🔋 MCP-Based EV Digital Twin Agent with Gemma-Powered Tool Calling

An AI-powered EV Digital Twin platform that combines battery analytics, predictive maintenance, fleet monitoring, MCP-based tooling, and Gemma-powered intelligent decision support.

The system allows users to analyze battery health, estimate remaining useful life (RUL), monitor fleet-wide battery performance, detect data drift, and interact with EV analytics through a natural language interface.

---

Dashboard Preview

Battery Digital Twin

Fleet Analytics

Monitoring & Drift Detection

---

🚗 Key Features

Battery Digital Twin

• State of Health (SOH) Prediction
• Remaining Useful Life (RUL) Estimation
• Battery Range Estimation
• Future Degradation Simulation
• What-if Scenario Analysis

Charging Intelligence

• Charging Pattern Analysis
• Charging Risk Scoring
• Fast Charging Detection
• Battery Protection Recommendations

Driving Intelligence

• Driving Style Analysis
• Energy Consumption Evaluation
• Vehicle Efficiency Insights

Predictive Maintenance

• Maintenance Priority Assessment
• Maintenance Recommendations
• Battery Risk Identification

Fleet Analytics

• Fleet Health Monitoring
• Healthiest Battery Detection
• Riskiest Battery Detection
• Fleet Status Evaluation

Data Drift Detection

• SOH Drift Monitoring
• Fleet Stability Analysis
• Monitoring Dashboard

Gemma-Powered Tool Calling Agent

The system includes a tool-calling EV assistant powered by Gemma.

Examples:

• Compare B0005 and B0047
• Which battery is the riskiest?
• Which battery is the healthiest?
• Why is B0047 riskier than B0005?

The agent automatically selects the appropriate analytical tool and converts the tool output into a professional EV engineering explanation.

MCP Server Tools

Implemented MCP-compatible tools:

• predict_future_soh()
• estimate_rul()

The project includes MCP server infrastructure and tool-based EV analytics services.

---

🏗 Architecture

User
 │
 ▼
Gemma Tool-Calling Agent
 │
 ▼
EV Digital Twin Engine
 │
 ├── Battery Twin
 ├── Charging Intelligence
 ├── Driving Intelligence
 ├── Predictive Maintenance
 ├── Fleet Monitoring
 ├── Drift Detection
 └── MCP Server Tools
 │
 ▼
Streamlit Dashboard

---

📊 Datasets

NASA Battery Dataset

Used for:

• Battery degradation modeling
• SOH prediction
• RUL estimation

EV Charging Patterns Dataset

Used for:

• Charging behavior analysis
• Charging risk assessment

EV Telemetry Dataset

Used for:

• Driving intelligence
• Energy consumption analytics
• Vehicle efficiency evaluation

---

🛠 Technologies

• Python
• Streamlit
• Scikit-Learn
• Pandas
• Plotly
• Ollama
• Gemma 3
• MCP (Model Context Protocol)
• Machine Learning

---

Installation

git clone https://github.com/mahmutcanborann/mcp-ev-digital-twin-agent.git

cd mcp-ev-digital-twin-agent

pip install -r requirements.txt

---

Run Dashboard

streamlit run dashboard/app.py

---

🔮 Future Work

• Full MCP Client-Server Tool Orchestration
• Advanced Fleet AI Analyst
• Hugging Face Deployment
• Real-Time Vehicle Telemetry Integration
• Multi-Agent EV Diagnostics

---

👨‍💻 Author

Mahmut Can Boran

Computer Engineer | AI Engineer |

Areas of Interest:

• Automotive AI
• Digital Twins
• Agentic Systems
• Predictive Maintenance
• EV Software
• MCP Architectures