District-Heating Boiler Digital Twin
Integrated a district-heating boiler digital twin with a Siemens S7 PLC (Modbus) and SCADA, and prepared years of sensor history for predictive models.
- role
- Head Python Developer / Solution Architect
- period
- 2023
Problem
An energy-sector digital twin that had to run against real plant control systems and years of messy historical sensor data.
Training directly on the real boiler risks real harm and cost: chimney-mounted sensors would flag false emissions during combustion-quality tests, and some scenarios need to simulate conditions that threaten to burst the boiler. Downtime costs millions per day, and a real failure triggers a formal commission, since this is critical infrastructure. The digital twin solves this by simulating the real boiler and its behavior with high fidelity, enabling training and assessment on hazardous scenarios without touching the actual equipment.
What was built
- Integration with a Siemens S7-series PLC over Modbus, and with SCADA.
- UI exposing the ML pipeline: read PLC addresses → generate predictions → write results back to PLC addresses → present on SCADA.
- Four years of history across 150+ sensor variables, feature-engineered for the predictive models.
- A dashboard for managing the simulation, including time-speed control.
Outcome
- Correct PLC/SCADA communication for 150+ parameters at set intervals, with working models; continued past the PoC phase.
- Digital twin behaves like the real boiler.
- Publicly documented EU-funded project.