Geothermal steam monitoring

Connecting plant signals to a live view of estimated steam quality.

I developed the acquisition hardware and software, integrated the estimator, and connected storage and dashboard delivery with the research team.

R&D Engineer

Universitas Padjadjaran × Pertamina Geothermal Energy

2024 to Jan 2025

Completed Jan 2025

System overview

Input: Pressure, temperature and flow from 4-20 mA transmitters
Processing: Parallel signal acquisition, scaling and model inference on an Intel NUC
Output: Live and historical views with estimated steam dryness

I worked as the R&D Engineer with Universitas Padjadjaran researchers and PGE engineers. The team continued implementation after handoff.

Built hardware and monitoring view

Perta-SMART added a monitoring path beside the existing DCS. Physical steam-dryness tests remained the reference; the model supplied an estimate between tests.

Closed Perta-SMART enclosure with the project and company marks. — Enclosure built for the geothermal monitoring system.

Open enclosure showing the power supply, Intel NUC, Ebyte ME31 acquisition module and cooling fan. — Edge computer and acquisition components inside the enclosure.

Dashboard demonstration with estimated dryness, temperature, pressure, flow and trend charts. — Measured signals appear beside estimated dryness. Displayed values are demonstration data.

Geothermal plant buildings with steam rising above a forested site. — Steam rising above the Ulubelu plant buildings.

Preserve the existing signal route

The installed transmitters continued to feed the DCS while splitters supplied a parallel 4-20 mA path to the Ebyte ME31 module.

Pressure, temperature and flow transmitters feed the DCS and a parallel Ebyte ME31 acquisition path, then connect to the Intel NUC and dashboard. — The splitters retain the existing DCS route while feeding Perta-SMART acquisition.

Scale, estimate, store and deliver

The NUC reads the acquisition module over Modbus TCP/IP. Python scales current readings with the transmitter ranges before the model, storage and dashboard paths use them.

Diagram of Modbus readings entering Python scaling and model processing on the Intel NUC, with readings stored in SQL and delivered through a WebSocket gateway. — The design shows scaling, estimation, storage and dashboard delivery. Its analog return route is a design path, not a verified control loop.

Stage	System behavior and my work
Instrumentation	Splitters preserve the DCS route and feed Ebyte ME31 acquisition. My work: developed the parallel signal-acquisition hardware and software.
Read and scale	The NUC reads Modbus TCP/IP and Python converts current values using transmitter ranges. My work: configured the acquisition path and scaling.
Estimate and store	The model estimates dryness from measured inputs; PostgreSQL and TimescaleDB retain readings and ten-minute aggregates. My work: integrated the estimator and connected storage with the research team.
Deliver	WebSocket supplies current updates; HTTP serves historical queries. My work: connected data delivery to the dashboard workflow with the team.

Field integration and checks

Project team reviewing a hardware setup around a table in a meeting room. — Project team members review the equipment during development work.

Geothermal project team in PPE holding a Perta-SMART banner outside a facility. — Team context at the project site; this is not a performance result.

Check	Result and conditions
Laboratory input path	Average current-loop error around 0.048 mA; largest reported error 0.081 mA in laboratory testing.
Controlled service test	Mean delay of 35.047 ms for WebSocket updates and 232.261 ms for HTTP history in a test with 50 virtual users.

These results describe their test conditions, not plant-wide delivery performance.