config.ini.example

[GENERAL]
# Use ct002/ct003 for multiple storage devices; use shelly* types otherwise.
# Comma-separated list of device types to emulate (ct002, ct003, shellypro3em, shellyemg3, shellyproem50)
DEVICE_TYPE = shellypro3em
# Comma-separated list of device IDs (optional, auto-generated if not set)
# DEVICE_IDS = shellypro3em-c59b15461a21
# Skip initial powermeter test on startup
SKIP_POWERMETER_TEST = False
# The web-based configuration editor at http://<host>:<port>/config.  Left
# unset it follows DASHBOARD_ENABLED below, whose Configuration tab is that
# editor.  Set it to True to serve the editor with no dashboard, or to False
# to refuse it even with one — False also takes the dashboard's Configuration
# tab away, and is the way to keep an unauthenticated port from editing this
# file at all.
# WEB_CONFIG_ENABLED = False
# The live status dashboard at http://<host>:<port>/ (default True).
# Shows grid power, every battery's target and reported power, power-source
# health and the balancer's internal state.  Set this to False to serve only
# the health check (and /config, if WEB_CONFIG_ENABLED above is on).  Ignored
# by the Home Assistant add-on, where the dashboard always runs and opens from
# the sidebar.
# DASHBOARD_ENABLED = True
# Let the dashboard change this file and control batteries (default True).
# There is NO authentication on this port, so anyone who can reach it can use
# those controls; set this to False for a read-only page.
# DASHBOARD_ALLOW_WRITE = True
# Only relevant when this file is used BY the Home Assistant add-on: there the
# dashboard opens from the sidebar and port 52500 is refused unless you opt in
# here.  Running AstraMeter yourself (Docker or Python), the port is the only
# way in and the dashboard is already served there.
# DASHBOARD_DIRECT_ACCESS = False
# Extra host names the web port answers under, comma-separated.  IP addresses,
# localhost, .local and .home.arpa names always work, so most setups need
# nothing here.  Add a name if you reach the dashboard through one that
# resolves via a nameserver — a reverse proxy, a private DNS entry, or a
# router-assigned name such as astrameter.fritz.box.  Names are refused by
# default because a name is the one part of the address another website can
# aim at this port from your browser (DNS rebinding), and this port has no
# login to fall back on.
# DASHBOARD_ALLOWED_HOSTS = astrameter.example.lan, proxy.example.lan
# Port for the embedded web server (health check + dashboard + config editor). Default: 52500
# WEB_SERVER_PORT = 52500
# Global throttling configuration (in seconds) - applies to all powermeters unless overridden
# Enforces minimum time between power meter readings by waiting when called too frequently
# This helps prevent instability or oscillation for slower data sources
# Set to 0 to disable throttling (default)
# Recommended values: 1-3 seconds for slower data sources
# Can be overridden per powermeter section
THROTTLE_INTERVAL = 0
# Wait briefly (up to 2s) for a fresh push from event-driven powermeters
# (MQTT, Home Assistant, HomeWizard, SMA energy meter, ...) before responding
# to the battery. Keeps battery control loops on the freshest data.
# Set to false to skip the wait entirely and always serve the last-known
# value — recommended for sources that update slower than 2s (e.g. a P1 smart
# meter behind Home Assistant) or whenever you'd rather have an instant
# response than a fresh one. Defaults to true.
# Can be overridden per powermeter section.
# WAIT_FOR_NEXT_MESSAGE = true
# Ignore repeated requests from the same emulator client within this window
# (seconds). Applies to CT002/CT003 (keyed by consumer id) and Shelly (keyed
# by battery IP). Can be overridden in the [CT002]/[CT003] section. 0 disables.
# A suppressed request is left UNANSWERED, so a battery's answered rate can only
# fall to a whole fraction of its own poll rate (half, a third, ...) — the value
# is honoured exactly, but its effect changes in steps, not smoothly.
# DEDUPE_TIME_WINDOW = 0

# [CT002]
## CT type is derived from the emulated device (ct002 -> HME-4, ct003 -> HME-3).
## Note: CT003 can optionally use its own [CT003] section with the same keys.
## CT MAC (12 hex digits, from Marstek app).
## If empty, the emulator accepts any request CT MAC and echoes the request's
## CT MAC in responses. If set, the emulator responds only to the configured MAC.
# CT_MAC = 001122334455
## UDP port to bind for CT002/CT003 (default 12345).
# UDP_PORT = 12345
## WiFi RSSI reported to the storage system
# WIFI_RSSI = -50
## Ignore repeated requests from the same consumer within this window (seconds).
## Overrides the [GENERAL] DEDUPE_TIME_WINDOW for this section.
# DEDUPE_TIME_WINDOW = 0
## Forget consumers after this many seconds without updates (multi-consumer support).
## Unset (default): adaptive — a battery is dropped after missing ~2 of its own
## poll cycles (min 5s), like the real CT. Set a number for a fixed window.
# CONSUMER_TTL = 120
## Log concise status (phase consumption + consumer charge/discharge) on each request when enabled
# DEBUG_STATUS = False
## --- HTTP cloud reporting (opt-in, experimental) ---
## Mimic what a real CT does: periodically GET hamedata.com's setCtReporting with
## live grid power, the charge/discharge buckets, RSSI, slave count and link
## flags (the HME-4/HME-3 field layout follows the emulated device type). Plain
## HTTP, no TLS. OFF by default. See docs/marstek-mqtt-http.md (§6) for the field
## meanings and the open unknowns (cadence, energy/V-I values, account binding).
# CLOUD_REPORTING = False
## Host to report to (default eu.hamedata.com; other regions swap the host).
# CLOUD_REPORTING_HOST = eu.hamedata.com
## The id sent as `id`/`uid` is the CT MAC: the device AstraMeter registers in
## your Marstek account (the [MARSTEK] section) if available, otherwise CT_MAC.
## The cloud accepts reports only for a device id it already knows. (The model
## and firmware version are sent automatically as part of the handshake.)
## Seconds between reports. The real cadence isn't documented — measure it from a
## device and set it here (default 60).
# CLOUD_REPORTING_INTERVAL = 60
## --- Active control ---
## When True (default), the emulator smooths the grid reading, splits the target
## across batteries, and balances their load. When False, raw meter values are
## relayed and batteries decide on their own.
# ACTIVE_CONTROL = True

## --- Fair distribution (multi-battery balancing) ---
## Adjust each battery's target so they share the load evenly.
## Only matters with 2+ batteries.
# FAIR_DISTRIBUTION = True
## How aggressively to correct imbalance between batteries.
## 0 = no correction (equal split only); 0.3-0.5 = faster rebalancing.
# BALANCE_GAIN = 0.2
## Ignore imbalance smaller than this (W). Prevents micro-corrections. Kept
## above the battery firmware's own +-20 W input deadband so corrections the
## battery would ignore are never sent.
# BALANCE_DEADBAND = 25
## Cap on the per-cycle balance correction (W). Limits how fast a single
## battery's target can deviate from its fair share.
# MAX_CORRECTION_PER_STEP = 80
## When imbalance exceeds ERROR_BOOST_THRESHOLD (W), the balance gain is
## multiplied by up to (1 + ERROR_BOOST_MAX). For example with the defaults
## (gain=0.2, threshold=150, max=0.5): at 150 W imbalance the effective gain
## is 0.2 * 1.5 = 0.3; below the threshold gain stays at 0.2.
# ERROR_BOOST_THRESHOLD = 150
# ERROR_BOOST_MAX = 0.5
## Below this imbalance (W), the gain is scaled down proportionally for
## gentler corrections as batteries approach equilibrium.
# ERROR_REDUCE_THRESHOLD = 20
## Hard clamp on target change relative to current battery output (W). 0 disables.
# MAX_TARGET_STEP = 0
## Ramp pacing: cap each battery's per-poll command delta. The cap starts at
## PACE_BASE_STEP (matching the battery firmware ramp's first step) and
## grows toward PACE_MAX_STEP only while the battery demonstrably follows the
## command; it resets when the direction reverses. Keeps the firmware's
## accelerating ramp from overshooting on meter latency. 0 disables pacing.
# PACE_BASE_STEP = 30
# PACE_MAX_STEP = 100
## Steady-import trim (W): every Marstek firmware parks the grid a few watts to
## the import side of zero in steady state (its input deadband won't chase a
## small residual), leaving real load to be imported at the retail tariff. Once
## the grid has held steady inside a small import band for a few polls (never
## during a load step, so no added overshoot) this nudges the batteries to cover
## it, recovering that self-consumption. Stays clear of a full/empty pack. 0
## disables.
# IMPORT_TRIM_W = 15

## --- DC battery keep-alive ---
## Applies per battery (independent of balancing — also relevant with a single
## battery). Minimum discharge (W) to keep a DC battery (e.g. Marstek B2500) from
## switching its inverter off at 0 W and falling asleep under high PV surplus.
## Only affects DC-only batteries that have no inverter of their own (Venus and
## Jupiter are unaffected); 0 disables (default). Can also be set per battery from
## Home Assistant. Recommended >= 20.
# MIN_DC_OUTPUT = 0

## --- Saturation detection (full/empty battery handling) ---
## Track how well each battery follows its target. When a battery can't deliver
## (full or empty), its share is reduced and redistributed to others.
# SATURATION_DETECTION = True
## EMA factor for the saturation score. Lower = slower to declare saturated.
# SATURATION_ALPHA = 0.15
## Ignore saturation tracking when target is below this (W).
# MIN_TARGET_FOR_SATURATION = 20

## --- Efficiency optimization (opt-in low-demand power concentration + probe handoff) ---
## At low total demand, concentrates power on fewer batteries to avoid
## inefficient low-power operation. Each active battery gets at least this
## many watts. 0 disables (default). Batteries rotate for fairness, and newly
## promoted batteries are probed before the previous active battery fades out.
## Example: 2 batteries, 200W demand, threshold 150 → one gets 200W, other 0W.
# MIN_EFFICIENT_POWER = 0
## Seconds between rotating which battery has priority (minimum 10).
# EFFICIENCY_ROTATION_INTERVAL = 900
## Minimum target (W) sent when probing a newly promoted battery.
## Batteries that ignore tiny commands need a higher floor (default 80).
# PROBE_MIN_POWER = 80
## How quickly the old battery fades out after a successful probe
## (0.01=very slow, 1.0=instant).
# EFFICIENCY_FADE_ALPHA = 0.15
## Force-rotate when a battery can't follow its target (e.g. externally limited
## to 0 W or fully discharged). After a probe succeeds, this still swaps out an
## active battery within a few seconds instead of waiting for the full rotation
## interval. 0 disables. The saturation EMA is time-weighted, so slow
## powermeters (>10 s updates) accumulate saturation faster per sample; raise
## this value (e.g. 0.8) if you see unnecessary swaps with a slow meter.
# EFFICIENCY_SATURATION_THRESHOLD = 0.4
## Smooths the household-demand estimate that decides how many batteries stay
## active, so meter noise can't flip a battery in and out of the active pool
## every few readings (which wears the setpoints with no benefit — a battery
## can't track noise anyway). Lower = smoother/slower to react to a sustained
## load change; 1.0 disables the smoothing (react to every reading). Only
## relevant with MIN_EFFICIENT_POWER set and two or more batteries.
# EFFICIENCY_DEMAND_ALPHA = 0.1
## How quickly a swapped-out battery becomes eligible again. Applied each cycle
## while the battery has no target. 1.0 = never recover, lower = faster recovery.
# SATURATION_DECAY_FACTOR = 0.995
## Probe window for a newly promoted battery. During this time the previous
## active battery stays online as backup and covers the residual shortfall until
## the promoted battery shows meaningful real output.
# SATURATION_GRACE_SECONDS = 90
## Stall escape for non-probe grace cases such as batteries rejoining auto mode.
## Probe handoffs use the probe window above as the primary timer.
# SATURATION_STALL_TIMEOUT_SECONDS = 60

# [MARSTEK]
## Optional: auto-register managed fake CT devices in Marstek cloud on startup.
# ENABLE = False
# BASE_URL = https://eu.hamedata.com
# MAILBOX = your@email.example
# PASSWORD = your_password
## Prefix is fixed to 02b250 (locally administered MAC space) for managed fake devices.
## Used for add-device request
# TIMEZONE = Europe/Berlin
## Optional: live grid in the Marstek mobile app over MQTT uses [MQTT_INSIGHTS]
## on the same broker as hame-relay (https://github.com/tomquist/hame-relay) >= 1.3.5
## (HA discovery does not depend on this).

# [SHELLY]
# TYPE = 1PM #PLUS1PM #EM #3EM #3EMPRO
# IP = 192.168.1.100
# USER = username
# PASS = password
# METER_INDEX = meter1
## Per-powermeter throttling override (optional)
## Shelly devices are typically fast, so throttling may not be needed
# THROTTLE_INTERVAL = 0
## Per-powermeter smoothing (optional, defaults come from [GENERAL])
## EMA alpha (0.0-1.0). Higher = tracks faster; lower = filters noise.
## 0.9 works well at >= 1 Hz; reduce toward 0.3 for slower powermeters.
# SMOOTH_TARGET_ALPHA = 0.9
## Return zeros when |grid total| < DEADBAND (W). 10-30 W sensible; 0 disables.
# DEADBAND = 20
## Max watts the smoothed target may change per cycle (slew-rate limit). 0 disables.
# MAX_SMOOTH_STEP = 0

# [TASMOTA]
# IP = 192.168.1.101
# USER = tasmota_user
# PASS = tasmota_pass
# JSON_STATUS = StatusSNS
# JSON_PAYLOAD_MQTT_PREFIX = SML
# JSON_POWER_MQTT_LABEL = Power
## For 3-phase meters, use comma-separated labels:
# JSON_POWER_MQTT_LABEL = Power_L1,Power_L2,Power_L3
# JSON_POWER_INPUT_MQTT_LABEL = Power1
# JSON_POWER_OUTPUT_MQTT_LABEL = Power2
# JSON_POWER_CALCULATE = True
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [SHRDZM]
# IP = 192.168.1.102
# USER = shrdzm_user
# PASS = shrdzm_pass
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [EMLOG]
# IP = 192.168.1.103
# METER_INDEX = 0
# JSON_POWER_CALCULATE = True
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [IOBROKER]
# IP = 192.168.1.104
# PORT = 8087
# CURRENT_POWER_ALIAS = Alias.0.power
# POWER_CALCULATE = True
# POWER_INPUT_ALIAS = Alias.0.power_in
# POWER_OUTPUT_ALIAS = Alias.0.power_out
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [HOMEASSISTANT]
# IP = 192.168.1.105
# PORT = 8123
# HTTPS = True
# ACCESSTOKEN = YOUR_ACCESS_TOKEN

## Single sensor mode - comma-separated list for multiple phases
# CURRENT_POWER_ENTITY = sensor.phase1_power, sensor.phase2_power, sensor.phase3_power

## Or for separate input/output sensors (use entity IDs)
# POWER_CALCULATE = True
# POWER_INPUT_ALIAS = sensor.phase1_input, sensor.phase2_input, sensor.phase3_input
# POWER_OUTPUT_ALIAS = sensor.phase1_output, sensor.phase2_output, sensor.phase3_output

## Per-powermeter throttling override (optional)
## HomeAssistant typically needs 2-3 seconds due to network latency
# THROTTLE_INTERVAL = 2
## Skip the up-to-2s wait for a fresh push from this powermeter and always
## serve the last-known value. Useful when the underlying sensor (e.g. P1
## smart meter) updates slower than 2s, so the wait would always time out
## and the cached value would be served anyway. Defaults to the global
## [GENERAL] WAIT_FOR_NEXT_MESSAGE setting (true).
# WAIT_FOR_NEXT_MESSAGE = false

# [VZLOGGER]
# IP = 192.168.1.106
# PORT = 8080
# UUID = your-uuid
## For 3-phase meters, use comma-separated UUIDs (one per phase):
# UUID = uuid-l1, uuid-l2, uuid-l3
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [ESPHOME]
# IP = 192.168.1.107
# PORT = 6052
# DOMAIN = your_domain
# ID = your_id
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [ESPHOMENATIVE]
# ADDRESS = powermeter.local
# PORT = 6053
# API_KEY = 5BqtR16i91/+rwUl+QrJewKFOnyS/whHc3v9ySSKpb8=
# OBJECT_ID = grid_power
## Connection Info string (optional)
# CLIENT_INFO = AstraMeter

# [AMIS_READER]
# IP = 192.168.1.108
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [MODBUS]
# HOST = 192.168.1.100
# PORT = 502
# UNIT_ID = 1
# ADDRESS = 0
# COUNT = 1
# DATA_TYPE = UINT16
# BYTE_ORDER = BIG
# WORD_ORDER = BIG
# REGISTER_TYPE = HOLDING
## Transport: TCP (default) or UDP
# TRANSPORT = TCP
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [MQTT]
# BROKER = broker.example.com
# PORT = 1883
## Optional: enable TLS to connect over mqtts:// (default false)
# TLS = false
## Alternatively, provide a single MQTT URI in place of BROKER/PORT/USERNAME/PASSWORD/TLS:
##   mqtt://user:pass@broker.example.com:1883
##   mqtts://user:pass@broker.example.com:8883
# URI = mqtt://user:pass@broker.example.com:1883
# TOPIC = home/powermeter
# JSON_PATH = path.to.value (Optional)
# USERNAME = mqtt_user (Optional)
# PASSWORD = mqtt_pass (Optional)
## Per-powermeter throttling override (optional)
## MQTT can have variable latency depending on broker and network
# THROTTLE_INTERVAL = 1

## Multi-phase: replace the [MQTT] section above with one of these alternatives
## Option 1: one topic per phase (comma-separated)
# [MQTT]
# BROKER = broker.example.com
# TOPICS = home/power/l1, home/power/l2, home/power/l3

## Option 2: single topic with multiple JSON paths
# [MQTT]
# BROKER = broker.example.com
# TOPIC = home/powermeter
# JSON_PATHS = $.phases[0].power, $.phases[1].power, $.phases[2].power

# [JSON_HTTP]
# URL = http://example.com/api
# JSON_PATHS = $.power
## JSON_PATHS supports jsonpath-ng extensions; strip a unit suffix like
## "331.74 W" with `split` or `sub`:
##JSON_PATHS = $.state.`split( , 0, -1)`
##JSON_PATHS = $.state.`sub(/[^0-9.\-]+$/, )`
# USERNAME = user
# PASSWORD = pass
# HEADERS = Authorization: Bearer token

# [TQ_EM]
# IP = 192.168.1.100
# PASSWORD = secret (Optional)
# TIMEOUT = 5.0 (Optional)
# THROTTLE_INTERVAL = 1

# [HOMEWIZARD]
## HomeWizard P1 Meter via WebSocket API (v2)
## Connects over WSS and subscribes to real-time measurement updates
## Token must be obtained once via POST /api/user while pressing the device button
## See: https://api-documentation.homewizard.com/docs/v2/
# IP = 192.168.1.110
# TOKEN = YOUR_32_CHAR_HEX_TOKEN
# SERIAL = your_device_serial
## Verify server TLS using the bundled HomeWizard CA (default True).
## Set False only if you get certificate errors on a trusted LAN (insecure).
# VERIFY_SSL = True
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 0

# [ENVOY]
## Enphase IQ Gateway (Envoy) via local HTTPS API
## Reads grid power from /production.json?details=1 (net-consumption).
## Auto-detects single- vs three-phase from the response.
# HOST = 192.168.1.120
## Option A: long-lived JWT token from https://entrez.enphaseenergy.com/
# TOKEN = eyJ...
## Option B: let AstraMeter obtain a token via the Enphase Enlighten cloud
##   (auto-refreshes on 401; not compatible with MFA-enabled Enlighten accounts)
# USERNAME = you@example.com
# PASSWORD = your-enphase-password
# SERIAL = 123456789012
## Envoy uses a self-signed certificate; verification is disabled by default.
## Affects only the local Envoy connection — cloud requests always verify TLS.
# VERIFY_SSL = False

# [SMA_ENERGY_METER]
## SMA Energy Meter / Sunny Home Manager via Speedwire multicast
## Listens for UDP multicast broadcasts and reports per-phase power (L1, L2, L3)
# MULTICAST_GROUP = 239.12.255.254
# PORT = 9522
## Serial number of the meter (0 = auto-detect first meter found)
# SERIAL_NUMBER = 0
## Network interface IP to bind to (optional, leave empty for all interfaces)
# INTERFACE =
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 0

# [FRITZ]
## AVM FRITZ!Smart Energy 250 smart-meter read head, via the FRITZ!Box
## AHA-HTTP-Interface (getdevicelistinfos). The read head pairs with the
## FRITZ!Box over DECT; AstraMeter polls the box for its current power.
# HOST = fritz.box
## A FRITZ!Box user with Smart Home permission (Home Network > FRITZ!Box Users).
# USER = smarthome
# PASSWORD = your-fritzbox-password
## AIN of the read head as shown under Home Network > SmartHome (with or without
## the blank). The signed grid-import (-1) branch is read by default; append
## -1 (Strombezug) or -2 (Einspeisung) explicitly to pick a branch.
# AIN = 12345 0123456
## Reach the FRITZ!Box over HTTPS (default False = http). Self-signed certs are
## common, so set VERIFY_SSL = False if HTTPS fails verification.
# HTTPS = False
# VERIFY_SSL = True
# TIMEOUT = 10.0
## IMPORTANT: power the read head over USB. On battery it only refreshes every
## ~2 min — too slow for AstraMeter's ~1 s control loop, so battery balancing
## won't work. USB raises the update rate to ~10 s; this throttle matches that
## so AstraMeter doesn't hammer the box between fresh readings.
# THROTTLE_INTERVAL = 10

# [FRONIUS]
## Fronius Smart Meter via a Fronius inverter's local Solar API.
## Reads the signed PowerReal_P_Sum (positive = grid import, negative = feed-in)
## from GetMeterRealtimeData.cgi. No token/login needed on the local network.
# IP = 192.168.1.130
## Solar API meter device id; 0 is the first/only meter (default).
# DEVICE_ID = 0
## Report per-phase power (PowerReal_P_Phase_1..3) as L1/L2/L3 instead of the
## aggregate sum. Only enable if your meter reports SIGNED per-phase power;
## some firmwares report it unsigned, which breaks export readings.
# PER_PHASE = True
## If readings are reversed, flip the sign:
# POWER_MULTIPLIER = -1

# [REFOSS]
## Refoss / Meross energy monitor (EM01P, EM06P, EM16P) via local Open API.
## Polls Em.Status.Get and reads signed channel power (W): positive = import,
## negative = feed-in. [MEROSS] is an alias for the same driver.
## The device API is cleartext HTTP — use only on a trusted local network.
# IP = 192.168.1.150
## CT channel id(s). One id for single-phase; three for L1/L2/L3 (e.g. 1,2,3).
# CHANNELS = 1
## Prefer a numeric IP in Docker — *.local mDNS often does not resolve in bridge mode.
# POWER_MULTIPLIER = -1

# [TIBBER_PULSE]
## Tibber Pulse read locally via the Pulse Bridge (no Tibber cloud). Takes each
## SML telegram live as the bridge pushes it (ws://<bridge>/ws), polling
## /node_data.json (/data.json on older firmware) wherever push isn't available.
## Enable the bridge's local webserver first: params page >
## webserver-force-enable = true > Store params.
# IP = 192.168.1.140
## Username is always "admin"; password is the nine-character code printed on the
## bridge (with the dash), e.g. AD56-54BA.
# PASSWORD = AD56-54BA
## Optional: node id (see http://<bridge>/nodes/), defaults to 1.
# NODE_ID = 1
## Optional: override the Basic-auth user (defaults to "admin").
# USER = admin
## Optional: request timeout in seconds (default 5). The bridge's webserver can
## be slow to respond — raise this if readings drop with connection timeouts.
# TIMEOUT = 5.0
## Optional: poll over HTTP instead of using push (default False). Push already
## falls back to polling on its own when the bridge can't push.
# FORCE_POLLING = False
## Optional OBIS overrides (12 hex digits; omit to use eHZ-style defaults)
# OBIS_POWER_CURRENT = 0100100700ff
# OBIS_POWER_L1 = 0100240700ff
# OBIS_POWER_L2 = 0100380700ff
# OBIS_POWER_L3 = 01004c0700ff
## When polling, throttle to avoid hammering the bridge's slow webserver.
# THROTTLE_INTERVAL = 2
## If readings are reversed, flip the sign:
# POWER_MULTIPLIER = -1

# [SCRIPT]
# COMMAND = /path/to/your/script.sh
## Per-powermeter throttling override (optional)
# THROTTLE_INTERVAL = 1

# [SML]
# SERIAL = /dev/ttyUSB0
# Required: device path to the serial interface.
## Optional OBIS overrides (12 hex digits; omit to use eHZ-style defaults)
# OBIS_POWER_CURRENT = 0100100700ff
# OBIS_POWER_L1 = 0100240700ff
# OBIS_POWER_L2 = 0100380700ff
# OBIS_POWER_L3 = 01004c0700ff

## --- PID controller (grid-balance feedback loop) ---
## When PID_KP > 0, a Proportional-Integral-Derivative controller is layered
## on top of the configured powermeter.  It uses the measured grid power as its
## process variable and steers the reported value toward zero (net-zero grid).
##
## Recommended starting point (P-only, bias mode):
##   PID_KP         = 0.5   # gain; 0.5 is a safe start for one storage device
##   PID_KI         = 0     # integral — usually not needed; risks windup
##   PID_KD         = 0     # derivative — noisy on real meters; leave at 0
##   PID_OUTPUT_MAX = 800   # cap the PID output at ± 800 W
##   PID_MODE       = bias  # add PID output to raw reading (alternative: replace)
##
## In bias mode the PID and the storage device's own closed-loop controller
## act together; the system is stable for 0 < Kp < 1.
## In replace mode the PID output replaces the raw reading entirely.
##
## To keep a small import safety buffer (prevent accidental export), combine
## with a negative POWER_OFFSET applied before the PID:
##   POWER_OFFSET   = -20   # bias 20 W toward import
##
## Parameters can be set globally in [GENERAL] or per powermeter section.
## Per-section values override the global ones.
# PID_KP = 0.5
# PID_KI = 0
# PID_KD = 0
# PID_OUTPUT_MAX = 800
# PID_MODE = bias

## --- Hampel outlier filter (optional) ---
## Stateful rolling-median rejection of wild samples (e.g. MQTT/WiFi glitches).
## Outliers are replaced with the window median; normal samples pass through
## unchanged.  Disabled by default.
##
## When HAMPEL_WINDOW > 0, the wrapper is inserted after throttling and before
## EMA smoothing.  This is orthogonal to the EMA: Hampel rejects impulse noise,
## the EMA smooths the cleaned signal.
##
## A *sustained* change (solar arriving, an oven switching on) is held back
## for about half a window before it is adopted — it is indistinguishable
## from a spike until enough samples agree.  A bigger window rejects noise
## better and adopts real changes more slowly; that is the whole trade.
##
## Recommended starting point for flaky HTTP/MQTT sources:
##   HAMPEL_WINDOW        = 5    # odd sizes recommended; 5–7 works well
##   HAMPEL_N_SIGMA       = 3.0  # reject beyond ~3σ (MAD-derived)
##   HAMPEL_MIN_THRESHOLD = 50   # watts; floor for constant-signal (MAD=0) case
##
## Parameters can be set globally in [GENERAL] or per powermeter section.
## Per-section values override the global ones.
# HAMPEL_WINDOW = 5
# HAMPEL_N_SIGMA = 3.0
# HAMPEL_MIN_THRESHOLD = 50

## --- MQTT Insights (optional) ---
## Mainly for Home Assistant: publishes internal state (grid power, targets,
## saturation, consumer topology) to MQTT with HA Device Discovery.
## When running as a Home Assistant app with Mosquitto, this is auto-configured.
## With HA Discovery on, each battery also exposes config entities you set live
## from Home Assistant — Manual Target, Auto Target, Active, and a per-battery
## Distribution Weight (default 1.0) that biases the load split across batteries
## (e.g. 1.5 vs 1.0 for ~60:40). No extra keys needed; these persist across
## restarts via retained MQTT.
## Optional add-on: Marstek app poll replies use the same broker (see MARSTEK_* below).
# [MQTT_INSIGHTS]
# BROKER = 192.168.1.100
# PORT = 1883
# USERNAME = mqtt_user
# PASSWORD = mqtt_pass
## Enable TLS (default: false)
# TLS = false
## Alternatively, provide a single MQTT URI in place of BROKER/PORT/USERNAME/PASSWORD/TLS:
##   mqtt://user:pass@broker.example.com:1883
##   mqtts://user:pass@broker.example.com:8883
# URI = mqtt://user:pass@192.168.1.100:1883
## Base topic for all MQTT Insights messages (default: astrameter)
# BASE_TOPIC = astrameter
## Enable Home Assistant MQTT Device Discovery (default: true)
# HA_DISCOVERY = true
## HA discovery prefix (default: homeassistant)
# HA_DISCOVERY_PREFIX = homeassistant
## Optional Marstek mobile app: answer CT002/CT003 MQTT polls on this broker (default: true).
## Live readings in the app need hame-relay (https://github.com/tomquist/hame-relay) >= 1.3.5
## on the same broker; older relays may not forward poll/replies. Requires [MARSTEK] so the
## topic MAC matches the managed fake device. Set false for HA MQTT Insights only.
# MARSTEK_MQTT_ENABLED = true
## Seconds between optional Marstek aggregate (cd=1) broadcasts when the app is quiet; 0 = polls only.
# MARSTEK_MQTT_INTERVAL = 300
## Smallest gap between two state publishes for the same battery (seconds); 0 = every poll (default).
## A battery polls about once a second, so without this its state topic goes out at that rate and
## every subscriber pays to parse it. Raising it trades resolution in Home Assistant's history for a
## quieter broker. Availability, removals, discovery and commands are never throttled.
# STATE_THROTTLE_INTERVAL = 0
## Seconds between per-powermeter "Online" diagnostic sensor updates (default: 30); 0 disables it.
## Each configured powermeter section gets an "AstraMeter Powermeter <section>" device with an
## Online binary sensor that flips off when the source stops delivering fresh/usable readings.
# POWERMETER_HEALTH_INTERVAL = 30