What Is Cold Storage?
I got asked this question at a trade show last year. A guy from a European ESS integrator wanted to know why his supplier was charging extra for "cold chain logistics." He thought it was a scam. It was not.
Lithium cells degrade when sitting on a shelf. The hotter the warehouse, the faster they lose capacity. This is not a small effect. I have seen test data from a cell manufacturer showing 2-3% capacity loss over six months at 35°C versus maybe half a percent at 20°C. The cells were not even being used. Just sitting there.
Cold storage is exactly what it sounds like. You put the cells in a temperature-controlled warehouse, keep it somewhere around 15-25°C, and control humidity. The side reactions inside the cell slow down. SEI layer grows slower. You lose less lithium.
After the summer of 2022, nobody in China argues about whether cold storage matters. Eastern China hit 40°C for weeks. I heard from someone at a Jiangsu pack maker that a couple of cell suppliers had to write off inventory because the cells came out of storage with noticeably lower capacity than spec. Whether that story is exaggerated, I cannot say, but the timing lines up with when cold storage upgrades started showing up in everyone's capex plans.
SOC Before Storage
You cannot just throw cells into cold storage at any state of charge. Full charge is bad for the cathode. The layered oxide structure in high-nickel NMC does not like sitting at 4.2V for months. Something about irreversible phase changes. I am not a materials scientist, so I will not pretend I fully understand the mechanism, but the result is measurable capacity loss.
Empty cells are worse. If voltage drops too low, the copper current collector starts dissolving. You charge the cell back up and the copper deposits somewhere it should not. Internal resistance goes up. Sometimes the cell just fails outright.
The sweet spot is somewhere between 30% and 50% SOC. BYD apparently requires 40% plus or minus 5% for Blade cells going into storage. I have not seen the actual document but multiple people have mentioned this number. LFP is tricky because the voltage curve is so flat. Hard to know exactly what SOC you are at just from measuring voltage. Some manufacturers do a full charge-discharge cycle before storage just to calibrate.
Temperature Specs Vary
LFP handles cold better than ternary. Gotion lists storage temperature down to negative 20°C on their datasheet. High-nickel NCA and NCM are pickier. Samsung SDI NCA cells want 15-25°C if I remember the spec correctly. Go below 10°C and the electrolyte gets too viscous. Affects formation quality apparently.
SVOLT has even tighter requirements on their short-blade cells. Something around 18-23°C with only 2 degrees tolerance either way. Not easy to maintain year-round in southern China where summer outdoor temps hit 40. Compressors run nonstop. Electricity bills get ugly. One warehouse operator I talked to said their July power bill was over 100,000 RMB and that was considered normal for a mid-sized facility.
Some manufacturers quietly relax the upper limit to 28°C during peak summer. Then they run an extra OCV screen before shipping to catch any cells that degraded too much. Not ideal but the economics are what they are.
The Condensation Problem
Move a cold cell into a warm room and water condenses on the surface. Everyone knows this. What not everyone realizes is how much this matters for pouch cells. Water on the aluminum laminate film is not just a cosmetic issue. It gets into the tab welds. Causes oxidation. In bad cases moisture works its way through the heat seal and reacts with the electrolyte.
The proper way to handle this is a buffer zone. Cells sit in an intermediate temperature room for four to six hours before moving to the production floor. Temperatures equalize gradually. No condensation.
Smaller operations skip this step. I visited one pack assembly shop where they were pulling cells straight from a 5°C cooler into a 30°C workshop. Middle of August. Every cell surface was wet. The production manager admitted their module welding reject rate was running high that month. He blamed the welding equipment. I had my own theory.
How Long Can You Store
Six months is the common limit. After that you need to retest. OCV, IR, capacity, the whole suite. Cells that fail get downgraded or scrapped. EVE supposedly requires cells stored beyond nine months to go back to the factory for re-grading. I have not confirmed this directly but it came from someone who would know.
This became a real headache when the ESS market slowed down in late 2023. Integrators were sitting on inventory they could not move. Projects kept getting delayed. By the time deals finally closed, cells had exceeded their storage window. Cue months of arguments between integrators and cell suppliers about whose problem this was. Purchase contracts now often include language requiring cells to be manufactured within 60 or 90 days of delivery. Lesson learned.
Fire and Insurance
Standard warehouse sprinklers do not work well on lithium battery fires. Water hitting a cell in thermal runaway can make things worse. Generates hydrogen. The preferred systems now are perfluorohexanone or fine water mist, but a perfluorohexanone setup costs millions of RMB. A lot of warehouses cannot justify that spend.
Insurance has become the bigger constraint. CPIC and Ping An both jacked up rates for lithium battery storage starting around 2024. Three to four times what a normal warehouse pays, from what I have heard. Some insurers will not touch these projects at all. Without insurance coverage, good luck getting bank financing. Everything downstream gets stuck.
Day-to-Day Operations
Cells need periodic voltage checks while in storage. Normal self-discharge is a few millivolts per month. If you see a cell dropping 15mV or more, something is probably wrong internally. Micro short circuit most likely. Pull it out and deal with it separately.
European customers want to see temperature logs for the entire storage period. Some warehouses attach data loggers to each pallet that travel with the shipment. Customer pulls the data at receiving. One out-of-spec excursion on that log and they might reject the whole batch. Happened to a shipper I know. Forty-foot container turned around at the port.
FIFO should be simple. First in, first out. In practice, forklift drivers grab whatever pallet is easiest to reach. The stuff in the back of the rack sits there for months until someone notices. MES systems help if you have them. Many warehouses still run on spreadsheets and hope for the best.
