Air Quality Control System Simulator

A comprehensive simulation and test-prototype system for intelligent air quality control in indoor environments, focusing on carbon monoxide (CO) monitoring and automated ventilation management.

Project Overview

This repository presents a complete air quality control system that demonstrates real-time monitoring, predictive modeling, and automated control of indoor air quality through intelligent ventilation management. The system serves as both a functional prototype and a showcase of modern approaches to environmental control using machine learning and real-time simulation.

What This Project Showcases

This repository demonstrates several key concepts and technologies:

1. Model Predictive Control (MPC): Implementation of a complete MPC framework for environmental control systems, featuring prediction horizon, optimization, and receding horizon control.

2. Predictive Environmental Control: Using machine learning models to predict future air quality conditions and optimize ventilation systems proactively rather than reactively.

3. Real-time Simulation Framework: A comprehensive simulation environment that models the complex interactions between occupancy, ventilation systems, and air quality parameters.

4. Interactive Web-based Interface: A modern dashboard built with Dash/Plotly that provides real-time visualization and control capabilities.

5. Test-Prototype Methodology: A systematic approach to developing and validating environmental control systems through simulation before physical implementation.

System Architecture

Core Components

1. Data Generation and Modeling (data_creation/)

• data_creation.ipynb: Synthetic data generation for training machine learning models
• data/data.csv: Comprehensive dataset containing CO levels, occupancy patterns, and ventilation parameters
• Simulates realistic indoor air quality scenarios with varying occupancy and ventilation conditions

2. Predictive Models

• predictive_model.ipynb: Development and training of neural networks and regression models
• predict_model.keras: Trained TensorFlow/Keras neural network for CO prediction
• linear_regressor.pkl: Linear regression model for comparison and fallback
• scaler.pkl: Data preprocessing scaler for model inputs

3. Simulation Engine (simulator.py)

The heart of the system, featuring a Model Predictive Control (MPC) implementation:

• Real-time air quality simulation based on physical parameters
• Model Predictive Control (MPC) algorithm that predicts CO levels over a 10-minute horizon
• Predictive optimization using trained models to forecast CO levels and optimize control actions
• Automated pump control with power optimization algorithms
• Adaptive thresholding for different environmental conditions

4. Web Application (app.py, pages/)

• Interactive dashboard with real-time visualization
• Parameter control interface for system configuration
• Live simulation monitoring with predictive forecasting display
• User-friendly controls for adjusting room volume, occupancy, and ventilation parameters

5. Testing and Validation (test_predict_model.ipynb)

• Model validation and performance testing
• Scenario testing with various environmental conditions
• Comparative analysis between different prediction approaches

---

Key Features

Intelligent Control System

• Model Predictive Control (MPC): The system implements a full MPC framework that predicts CO levels 10 minutes into the future and optimizes pump power accordingly
• Predictive Horizon: Uses a 10-step prediction horizon to anticipate future air quality conditions
• Optimization Algorithm: Automatically determines optimal pump power (0-100%) by evaluating multiple control scenarios
• Receding Horizon Control: Updates control actions at each time step based on new measurements and updated predictions
• Adaptive Response: Considers occupancy patterns, room volume, and current air quality conditions

Real-time Monitoring

• Live CO Concentration Tracking: Real-time display of current and predicted CO levels
• Threshold Management: Configurable safety thresholds with visual alerts
• Historical Data Visualization: Trend analysis and pattern recognition

Environmental Parameters

• Room Volume: Adjustable space size (10-30 m³)
• Occupancy: Real-time people count (0-10 persons)
• Ventilation Capacity: Configurable pump capacity (141-708 L/min)
• System Scaling: Multiple pump configuration (1-8 pumps)

The Test-Prototype Significance

Why This Approach Matters

This project represents a test-prototype methodology that bridges the gap between theoretical environmental control systems and real-world implementation. The significance includes:

1. Risk Mitigation: Testing control algorithms in simulation before deploying to actual buildings prevents potentially dangerous situations with air quality.

2. Cost-Effective Development: Validating system behavior across thousands of scenarios without physical hardware reduces development costs and time.

3. Performance Optimization: Fine-tuning control parameters and algorithms in a controlled environment ensures optimal performance.

4. Scalability Testing: Evaluating system behavior across different room sizes, occupancy levels, and ventilation configurations.

Real-World Applications

The prototype demonstrates applicability to:

• Office Buildings: Automated air quality management for varying occupancy
• Residential Spaces: Smart home integration for health-conscious living
• Industrial Facilities: Worker safety through predictive air quality control
• Healthcare Facilities: Critical air quality management in sensitive environments

Technical Implementation

Model Predictive Control (MPC) Framework

The system implements a complete MPC architecture with the following components:

1. Prediction Model

• Neural Network Predictor: Uses a trained TensorFlow/Keras model to predict CO concentrations
• Multi-step Prediction: Generates 10-minute ahead predictions iteratively
• State Variables: Considers current CO levels, occupancy, room volume, and ventilation parameters

2. Optimization Problem

• Objective Function: Minimizes deviation from CO threshold over the prediction horizon
• Control Variables: Pump power percentage (0-100%)
• Constraints: Physical limits on pump capacity and power consumption
• Discrete Optimization: Evaluates multiple pump power levels [0, 25, 50, 60, 65, 70, 75, 80, 85, 90, 95, 100]

3. Receding Horizon Implementation

• Control Update: Optimal pump power is applied for one time step
• State Update: System state is updated with new measurements
• Horizon Shift: Prediction horizon shifts forward for next optimization cycle

4. Key MPC Functions in simulator.py

• predict(): Multi-step prediction over 10-minute horizon
• optimize(): Evaluates different pump power levels and selects optimal control action
• simulate_time_step(): Implements the receding horizon control loop

Machine Learning Pipeline

1. Data Collection: Synthetic data generation based on real-world parameters
2. Feature Engineering: Processing occupancy patterns, environmental conditions, and ventilation parameters
3. Model Training: Multiple approaches including neural networks and linear regression
4. Validation: Comprehensive testing across various scenarios
5. Deployment: Real-time model inference in the simulation environment

Simulation Framework

• Physics-Based Modeling: Realistic air quality dynamics considering room volume, occupancy, and ventilation
• Model Predictive Control Implementation: Complete MPC framework with prediction model, optimization, and receding horizon control
• Stochastic Elements: Random variations to simulate real-world uncertainty
• Optimization Algorithms: Smart pump control using predictive models with discrete optimization over pump power levels
• Real-time Processing: Live simulation with configurable time steps

Installation and Usage

Prerequisites

• Python 3.10.12
• Required packages: dash, plotly, tensorflow, pandas, numpy, scikit-learn

Quick Start

1. Clone the repository
2. Install dependencies: pip install -r requirements.txt (if available)
3. Run the application: python app.py
4. Access the dashboard at http://localhost:8050

Using the Simulator

1. Configure Parameters: Set room volume, occupancy, and ventilation capacity
2. Start Simulation: Begin real-time air quality monitoring
3. Monitor Performance: Observe CO levels, pump power, and predictions
4. Adjust Settings: Modify parameters to see system response

Future Enhancements

This test-prototype provides a foundation for:

• IoT Integration: Connecting real sensors and actuators
• Multi-pollutant Monitoring: Expanding beyond CO to include CO₂, PM2.5, VOCs
• Building Integration: HVAC system integration and building automation
• Mobile Applications: Remote monitoring and control capabilities
• Energy Optimization: Balancing air quality with energy efficiency

License

This project is intended for educational and research purposes, demonstrating the application of machine learning and simulation techniques to environmental control systems.

---

This repository showcases the potential of combining Model Predictive Control, real-time simulation, and intelligent control systems to create safer, more efficient indoor environments.