Chief Technology Officer / Singapore / Southeast Asia
Choices that cost availability, speed or code
- Plants in the fleet
- 52
- 38 island solar-and-battery plants and 14 factory battery installations
- Island battery storage
- 412 MWh
- Alongside 186 MWp of solar on Philippine and Indonesian island grids
- Company revenue, 2025
- US$118m
- Lautaneri Energy as a whole; not a technology budget
- Technology team
- 86
- Control systems, battery safety, security, platform and operations
Decisions & consequences
01Derating twenty-three battery containers after one overheated on a Visayas island
At 15:40 on a dry-season afternoon in April 2024, the cooling unit on one battery container at our plant on a Visayas island failed, and its hottest cells reached 52°C before the battery management system took the container offline. Nothing vented and nothing burned. Within four days our investigation found the same cooling model, with the same alarm setting, on twenty-three containers at nine island plants. That setting delayed the high-temperature warning by about twelve minutes. Our supply agreements commit us to available battery capacity for the evening peak, so any derating would push islands back onto diesel and cost us availability credits.
Our commercial director proposed derating only the six containers with the worst temperature records and retrofitting the rest during routine maintenance. It was a fair argument; the other seventeen had shown nothing abnormal. I decided to derate all twenty-three to 60 per cent of rated power until each had an independent temperature sensor, a second cooling unit and corrected alarm logic. A hazard that comes from a design cannot be cleared by a clean history, and I wanted our chief executive and the risk and safety committee to hear it from me that week, not after a second event.
The retrofit took fourteen weeks and cost US$2.1 million. Availability credits to five cooperatives came to US$1.3 million. The affected islands ran diesel for longer each evening, costing the cooperatives fuel and their customers' patience. I wrote to each general manager explaining why.
Cooling-related trips across the fleet fell from eleven in 2023 to one in 2025, and none has approached the April temperature. Our battery safety lead now has authority to derate equipment without commercial approval, and every new container design must pass a deliberate cooling-failure test at the factory.
Next time I would visit the five general managers in person during the first week. My letter explained the engineering but not why their evenings had to change, and two of them told me so.
Removing an inverter manufacturer's standing access to our fleet
In September 2022, while mapping every network connection into our island plants, my security lead found that the manufacturer of our main inverter model had kept its own cellular link to 410 inverters at thirty-one sites since commissioning. Through it, the manufacturer's service team could read data, change settings and push firmware without any approval from us. Nobody had concealed it. The link came with the warranty support package for the inverter we standardised on in 2018, under an engineering standard I had written, and none of us had asked what it allowed. A mistake or a compromise at a supplier's service centre could therefore change how our whole fleet behaved on grids where we carry much of the evening load.
Our operations director and the manufacturer both resisted closing it. Remote diagnosis was fast, the warranty assumed the access, and our technicians valued the help. I decided to remove the direct links and route all supplier access through a gateway in our Singapore data centre. Each session would need approval from the Cebu control room, would be recorded and would expire after four hours. Firmware would be installed first on a test bench in Cebu and released to sites in stages.
New modems and the gateway cost US$780,000. The manufacturer amended its warranty terms but added a service charge of US$160,000 a year. In the first six months, the median time to resolve a fault needing its help rose from nine hours to fifteen; it has since settled at eleven.
In May 2024 a firmware release on the bench reset reactive-power settings to factory defaults. At sixteen sites those defaults would have breached the interconnection settings agreed with the cooperatives. We held it until the manufacturer issued a correction.
Supplier access is now specified in every equipment contract before purchase, including the gateway route and its charge. I would have written that requirement into the 2018 standard, when negotiating it would have cost far less than retrofitting it.
Replacing our own site controller with an independently tested product
By early 2023 the software that decided, every few seconds, how solar, batteries and the cooperative's diesel sets shared the load at each island plant was eight years old and still largely the work of one person, our founding engineer Edwin Kusumarta. It worked well. But each new interconnection now required test evidence that the plant would respond correctly to frequency changes, limit its ramp rates and stay connected through brief voltage dips. Because our controller had no independent type-test record, every new site needed bespoke testing averaging nineteen days, and two cooperatives had asked for an outside engineer to review its logic. Every change also depended on Edwin and two colleagues.
Edwin argued for keeping it and hiring engineers to document it. He was right that it was more flexible and that licences would cost more than his team. I recommended buying a commercial hybrid-plant controller with independent test documentation for site-level control, while keeping in-house the fleet forecasting, scheduling and reporting layer that encodes each cooperative's contract. Our chief executive and the board agreed.
Licences and migration were estimated at US$3.6 million over three years, against about US$1.1 million a year for the in-house controller team. We wrote off US$1.9 million of capitalised development. Twenty-six island plants had migrated by December 2025. At two of them the new controller dispatched the battery too cautiously for three weeks, so diesel ran longer than necessary until we corrected its settings.
Interconnection testing at a new island plant now averages eight days, and no cooperative has since asked for an outside review. Edwin leads the fleet scheduling team, which is where our knowledge of the cooperatives' contracts actually lives.
With hindsight, I would run the old and new controllers in parallel for longer at the first two sites. I would also ask each cooperative's dispatchers to approve the battery behaviour before cutover rather than discovering their objections afterwards.