docs: power_to_user does track grid import; correct the clamp note (#373)

The note against box_grid_power said a grid-bounded limit was not buildable from
this inverter's own registers on this model. That was too strong, and the
reporter corrected it against his own evidence rather than let it stand because
it quoted him.

Across 637 samples during EV charging, power_to_user (3041/3042) was non-zero in
632 of 632, median -4 W against a P1 meter, 98% within 200 W. For sustained
import it tracks the grid closely, and sustained import is the failure mode a
clamp exists to prevent.

The limitation is real but narrower: power_to_user also read 0.0 through eight
transient events, including a confirmed 2.2 kW load running ten seconds while
the inverter's own discharge register moved by 2.2 kW to meet it. Trustworthy
for sustained flow, blind to transients on roughly that timescale - which is the
normal profile of a kettle, espresso machine or boiler, so a design using it
should expect to miss those and say so rather than discover it later.

box_grid_power stays mapped and stays noted as dead on this model. That half is
firmer than when written: 4430 consecutive zero samples now, and the conclusion
no longer rests on the transient spikes at all.

Co-Authored-By: Claude <noreply@anthropic.com>

Author: 0xAHA

custom_components/growatt_modbus/profiles/mod.py

@@ -213,9 +213,8 @@
         3286: {'name': 'box_temperature',   'scale': 1,   'unit': '°C', 'signed': True, 'desc': 'NTC temperature, Int8, -40 to 100°C'},
         3287: {'name': 'box_grid_voltage',  'scale': 0.1, 'unit': 'V',  'desc': 'Grid voltage'},
         # box_grid_power has been observed constant at zero on a MOD 10KTL3-XH (DN1.0):
-        # 2509 samples over two nights, never non-zero, including 11 samples where the
-        # house meter showed real grid flow of up to 2064 W. In those same 11, power_to_user
-        # read 0 in all of them and power_to_grid in 10 of 11 (#373).
+        # 4430 consecutive samples over two nights, never once non-zero, including samples
+        # where the house meter showed real grid flow of up to 2064 W (#373).
         #
         # NOT removed on that evidence. The box's other sensors work on the same unit
         # (temperature, grid voltage, work mode, bypass status all report), so this is not
@@ -224,10 +223,25 @@
         # settle it is box_work_mode (3282) during a sample with known grid flow; nobody
         # has captured that pairing yet.
         #
-        # Practical consequence, and the reason it is recorded here: a grid-bounded limit
-        # ("never command anything that causes import") is not buildable from this
-        # inverter's own registers on this model. A PV-bounded one is — PV power at 1/2,
-        # 5/6, 9/10 is validated, as is battery power at 3178-3181.
+        # USE power_to_user (3041/3042) FOR GRID IMPORT ON THIS MODEL, not this register.
+        #
+        # An earlier version of this note claimed a grid-bounded limit ("never command
+        # anything that causes import") was not buildable from this inverter's own
+        # registers. That was too strong, and the reporter said so rather than let it
+        # stand: across 637 samples during EV charging, power_to_user was non-zero in
+        # 632 of 632, with a median difference of -4 W against a P1 meter and 98% of
+        # samples within 200 W. For sustained import it tracks the grid closely.
+        #
+        # The limitation is real but narrower than stated. power_to_user also read 0.0
+        # through eight transient events, including a confirmed 2.2 kW load running for
+        # ten seconds while the inverter's own discharge register moved by 2.2 kW to meet
+        # it. So it is trustworthy for sustained flow and blind to transients on roughly
+        # that timescale — which is the normal profile of a kettle, an espresso machine or
+        # a boiler, so any design using it should expect to miss those.
+        #
+        # For a clamp that exists to stop a *command* causing continuous import, that gap
+        # is acceptable and should be stated rather than discovered. PV power (1/2, 5/6,
+        # 9/10) and battery power (3178-3181) are also validated on this model.
         3289: {'name': 'box_grid_power_high', 'scale': 1, 'unit': '', 'pair': 3290, 'signed': True, 'desc': 'Grid power HIGH (Int32, positive=import). Reads 0 on at least one DN1.0 unit — see note above'},
         3290: {'name': 'box_grid_power_low',  'scale': 1, 'unit': '', 'pair': 3289, 'combined_scale': 0.1, 'combined_unit': 'W', 'signed': True, 'desc': 'Grid power LOW'},
         3297: {'name': 'box_load_power_high', 'scale': 1, 'unit': '', 'pair': 3298, 'desc': 'Load power HIGH (Uint32)'},