Field reference for grid scale battery storage technicians. 10 entries, taken from the StoreWatt app.
Reference notes from StoreWatt, the offline field toolkit for grid scale battery energy storage technicians. It works with no cell signal, because the sites do not have any.
Get StoreWatt on the App StoreCAN is the backbone of BMS communication. Physical rules: a linear bus with 120 ohm termination at the two physical ends only, twisted pair wiring, stubs as short as possible, and every node at the same bit rate (250 kbit and 500 kbit are common in BESS). With power off, a healthy bus reads about 60 ohms between CAN high and CAN low. Each message carries an identifier that doubles as its priority. Vendors publish (or guard) a database file mapping identifiers to signals; with it, a cheap USB CAN adapter turns into a powerful diagnostic window.
RS 485 carries Modbus RTU across plants and fails for the same short list of reasons everywhere. A and B polarity swapped on one device scrambles that device or the whole segment. Termination belongs at the two ends of the trunk only, roughly 120 ohms. One set of bias resistors holds the line in a defined idle state. Daisy chain topology is correct; star topology invites reflections. Ground the shield at one end only to avoid ground loops. Vendors disagree about which terminal is A and which is B. When in doubt, swap the pair on the misbehaving device before deeper surgery.
Serial Modbus over RS 485: one master polls up to 247 addressed slaves. Every device on a segment must match baud rate, parity, and stop bits exactly; a single mismatched device can disturb the whole segment. Each slave needs a unique address, and duplicated addresses produce intermittent garbage that looks like wiring trouble. Timeouts and silent intervals matter at low baud rates; an impatient master retries into slow devices and makes congestion worse.
Modbus framed over Ethernet, port 502. The serial rules disappear and network rules take over: IP addressing, subnets, switches, and firewalls. Multiple masters can poll one device, which enables monitoring but also lets a rogue poller overload a slow device. Unit identifier still matters when a gateway bridges TCP to serial devices behind it; the wrong unit id returns another device's data with no error, a nasty silent failure.
The four classic Modbus integration bugs. One: holding registers versus input registers; function codes 03 and 04 read different tables, and polling the wrong table returns zeros or junk without an error. Two: zero based versus one based addressing; register 40001 in one document is offset 0 in another, so everything lands one register off. Three: byte and word order; 32 bit values split across two registers can be swapped at either level, producing absurd magnitudes. Four: scaling factors; a value ten or one hundred times off means a missed multiplier, not a broken sensor. When a new integration reads garbage, walk these four in order; one of them is the answer far more often than not.
A device at the wrong baud rate is completely silent or produces framing errors on every poll. A parity mismatch is crueler: many polls succeed and a fraction fail, which reads as flaky wiring. Device swap outs are the classic trigger, because replacements ship at factory defaults. Rule of thumb: total silence points at baud, address, or wiring; intermittent CRC and framing errors point at parity, marginal wiring, or a duplicate address.
Plant networks are full of statically addressed devices, and a duplicate IP produces one of the most confusing failures available: two devices answer alternately as ARP caches flip, so the point flickers between two data sets or drops in and out. Laptop technicians who joined the plant subnet with a static address are a recurring cause. Subnet mask errors produce one way visibility: a device that answers pings from one host and not another. Keep the plant IP plan printed in the O&M binder; discovery by trial is how duplicates happen.
Long runs between containers, collection rooms, and the control house ride on fiber, with media converters or SFP ports at each end. Failure modes: dirty or damaged connectors, bent patch cords under cable management, failed converter power supplies, and mismatched fiber types or wavelengths after a swap. Link lights at both ends are the first check; a link up with heavy errors suggests a dirty connector, and cleaning kits solve more fiber tickets than replacement parts do.
DNP3 is the utility SCADA protocol of choice for the interconnection: it carries timestamped events, quality flags, and report by exception, which suits telemetry to a control center. On a BESS site it usually appears between the plant RTU or gateway and the utility, not inside the plant. When the utility reports bad data, check the DNP3 point map and the quality flags first; a point marked offline or restart carries its own explanation.
IEC 61850 lives in the substation: relays and merging units exchanging GOOSE messages and reports over Ethernet with strict timing. BESS technicians meet it at the plant substation boundary, usually when protection engineers are involved. Field posture: do not patch laptops into 61850 process networks casually, and treat GOOSE misconfiguration alarms as protection issues for the relay engineer, not network tickets.
Reference notes from StoreWatt, the offline field toolkit for grid scale battery energy storage technicians. It works with no cell signal, because the sites do not have any.
Get StoreWatt on the App StoreThese notes are a field aid, not a substitute for the governing codes, the stamped drawings, the authority having jurisdiction, or manufacturer manuals. Verify against the current documentation for your installed equipment.