This project review was written by the engineers who designed and commissioned the system. It explains why this Battery Energy Storage System was built the way it was: what the customer's load profile looked like, why the existing supply was no longer comfortable, how the C&I ESS was sized, and what changed after it entered service.
The customer is a metal fabrication company in an industrial park in Binh Duong Province, Vietnam. They produce precision sheet-metal parts, laser-cut components and welded assemblies for HVAC and elevator OEMs, mostly for export. The plant runs two production shifts from Monday to Saturday, with a smaller night shift on the welding lines. The equipment base includes CNC punching and laser cutting machines, press brakes, robotic welding cells, a powder coating line and a centralized compressed air system.
We first visited after the customer received a utility quotation for a transformer upgrade and wanted a second opinion; the audit that followed became the basis of the storage project.
We started by recording the plant's real load. A temporary power analyzer on the low-voltage side of the main transformer logged three weeks of data, including two full production weeks.
During the morning shift the load sat at 450–650 kW, climbed after lunch and peaked between about 13:30 and 16:30, when the CNC machines, press brakes, compressors and the coating oven ran together. The highest 15-minute average recorded was around 980 kW, against an average daytime load of roughly 620 kW — a far wider gap than the production team had realized.
That profile had two consequences. The provincial tariff includes a demand charge based on the highest 15-minute average demand each month, so the afternoon peak inflated the fixed part of the bill. And the peak was still growing: a new laser cutting line was already on order.
The plant is supplied by a single 22/0.4 kV transformer rated 1,250 kVA. At 980 kW and a measured power factor around 0.92, it was carrying roughly 1,065 kVA — above 85% of rating. Not an immediate overload, but no headroom, and it explained two complaints from the maintenance manager.
First, voltage: during the afternoon peak the far end of the factory busbar sagged, and CNC controllers occasionally tripped on undervoltage in the hottest months. Second, capacity: the new laser line would push the transformer close to its limit within a year.
The utility proposed a new 2,000 kVA transformer and a substation reconfiguration — a long lead time and significant capital cost. We looked at the problem differently: the transformer was only stressed for two to three hours a day and had ample margin the rest of the time. That asymmetry is exactly what storage corrects, so instead of upgrading the transformer to serve a three-hour peak, we proposed an ESS that shaves the peak and leaves the existing infrastructure in place.
The agreed scope was a grid-connected Battery Energy Storage System with three functions: peak shaving in the afternoon, load shifting by charging at night when the tariff is lower, and backup for a defined list of critical loads during grid faults. The result was a 500 kW AC system with 1,075 kWh of nominal LiFePO4 capacity in five outdoor cabinets. The table summarizes the parameters; the sections below explain why each value was chosen.
| Parameter | Value |
|---|---|
| Transformer capacity (existing) | 1,250 kVA, 22/0.4 kV |
| Average daytime load | ~620 kW |
| Recorded peak load (15-min avg) | ~980 kW |
| ESS rated power | 500 kW AC (5 × 100 kW) |
| Battery capacity | 1,075 kWh nominal, LiFePO4 |
| SOC operating range | 10–90%, with configurable backup reserve |
| Charge window | 22:30–06:30, power limited by EMS |
| Discharge window | 13:30–16:30, demand cap 850 kW |
| Cooling | Air-cooled, inverter-driven fans |
| Enclosure | Outdoor cabinet, IP54, concrete plinth |
| Communication | Modbus TCP to PCS/BMS, site LAN with 4G fallback |
| Grid connection | Low-voltage, behind the meter, no export |
LiFePO4 was selected for two site-specific reasons: it tolerates the warm outdoor environment (ambient temperatures regularly reach 35–37°C) well, and the customer's electrical team had no lithium experience, so safety and predictable ageing mattered more than peak performance.
The 1,075 kWh figure was not taken from a catalogue. Holding the demand cap at 850 kW through a typical afternoon requires roughly 130 kW average discharge over two to three hours, so nominal capacity was set well above that to keep the daily cycle inside a comfortable depth of discharge. The SOC window is 10–90%: the upper limit slows ageing in hot weather, the lower limit protects against deep discharge, and the EMS adds a configurable backup reserve during working hours so peak shaving never empties the battery to the floor.
Five 100 kW / 215 kWh cabinets from our industrial and commercial ESS cabinet range replaced what would otherwise have been a container, because the only available space near the transformer compound was long and narrow. The modular layout also gives an easy expansion path: if the new laser line pushes the peak higher, one or two cabinets can be added without touching the rest of the system.
Each cabinet has its own 100 kW bi-directional PCS, 500 kW AC in total, handling grid synchronization, current control and islanding detection. Distributed conversion matches the modular battery blocks, lets one unit be isolated for service without stopping the system, and keeps AC cabling simple.
The Energy Management System (EMS), in a panel beside the cabinet row, reads plant load from current transformers on the main low-voltage incomer and directs every PCS and BMS over Modbus TCP in real time. All setpoints, schedules and the SOC reserve are adjustable remotely; a site LAN connection with 4G fallback lets the customer's electricians check the system from home and lets us review operating data without a visit.
The cabinets are air-cooled, deliberately. Operating rates stay at or below 0.5C, within air cooling capability. Liquid cooling would have bought tighter cell temperature control and a higher continuous C-rate, but it adds coolant loops, pumps and a maintenance burden the customer's small team would carry for years.
Each cabinet is IP54, mounted outdoors on a concrete plinth about 300 mm above grade, clear of rain splash and road dust. Accessible intake filters are on the customer's monthly maintenance list. Cell-level voltage and temperature monitoring runs through each BMS, with smoke and overtemperature sensors able to trigger a controlled shutdown; DC and AC sides have independent breakers and the row has a clearly marked emergency stop.
Installation was constrained by the site. The only suitable location was a strip between the transformer compound and the park road, about 30 meters of cable from the switchroom, so we poured one plinth, set the five cabinets on it with service clearance, and fed each from a new distribution board tapped after the main breaker. The plant allowed only one short production stop, so all live-busbar work was done on a Sunday, coordinated with the utility and the shift plan.
Commissioning took about ten working days: insulation and earth continuity tests, torque checks, and verification of CT polarity and phase sequence by comparing EMS readings with the utility meter over a full day. A full charge followed, letting the BMS balance cells and correct the SOC reference — if the EMS thinks the battery is fuller than it is, every dispatch decision drifts.
The protection tests mattered most to the customer. We tested anti-islanding by opening the upstream utility breaker and confirming the PCS stopped exporting within the time required by the local grid code. We then tested backup: with the main breaker open, the PCS switched to island mode and carried the critical bus — CNC controls, the compressor PLC, the coating line panel and emergency lighting. The acceptance criterion was simple: the PLCs and CNC controllers must not restart. They did not, and that test convinced the site manager this was more than a billing tool.
The system then ran in observation mode for two weeks — recording load and simulating decisions without charging — to give us a real profile to tune against.
On a normal weekday the system follows a rhythm the customer's electricians now recognize at a glance.
Charging happens at night. The EMS starts at 22:30 — deliberately after the welding night shift ends at about 22:15 — and normally finishes well before 06:30, limiting charge power to roughly 200–250 kW to keep transformer loading moderate and the battery gently handled. A typical night moves the SOC from about 15% back to the 90% ceiling: roughly 600–650 kWh.
In the morning the load stays below the cap and the ESS rests. After lunch, when the load approaches the 850 kW cap, the EMS discharges and modulates output to hold total intake at the cap, rotating dispatch between cabinets to keep states of charge balanced. On most days discharge lasts two to three hours and the battery ends the afternoon near its reserve, not at the floor.
If the grid dips or fails during working hours, the PCS units detect the abnormal condition, disconnect and form an island supplying the critical bus from the battery. When the grid returns, the system verifies voltage and frequency stability before resynchronizing and resuming its schedule. The reserve exists for this situation: the EMS never discharges below it during peak shaving, so energy is always left for a grid event, even late in the afternoon.
The system has run for several billing cycles. Storage performance claims are often overstated, so we report only what the monitoring data and the customer's records show.
The demand cap has held. Since active peak shaving began, the highest 15-minute demand on working days has stayed at the configured 850 kW, against the roughly 980 kW peak measured during the audit. The customer expected the demand charge to fall accordingly, and initial operating data indicates that expectation was realistic, though a full twelve-month comparison is needed before quoting a firm saving.
The backup function has proved itself too. Twice in the first months the feeder failed, which the utility attributed to external construction work. On the longer event the grid was absent for about 25 minutes; the ESS carried the critical bus throughout and finished above the reserve level. Not one CNC program was lost — previously a monthly risk in the storm season, the customer said.
Thermally, the highest cell temperature in the hottest week was about 42°C, inside the design limit, with fans at high speed only during the afternoon discharge. The maintenance manager noted the system has needed less attention than the air compressor next to it.
Several lessons are worth passing on to anyone planning a similar installation.
First, the load curve is the foundation. The customer said "the transformer is too small"; the measurements showed a three-hour peak on an otherwise healthy transformer. Sizing against a description instead of data is how projects get the wrong kW/kWh ratio.
Second, the tariff determines the value. This project pays because the demand charge is based on the 15-minute peak; without that, the same system would have a much longer payback — we said so during the audit.
Third, the EMS must stay adjustable. The customer has changed shift patterns twice, and each change required a small adjustment to the charge window or the cap. A schedule fixed at commissioning quietly becomes wrong within a year.
Fourth, local capability matters more than hardware features. We chose air cooling, a simple communication topology and a documented EMS partly because the customer's team will live with this system for years.
Finally, budget for utility paperwork: the grid connection approval took longer than the installation.
Every major choice traces to a site constraint. 500 kW covers the gap between the recorded peak and the target cap with margin to modulate smoothly. 1,075 kWh covers the afternoon window without deep cycling, protecting cycle life in a warm climate. LiFePO4 suits the heat and the customer's maintenance capability; air cooling suits the moderate C-rate; five modular cabinets suit the site layout and expansion plan; and grid-connected operation with backup was chosen over a full microgrid because only a defined critical bus needed island operation — the cost difference was significant.
The customer did not buy the largest system proposed, and we did not push one. We presented the sizing logic with the load data, and they chose the configuration that matched their measured peak, expansion plan and budget.
This project was not really about batteries. It was about a factory whose transformer was too small for three hours a day, whose tariff punished its afternoon peak, and whose production could not afford to lose CNC programs every time the grid blinked. A 500 kW / 1,075 kWh Battery Energy Storage System addressed all three with one piece of infrastructure.
At handover the customer said the system behaves like a predictable part of the factory's electrical infrastructure, not an experiment — for a Commercial Energy Storage installation, the best possible outcome. The plant is now discussing a Solar + Storage addition; the BESS was specified so adding PV needs only new inverter capacity and an EMS update.
The hardware described in this project comes from our standard industrial energy storage range. Customers evaluating a similar retrofit usually start with the following three items.
Tianon New Energy Technology Co., Ltd, founded in 2018, is the manufacturer behind the systems used in this project. Its product matrix covers residential lithium batteries, outdoor C&I ESS cabinets, liquid-cooled BESS, containerized energy storage and foldable solar containers, and it delivers complete MWh-level solar-battery-diesel hybrid solutions. With 80–90% of output exported and more than 3,000 customers served, the company supports every project with in-house system sizing, product integration, shipment and after-sales technical support. More information is available on our company profile and in the full product catalogue.
