Magnetics for Battery Energy Storage PCS: Loss at Part Load and Thermal Reality

Storage Converters Have a Different Magnetics Problem

A grid-tied battery energy storage system (PCS) shares its power stage family with solar inverters, but its duty profile is opposite: it must charge and discharge at high current for hours, ride through grid events, and stay efficient at part load for the majority of its life. That profile changes what matters in the magnetics — not peak efficiency at one operating point, but loss across a wide current range plus thermal stability in a container that can reach 50 °C ambient.

Three components dominate: the DC-DC stage inductor or transformer, the grid-side filter inductor, and the common mode choke at the grid interface.

Typical Configuration and Component Roles

Common PCS architectures and the magnetics each requires
Architecture	DC-DC stage	Grid interface	Dominant constraint
Two-stage, isolated	Isolation transformer (DAB, LLC, CLLC) at 100–500 kHz	LCL filter inductor + CMC	Transformer loss at partial load and PD at 800 V
Two-stage, non-isolated	Bidirectional buck-boost inductor	LCL filter inductor + CMC	Inductor DC bias at high current, ripple at low current
Single-stage / transformerless	Combined bridge inductor	Filter inductor and CMC	Core saturation under grid fault current

Battery voltages of 600–1,000 V and grid currents of 100 A and above mean that a 1 % efficiency loss in the inductor equals roughly a kilowatt of heat inside a sealed container. Thermal design, not magnetic design, becomes the limiting factor.

Loss at Part Load Is the Real Specification

Storage systems spend most of their operating hours between 20 % and 60 % of rated power, following grid frequency regulation or peak shaving profiles. A magnetic design optimized at full load may show poor efficiency in that band because core loss dominates when current is low and flux swing is small. Two practical consequences:

Choose a core material whose loss curve is flat across the flux range the application actually uses, and check loss at 30 % load rather than only at rated flux.
Size the winding for the weighted-average current, not the peak, so copper loss at part load stays low; a larger window with fewer turns often beats a compact design on annual energy.

A useful design check is to compute the weighted loss over the site’s dispatch profile rather than a single point. Two designs with the same full-load efficiency can differ by 20 % in annual losses.

Grid Interface: Filter Inductors and Common Mode Chokes

LCL filter inductors carry the full grid current with a small ripple, so their design is dominated by DC bias and by the harmonic spectrum the converter injects. Above 100 A, foil windings and distributed-gap powder cores are common; the trade is size against loss. The CMC handles conducted emissions from 150 kHz to 30 MHz and is usually a nanocrystalline toroid with split windings, because high impedance per turn matters more than saturation headroom.

Both must survive grid faults: a short-circuit event can push the filter inductor far past its rated current for tens of milliseconds. Specify the fault current and duration in the RFQ, and confirm the core does not saturate into a destructive state.

Container Thermal Reality

PCS containers are often liquid-cooled at the semiconductor level but air-cooled for magnetics, with internal ambient 45–55 °C. A magnetic component designed for a 40 °C bench will derate to a lower current in that environment, which shows up as a system that cannot hold rated power in summer. Specify the internal ambient explicitly, require the supplier’s temperature-rise data at that ambient, and ask for the hotspot location in the drawing.