BIWMS (Blockchain Integrated Water Management System)

A system for predicting and allocating water consumption across different boroughs.

Overview

This system uses LSTM neural networks to predict water consumption patterns based on historical data. The predictions are used to allocate water resources efficiently and are stored on IPFS for transparency and immutability.

Features

• AI model trained on historical water consumption data
• FastAPI backend for file uploads and prediction generation
• IPFS integration via Pinata for decentralized storage
• Blockchain oracle integration for on-chain data availability

Installation

Prerequisites

• Python 3.8+
• Node.js 14+
• pip
• npm

Python Dependencies

pip install -r requirements.txt

Node.js Dependencies

npm install

Environment Setup

Create a .env file in the root directory with the following variables:

# Pinata IPFS Credentials
PINATA_API_KEY=your_pinata_api_key
PINATA_SECRET_API_KEY=your_pinata_secret_key

# Blockchain Oracle Credentials (optional)
ORACLE_PRIVATE_KEY=your_private_key
ORACLE_CONTRACT_ADDRESS=your_contract_address
RPC_URL=your_rpc_url

Running the API

Start the FastAPI server:

uvicorn app:app --reload

The API will be available at http://localhost:8000.

API Usage

Prediction Endpoint

POST /predict

Upload a CSV file with water consumption data to generate predictions.

Example using curl:

curl -X POST "http://localhost:8000/predict" \
  -H "accept: application/json" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@/path/to/your/water_consumption_data.csv"

Response:

{
  "status": "success",
  "prediction_id": "550e8400-e29b-41d4-a716-446655440000",
  "ipfsHash": "QmYwAPJzv5CZsnA625s3Xf2nemtYgPpHdWEz79ojWnPbdG",
  "confidence_score": 95.6,
  "timestamp": "20230301_120000",
  "prediction_file": "prediction_20230301_120000.json"
}

Model Details

• Architecture: Bidirectional LSTM
• Features: Time-based, rolling statistics, lag features
• Performance: ~89% R² on test data

File Structure

├── AI_feeds/
│   ├── models/            # Trained models and artifacts
│   ├── outputs/           # Model predictions
│   ├── predict.py         # Prediction logic
│   └── train_lstm_model.py # Model training script
├── app.py                 # FastAPI application
├── pinata_uploader.py     # IPFS upload functionality
├── oracle_submit.js       # Blockchain oracle submission script
├── data_uploads/          # Uploaded data files
├── package.json           # Node.js dependencies
└── requirements.txt       # Python dependencies

License

ISC