This repository includes our fully data-driven unsupervised leak detection and localization framework in water distribution networks, employing 1D-CAE and SDNML change-point detector to realize real-time leak detection.

Python v3.8.11

Pytorch v1.8.0 cuda11.1

numpy v1.24.3

pandas v1.5.3

matplotlib v3.2.2

sklearn v1.2.2

epynet v1.1

rtchange

Our work is validated on the L-Town dataset. Dataset source: https://github.com/KIOS-Research/BattLeDIM

1. $ git clone https://github.com/Jolyne-lhy/WaterLeakDetection.git
   $ cd WaterLeakDetection
   

2. Install dependencies.

3. Download dataset.

4. Train and test the 1D-CAE model with 1D-CAE Trainer.ipynb and 1D-CAE Tester.ipynb, run Leak detection (Change Point Detection).ipynb to do change point detection.

5. Run Leak localization (impact factor).ipynb to do localization.