predict salinity intrusion at Vu Gia - Thu Bon River basin

Fig 3. The process of proposed method for project purpose 

2. Program Goals and Main Activities

2.1. Program goals

The proposed idea is to develop a practical AI model with control chart technique to predict the water quality at Vu Gia River, especially when water is salty. To achieve that goal, the specific objective of the project is:

To develop a practical model, combining the Transformer and Support Vector Data Description with control chart, to accurately predict the water pollution and salt intrusion process, thereby providing a valuable tool for water quality management.

The findings will assist to enhance early warning systems at Cau Do water plant, the government in making long- and short-term policies to reduce water pollution and minimize the saltwater intrusion process effectively. The new approach is also considered one of the most cost-effectiveness solution for local water plant, governments, agriculture and people.

*Method:

+ Transformer: Transformer is a new kind of neural architecture which encodes the input data as powerful features via the attention mechanism. Some studies about water quality prediction have shown high performance of transformer model or transformer combining models (14, 15). Project team members have also studied using transformer in some areas (6, 10, 12).

+ Support Vector Data Description: Support Vector Data Description (SVDD) is inspired by the Support Vector Classifier – Machine Learning. It obtains a spherically shaped boundary around a dataset and analogous to the Support Vector Classifier. Some studies about water quality prediction and classification obtained high accuracy (16-18). Some paper about using ML or combine ML with statistical processing control or in federated setting for abnormal detection are conducted by project team members (4, 5, 11, 13).

2.2. Activities

In order to meet the goal, main activities are planned with a time-line (dd/mm/yyyy) from the literature review phase, experiment model, and end report, particularly: 

-          Package 1: Literature review (01/01 – 31/02/2025)

• Review literature on the current situation of saline rivers in the world and Vietnam and theoretical models
• Review literature on methods using control charts, machine learning, and deep learning on predicting and classifying saline rivers (the world and Vietnam).

-          Package 2: Develop algorithm for models and experiment with dataset available (01/03 - 30/04/2025)

• Develop a model combining the Transformer with SVDD 
• Experiment with the proposed model with the dataset available (https://data.world/marineinstitute/b2348da6-873f-4ebb-a03c-5280ee2cf025)
• Evaluate the results of models and adjust parameters for the model to have optimal accuracy.

-          Package 3: Collect quality of water (2020 – 2024) from the Cau Do water Plant and clean data. (01/05 – 30/06/2025)

• Meet leaders of Cau Do water Plant, Department of Natural Resources and Environment of Quang Nam Province, and Danang City to ask permission for data collection and discuss the novel solution
•  Clean and manage data after collection

-          Package 4: Model Experimentation with local data (01/07 – 30/08/2025)

• Experiment models with water dataset, evaluate models with adjusted parameters to improve accuracy

-          Package 5: Reporting and publication (01/09 – 31/12/2025)

•  Write reports about the project's findings, finance, and impacts on communities, governments, and policies.
• Write a manuscript to submit paper in international journals 
• Write reports and present at international and national scientific conferences
• Transfer findings of the practical project into a short academic course for students studying AI and DS at Dong-A University. 
• Share findings via university websites and other platforms: https://donga.edu.vn/, https://iad.donga.edu.vn/, https://ai.donga.edu.vn/

3. Expected results

3.1. Expected results of research

The research results of the project include the following product/activity packages:

Package 1: Have a literature review

Package 2: A completed model can work on available dataset with the accuracy >80%

Package 3: Get permission for data collection from Cau Do water Plant and have data

Package 4: The model work on local dataset with accuracy > 95%

Package 5: Have 2 reports for the US-ASEAN Science, Technology and Innovation Cooperation (STIC) Program and local governments. Have a short academic course for students and sharing findings on 3 websites, 3 fan pages of university, institute.

3.2. Expected results of publication

No.	Publication	Quantity
1	Reputable international journals (Scientific journals selected from the SCI and SCIE catalogs, issued by the Foundation)	1
2	International and national scientific conferences	2

This grant will facilitate the development of Federated setting on edge computers, where data is distributed and trained at each edge device. In the next phase, the goal is to build and experiment edge AI system, which has advantages about data privacy, security, and higher accuracy. Most importantly, the final phase is AI algorithm models embedded into the intelligent water quality monitor system (LoRa-based Internet-of-Things), which can monitor and predict water quality in real-time. 

- Future funding sources: 

+        Local funds from the government, such Department of Science and Technology in Quang Nam, Da Nang, and the National Foundation for Science and Technology Development (NAFOSTED)

+        International funds from UNICEF, the United Nations Development Programme (UNDP), Ministry of Natural Resources and Environment.

Fig 4. The process of build and experiment edge AI system, LoRa-based Internet-of-Things monitor and predict water quality in real-time and benefits for Ecosystem

*Key personnel

Dr. Quoc-Thông Nguyen (PI) has a wealth of experience in machine learning (ML) and artificial intelligence (AI) applications, gained through a long research and academic career. His extensive experience spans technical development and management in numerous projects utilizing machine learning and AI, particularly in anomaly detection, quality control, classification, and forecasting. His research has recently extended to intelligent manufacturing and IoT systems (4, 5). This project on water quality monitoring leverages his strengths perfectly, as it proposes an IoT system embedded with AI/machine learning algorithms for real-time pollution detection, quality forecasting, and salt intrusion monitoring.

Ph.d student Kim Duc Tran is currently Permanent Researcher and Vice Director at  International Research Institute for Artificial Intelligence and Data Science (IAD), Dong A University. He is Ph.d student in the area of Artificial Intelligence and Data Science. He has over 15 publications in international journals and expertise in anomaly detection, explainable AI, and predictive maintenance. Leads multiple AI and IoT projects, including smart manufacturing and healthcare applications. His work on dynamic anomaly detection frameworks and machine learning for real-time monitoring aligns directly with the project's goals of developing AI and IoT systems for water quality monitoring and anomaly detection (6).

MPH. Tho Nguyen is working at IAD, Dong A University; his role is coordinator and researcher at IAD. He is a master of Public health major at Hanoi University of Public Health. He has experience and relevant research: Environmental problems are one of the main topics in public health. Implementing AI in this area is essential to solve the problem most effectively. MPH. Tho has experience studying environmental health, managing, cleaning, and analyzing data with machine, deep, and federated learning (4). These experiences align with the necessary activities in a research project.

Msc. Engineer Nguyen Dac Hieu is a lecturer and researcher at the Department of Artificial Intelligence at Thuy Loi University. He is a Master of Engineering in Information Technology at Hanoi University of Science and Technology—his relevant experience as a researcher specializing in prediction and anomaly detection at MIT, I focus on developing and implementing state-of-the-art models to improve the accuracy and reliability of salinity intrusion forecasts, particularly Transformer-based models. My goal is to enhance early warning systems and mitigate the impact of salinity intrusion on community life. Hieu concentrates on developing and implementing the state of art models to improve the accuracy and reliability of flood forecasting, particularly Transformer-based models, ultimately aiming to enhance early warning systems and mitigate the impact of flooding on communities and infrastructure (7).

Ph.D. Hoang Thanh Huong is a coordinator and researcher at the Science - Technology Management and International Cooperation Office, Danang University of Medical Technology and Pharmacy. She has a Master's in Environmental Management and a Ph.D. in Environmental Systems. She has researched waste-to-energy-oriented processing to sustainable development and participated in several other studies as a researcher, a surveyor, and an analyst. She is experienced in sustainable energy research, waste to energy, waste treatment, poverty alleviation, and policy making (8, 9).

Kim Phuc Tran is a Senior Associate Professor (Maître de Conférences HDR, equivalent to a UK Reader) of Artificial Intelligence and Data Science at the ENSAIT and the GEMTEX laboratory, University of Lille, France. He has a Master of Engineering in Automated Manufacturing. He obtained a Ph.D. in Automation and Applied Informatics at the University of Nantes, and an HDR (Doctor of Science or Dr. Habil.) in Computer Science and Automation at the University of Lille, France. He has published more numerous papers in peer-reviewed reputed journals and proceedings of international conferences. He edited 3 books. He is an Associate Editor, Editorial Board Member, and Guest Editor for several international journals such as IEEE Transactions on Intelligent Transportation Systems and Engineering Applications of Artificial Intelligence. Kim Phuc Tran has supervised 12 Ph.D. students and 3 Postdocs. In addition, as the project coordinator (PI), he conducted a national project about Healthcare Systems with Federated Learning. He has been or is involved (PI, co-PI, or member) in 13 national and European projects. He is an expert and an evaluator for the Public Service of Wallonia (SPW-EER), Belgium, the Natural Sciences and Engineering Research Council of Canada, ARN (Agence Nationale de la Recherche), ANRT (Association Nationale de la Recherche et de la Technologie), and CY Cergy Paris University, France. He received the Award for Scientific Excellence (Prime d’Encadrement Doctoral et de Recherche) from by the Ministry of Higher Education, Research and Innovation, France for 4 years from 2021 to 2025 in recognition of his outstanding scientific achievements. From 2017 until now, he has been the Senior Scientific Advisor at Dong A University and the International Research Institute for Artificial Intelligence and Data Science (IAD), Danang, Vietnam where he holds the International Chair in Data Science and Explainable Artificial Intelligence. His research interests include Explainable Trustworthy, and Transparent Artificial Intelligence, Intelligent Decision Support Systems, Digital Twins, and Applications of AI, Edge Computing, and Data Science in Industry 5.0 (4, 6, 10-13).

---

References

1. Ministry of Natural Resources and Environment. New management method in Vu Gia - Thu Bon river basin 2015 [cited 2024 Jan 10]. Available from: https://monre.gov.vn/English/Pages/New-management-method-in-Vu-Gia--Thu-Bon-river-basin.aspx.
2. Đinh Phùng Bảo. Xây dựng công nghệ dự báo dòng chảy cạn, xâm nhập mặn cho hệ thống sông Vu Gia - Thu Bồn. Tạp chí Khí tượng Thuỷ văn. 2017;10:48- 55.
3. Lê Hùng. Nghiên cứu ứng dụng mô hình Mike 11 tính toán xâm nhập mặn theo các kịch bản biến đổi khí hậu, nước biển dâng đến 2100 cho hạ lưu sông Vu Gia Thu Bồn. QUANG NAM: Quang Nam: Đề Tài Khoa Học Và Công Nghệ Cấp Đại Học Đà Nẵng; 2016. p. 1-24.
4. Nguyen T, Tran KD, Raza A, et al. Wearable technology for smart manufacturing in industry 5.0.  Artificial Intelligence for Smart Manufacturing: Methods, Applications, and Challenges: Springer; 2023. p. 225-54.
5. Tran PH, Ahmadi Nadi A, Nguyen TH, et al. Application of machine learning in statistical process control charts: A survey and perspective.  Control charts and machine learning for anomaly detection in manufacturing: Springer; 2022. p. 7-42.
6. Nguyen TTV, Heuchenne C, Tran KD, et al., editors. A Novel Transformer-Based Anomaly Detection Approach for ECG Monitoring Healthcare System. International Conference on Safety and Security in IoT; 2023: Springer.
7. Hieu TT, Chieu TQ, Quang DN, et al. Water level prediction using deep learning models: A case study of the Kien Giang River, Quang Binh Province. 2023;2(4):468-79.
8.  Hoang Thanh H. Design and Evaluation of the Economic-Environmental Benefits from Livestock Waste and Food Waste Treatment by Small-Scale Biogas Digesters in Farming Communities for Sustainable Development in Central Vietnam.
9. Hoang TH, Kato T, Hoang H. Using household-scale biogas digesters as a tool for poverty alleviation in central Vietnam. Journal of Engineering Science Technology. 2022;17(6):4235-52.
10. Truong HT, Ta BP, Le QA, et al. Light-weight federated learning-based anomaly detection for time-series data in industrial control systems. Computers in Industry. 2022;140:103692.
11. Tran KP. Control charts and machine learning for anomaly detection in manufacturing: Springer; 2022.
12. Raza A, Tran KP, Koehl L, et al. Lightweight transformer in federated setting for human activity recognition. arXiv preprint arXiv:00244. 2021.
13. Raza A, Li S, Tran K-P, et al. Using Anomaly Detection to Detect Poisoning Attacks in Federated Learning Applications. arXiv preprint arXiv:08486. 2022.
14. Wang R, Qi Y, Zhang Q, et al. A multi-step water quality prediction model based on the Savitzky-Golay filter and Transformer optimized network. Environmental Science Pollution Research. 2023;30(50):109299-314.
15. Peng L, Wu H, Gao M, et al. TLT: Recurrent fine-tuning transfer learning for water quality long-term prediction. Water Research. 2022;225:119171.
16. Abobakr Yahya AS, Ahmed AN, Binti Othman F, et al. Water quality prediction model based support vector machine model for ungauged river catchment under dual scenarios. Water. 2019;11(6):1231.
17. Jamshidzadeh Z, Ehteram M, Shabanian H. Bidirectional Long Short-Term Memory (BILSTM)-Support Vector Machine: A new machine learning model for predicting water quality parameters. Ain Shams Engineering Journal. 2024;15(3):102510.
18. Shams MY, Elshewey AM, El-kenawy E-SM, et al. Water quality prediction using machine learning models based on grid search method. Multimedia Tools Applications. 2023:1-28.