EV Smart Charging Bot is a Telegram bot that helps electric vehicle (EV) owners find the optimal charging schedule for their cars. By analyzing real-time and forecasted electricity prices and emission rates, the bot provides a detailed charging plan that minimizes cost and environmental impact. The bot uses machine learning and a sequence-to-sequence LSTM model for forecasting and integrates the Gemini LLM to handle user inputs in natural language.
- Real-time optimization: Retrieves real-time electricity price and emission data through APIs.
- Future prediction: Predicts the next 24 hours of electricity prices and emissions using a trained LSTM model.
- Custom user inputs: Allows users to input their EV’s battery state, capacity, charging rate, and departure time.
- Intelligent recommendations: Suggests the most cost-effective and eco-friendly charging schedule.
- Natural language interface: Utilizes the Gemini LLM to parse user inputs and refine user interaction.
- Interactive output: Provides a detailed breakdown of costs and emissions for various charging times.
This project comprises multiple components:
- Telegram Bot: Main interface for users to interact with the system.
- Data Processing: Handles data cleaning and preprocessing for both training and prediction.
- Machine Learning Model: LSTM model for forecasting future prices and emissions.
- Real-time Data Retrieval: API integration to fetch live electricity prices and emissions.
- Database: Stores user information and charging preferences.
Here’s an example interaction:
User: Battery is at 65%
Bot:
Current Status:
• Battery Level: 65.0%
• Departure Time: 8:00 AM
• Battery Capacity: 75.0 kWh
• Charging Rate: 7.0 kW
Analyzing optimal charging times...
It will take 3 hours and 45 minutes to charge your EV.
Here are your potential savings if you start charging at different times:
Start Time Cost Savings CO2 Savings
------------------------------------
10:03 PM $ 0.00 0.0 g
11:03 PM $ 3.52 -40.5 g
12:03 AM $ 8.65 -78.3 g
01:03 AM $ 4.67 -122.5 g
02:03 AM $ 2.70 -136.6 g
03:03 AM $ -1.40 -140.2 g
04:03 AM $ -3.60 -150.7 g
-
Clone this repository:
git clone https://github.com/srmadani/ev-charging-tg-bot.git cd ev-charging-tg-bot -
Install dependencies:
pip install -r requirements.txt
-
Set up your environment variables:
TELEGRAM_BOT_TOKEN: Your Telegram bot token.GOOGLE_API_KEY: Your gemini token.emission_api_token: Keys for accessing real-time emissions APIs.
-
Start the bot:
python main.py
main.py: Main entry point for running the Telegram bot.llm.py: Handles user inputs and converts them into structured data using the Gemini LLM.pred.py: Contains the LSTM model and prediction logic for price and emissions.reg.py: Manages user registration and updates user preferences.retrieve_data.py: Fetches real-time electricity prices and emissions data via APIs.bot_database.db: SQLite database storing user information and preferences.
data_collection.py: Scripts for collecting and preprocessing raw data for training.dtraining.py: Scripts for training the forecasting model and getting forecasts for the next 24 hours.trained_model.h5: Pre-trained LSTM model for predicting future prices and emissions.
- Emission Rates: Electricity Maps
- Electricity Prices: NYISO Custom Reports
Both datasets are used for training the model and fetching real-time data through APIs.
To reproduce the results:
- Train the LSTM model:
python training.py
- Use the
main.pyto test the bot's interaction and predictions in real-time.
This project is licensed under the MIT License.
Special thanks to:
- Electricity Maps and NYISO for the datasets.
- The developers of the Gemini LLM for its robust natural language capabilities.
Feel free to open an issue or reach out for any questions!