Your inverter says 100 percent, then 40 an hour later

The reading is not broken. The inverter is estimating state of charge from terminal voltage, and on a lithium iron phosphate pack that estimate is close to useless through the middle of the range.

A lead acid battery drops voltage steadily as it empties, which is what made voltage based fuel gauges work for decades. LiFePO4 does not do that. It sits near 52 volts from roughly 90 percent down to roughly 20 percent, then falls off a cliff. Every point in that flat region looks the same to an inverter reading terminal voltage.

What the jump actually is

Under load, voltage sags. Release the load and it recovers. An inverter watching voltage alone reads the sag as depletion and the recovery as a refill, so the percentage jumps around in step with your loads rather than with the energy left in the pack.

The reading is worst exactly when you care most, which is the evening, with the largest loads running and the least solar coming in.

The fix is a conversation, not a setting

Closed loop communication replaces the guess with a number the BMS calculates by counting coulombs in and out. The BMS knows what went in this morning and what has come out since. It reports that figure to the inverter over CAN or RS485, and the display stops lying.

Getting there needs three things: a battery with a comms port, an inverter with a matching port, and a protocol both units agree on. The third is where most attempts stall, because sharing a bus standard is not the same as sharing a message format.

If closed loop is not available

Set voltage limits conservatively and stop treating the percentage as data. A shunt based battery monitor wired at the negative terminal counts current directly and costs less than a hundred dollars, which is the cheapest honest answer for any open loop system.

Related pages on this site

• A pairing with documented closed loop
• Batteries that publish a comms protocol
• How the communication check works