CAPE – Smart Power Home Automation System

Overview

CAPE (Cognitive Automated Power Environment) is an IoT-based home power management and predictive maintenance system designed to monitor voltage, current, and power usage in real time.
It automatically generates usage-based billing, detects abnormal load behavior, and can cut off power or reduce load when unsafe or inefficient conditions arise — using MOSFETs and relay control circuits.

The system integrates AI models to predict device faults and abnormal power consumption patterns using historical sensor data.

Objectives

• Real-time monitoring of current, voltage, and power for multiple household devices.

• Automated billing system for tracking usage and generating cost summaries.

• Anomaly detection and fault prevention using machine learning.

• Remote power control (ON/OFF) through cloud dashboard and SMS alerts.

• Optimize energy usage and improve device lifespan through predictive analytics.

System Architecture

Hardware Setup

Core Components:

• Arduino Uno (Data acquisition and control)

• ACS712 Current Sensor

• ZMPT101B Voltage Sensor

• MOSFET (IRFZ44N) for PWM-based control

• Relays for circuit isolation and load cutoff

• Raspberry Pi (Edge compute and analytics)

  

Functional Flow

1. Sensor Acquisition → Current & voltage signals sampled by Arduino.

2. Computation → RMS values computed every 1 second, power = V × I × cos(φ).

3. Transmission → Data sent to Raspberry Pi via serial (115200 baud).

4. Storage & Processing → Pi logs to SQLite/CSV and runs predictive model.

5. Decision Layer → If power exceeds threshold, MOSFET reduces duty cycle or relay cuts load.

6. User Notification → Sends SMS with bill summary and warning via Twilio API.

📊 System Graphs & Visualizations

1️⃣ Real-time Power Consumption Graph
Displays live power usage (in watts) across multiple devices over time.

Graph: Line plot with timestamps on X-axis, Power (W) on Y-axis, color-coded per device.

2️⃣ Voltage and Current Waveforms
Captured at sensor input using ADC sampling.

Graph: Dual Y-axis plot – Blue line for Voltage (V), Orange line for Current (A).

3️⃣ Predicted vs Actual Load Curve (AI Model Output)
LightGBM regression output showing deviation between predicted healthy load and observed current spikes.

Graph: Smooth baseline (predicted) vs noisy peaks (actual) – anomaly marked in red.

4️⃣ Energy Consumption Breakdown (Pie Chart)
Shows % energy used by each appliance category (Lighting, HVAC, Kitchen, etc.).

Graph: Donut chart with device labels and percentage values.

5️⃣ Monthly Billing Trend (Bar Graph)

Graph: Bars representing total cost (₹) per month, highlighting reduction after system optimization.

Predictive Maintenance Model

Performance Metrics:

• RMSE: 0.07

• R² Score: 0.93

• Inference Speed: 12ms/sample

Mathematical Model

1. Power Equation:
P = V × I × cos(φ)

2. Energy Consumption:
E = Σ (P × Δt)

3. Cost Estimation:
Bill = E (kWh) × Tariff Rate

4. Fault Detection Threshold:
If |I_measured - I_predicted| > 15% of I_predicted → Fault Detected

Software Stack

Example Data Frame (Sample)

🔍 Insights & Results

✅ Reduced testbed energy consumption by ~20% after automated control logic.
✅ Achieved 92% accuracy in anomaly prediction with LightGBM model.
✅ Prevented overcurrent damage in simulation by relay cutoff within 80 ms.
✅ Monthly energy report auto-generated and sent via SMS.

Achievements

🏆 First Place – Best Hardware Project at Ignitex, BGSIT (2025)
Recognized for innovation in AI-integrated IoT power systems with practical scalability.

How to Run Locally

1. Install Dependencies:

pip install flask lightgbm pandas numpy pyserial twilio

2. Connect Arduino via USB:

   • Select COM port /dev/ttyUSB0 or similar.

3. Run Server:

python app.py

4. Open Dashboard:

   • Access at http://localhost:5000

   • View live data charts and fault alerts.

📂 File Structure

CAPE/
│
├── arduino/
│   ├── cape_main.ino
│   └── sensor_calibration.ino
│
├── pi_server/
│   ├── app.py
│   ├── model/
│   │   └── lightgbm_model.pkl
│   ├── static/
│   │   ├── js/
│   │   └── css/
│   ├── templates/
│   │   └── dashboard.html
│   ├── utils/
│   │   ├── serial_reader.py
│   │   ├── sms_alert.py
│   │   └── billing.py
│   └── data/
│       └── power_log.csv
│
└── README.md

📈 Visualization Reference (for Notion embeds)

You can later upload:

• Power vs Time (Line Graph)

• Energy Distribution (Pie Chart)

• Predicted vs Actual Load (Scatter/Line)

• Billing Trend (Bar Chart)

• Fault Detection Timeline (Highlighted Events)

License

Licensed under the MIT License

Contact

For collaboration or technical queries, reach out:
📧 mohammedrah1289@gmail.com