English | Deutsch

SLEMS is a Home Assistant integration that manages home batteries and controllable consumers. Its goal is to operate the batteries in a grid friendly way, e.g. by shifting the charging time so that the daily feed-in peak is absorbed, based on forecasts of consumption and PV production.

The name is a combination of Slug (our "favourite pets" and a funny word for a whole range of very interesting animals) and EMS (energy management system).

Status: beta. SLEMS has been running in production without problems for quite some time; new features and improvements keep coming often. In the operating mode Active it sends commands to batteries and switches consumers. Use it at your own risk: check the limits of your batteries and the fuses of your circuits. Please report faults as an issue with the diagnostics (see Quick start).

If you would like to use SLEMS with another battery, please open an issue with the model and how it is connected (Modbus, local API, Home Assistant integration), and whether you could test it.

• Does SLEMS fit your installation?
• Quick start
• Why SLEMS?
• Features
• Installation
  • HACS (custom repository)
  • Manual
• Setup
  • Smart meter over Modbus (optional)
  • Tariffs (optional)
  • PV forecast
  • Weather (optional)
  • Consumption forecast
• Without batteries
• Batteries
• Consumers
• Dashboard
• Operation and control
  • Operating mode
  • Control and options
• Language
• Roadmap
• Development
• License

What you need:

• Home Assistant 2026.10 or later and HACS.
• A smart meter in Home Assistant that reports the grid power (import and export) continuously, e.g. a Shelly Pro 3EM. The more often it reports, the more precisely SLEMS controls (e.g. every second).
• Optional: a PV forecast (e.g. Forecast.Solar), batteries and consumers that SLEMS should control.

In five steps:

1. Install SLEMS through HACS and restart Home Assistant (Installation).
2. Add the integration, select the smart meter and the PV systems and check the sign of the grid power (Setup).
3. Add batteries and consumers (Batteries, Consumers). Most batteries allow only one Modbus connection: disconnect them from other integrations first.
4. Let SLEMS run in the operating mode Observe only for a while (the default), so obvious problems show up early. The dashboard shows the forecast and the plan; nothing is controlled yet.
5. Turn off other controls of the same batteries and consumers (automations, other integrations) and set the operating mode to Active. New batteries first run their self-test (about 5–10 minutes), then SLEMS takes over.

The defaults are fine to start with; SLEMS learns losses, response times and forecast buffers by itself (see Learned values). Hints on problems are under Problems and notifications. For a fault report, attach the diagnostics: Settings → Devices & services → SLEMS → ⋮ → Download diagnostics. Addresses and serial numbers of the batteries are redacted in them.

Most home battery controls react to the moment: they keep the grid power at zero and charge whenever there is surplus. The battery is then full by late morning, and the midday peak goes into the grid anyway. SLEMS plans ahead and controls precisely:

• Grid friendly instead of full by 10 o'clock. From the PV and consumption forecasts SLEMS calculates a feed-in limit and charges the batteries with the surplus above it, so they absorb the feed-in peak and are still full in the evening. The limit is recalculated continuously and drops by itself when the day turns out worse than forecast.
• Its own consumption forecast. Learned from the long-term statistics of your house, with the weather and a separate heat pump model; vacation mode, night discharge to a forecast based reserve and a projection of the state of charge for today and tomorrow.
• Batteries and consumers in one plan. The surplus is distributed between batteries and controllable consumers (heating rod, heat pump, …) with priorities, battery priority until the charge is secured, minimum runtimes, external blocking and thermostat pauses. Daily targets (runtime, energy or temperature by a time of day) are met from the surplus first and only if needed from the batteries or the grid; per consumer you decide whether the batteries may supply it.
• Several batteries at their efficient point. SLEMS learns the conversion losses of every battery, runs as few batteries as useful and switches between them with a smooth transition.
• Feed-in cap. If you may feed in only part of the PV power (e.g. 60 % or nothing at all), SLEMS stores the energy above the limit and makes room in the batteries in time instead of curtailing PV.
• Import peaks shaved economically. With power based grid fees (a power price for the highest quarter hour of the month) SLEMS picks the import limit and the reserve with the lowest expected cost, keeps energy in the batteries for the peaks, lets forced consumers give way and tops the batteries up below the limit before the peaks.
• Prices in view. Tariffs as on the bill, with templates for Austria, Germany and Switzerland; day-ahead prices per quarter hour only with your consent, estimated until the next day's prices are published. The price aware control keeps the stored energy for expensive hours and can optionally charge cheaply from the grid or feed in when it pays; a monthly tariff comparison shows what another tariff would have cost.
• Precise control without oscillation. Event driven on every report of the smart meter; the learned response time of the batteries (also of starts out of standby and reversals) is compensated and the control gain adapts itself.
• Battery care and safety. State of charge window, power limits (e.g. 800 W), temperature charge limit, cell delta monitoring with active cell balancing, detection of batteries that do not respond, repair issues and notifications.
• Regular full charge without losing surplus. LFP batteries recalibrate their state of charge only when full. SLEMS charges the battery whose last full charge is the oldest first, one at a time and only with the surplus it would store anyway, and spares it when discharging so it starts the next day higher. It also recognises a full charge the BMS ends slightly below 100 %, and it measures the cell delta at the top.
• Learns instead of asking. Besides the consumption forecast, SLEMS learns the conversion losses and the response time of the batteries, the smart meter interval, the control gain and the consumption peaks; on request also the forecast buffers, the grid surplus targets, the control timing, the usable capacity of every battery and the power and thermostat behaviour of consumers (see Learned values). Every learned value can be switched back to a fixed one.
• Transparent and local. A dashboard with energy flow, forecast chart and all settings. Control, forecasts and learning run locally in Home Assistant, without a cloud account. SLEMS goes to the internet only for market prices and monthly market values (only with your consent), for Swiss tariffs from the open data of ElCom and once to suggest the region for public holidays (Nominatim, see Consumption forecast). The operating mode Observe only shows what SLEMS would do before it controls anything, also next to an existing battery integration.
• Try before you change. The simulation in the dashboard calculates the day plan with other settings (grid friendly charging, night discharge, feed-in cap, battery limits, forecasts ±%) and compares feed-in, grid import and state of charge with the current plan, without changing anything.
• Also without batteries. SLEMS can direct the PV surplus into consumers alone (heating rod, wallbox, dehumidifier, …), with daily targets, tariffs and forecasts (see without batteries).

Other batteries and electric cars: batteries that SLEMS does not yet support directly or through a Modbus profile can still be controlled if there is an integration for them in Home Assistant: SLEMS then controls them through its entities (experimental). SLEMS controls a wallbox with the surplus, with Charge now and with a daily target by the car's state of charge. evcc specialises in EV charging and complements SLEMS well (see the evcc guide).

The button opens SLEMS in HACS of your Home Assistant instance and adds the custom repository; then continue with step 2. Or by hand:

1. HACS → ⋮ → Custom repositories → add https://github.com/gojux/SLEMS, type Integration.
2. Install SLEMS and restart Home Assistant.

Copy custom_components/slems into the custom_components folder of your Home Assistant configuration and restart Home Assistant.

Requires Home Assistant 2026.10 or newer.

1. Settings → Devices & services → Add integration → SLEMS.
2. Select the devices of your smart meter and your PV systems (optional, preselected from the energy dashboard of Home Assistant; the devices that fit best come first, e.g. a Shelly Pro 3EM or the meter of an inverter). SLEMS suggests their sensors and fills in what the energy dashboard already knows, including the sign of its grid power, the PV forecast entries and the weather entity. Check the grid power entity (positive = import, negative = export; the form shows its current value; enable invert if your meter uses the opposite sign), and optionally PV power, PV forecast and weather. Several PV systems (e.g. roof and balcony) are summed up.
3. On the SLEMS integration page choose Add battery for every battery and Add consumer for every consumer you want to measure or control.

Phases (optional): if your smart meter measures each phase, you can select its Grid power L1, L2 and L3 in the SLEMS settings. SLEMS then learns which phase each device is on and shows it; this does not affect the control yet (see Phase of batteries and consumers).

The entity of the smart meter is updated once per polling cycle of its integration, often only about once per second, and then shows only a part of the meter's values. With Also read the smart meter over Modbus SLEMS reads the grid power directly from a SunSpec meter every 0.5 seconds (0.2 to 5 s). The control sees every value sooner; the gain is largest with fast batteries.

• SLEMS uses the shared Modbus connection of Home Assistant. As long as the inverter integration keeps the inverter's only Modbus TCP connection to itself, a Modbus proxy has to be set up, e.g. the app Modbus Proxy. SLEMS asks the shared connection for a timeout of 1 second; if another integration on the same device needs longer, the longer one applies. A meter that does not answer is reconnected at once, which other integrations on the same device notice as a short reconnect.
• SLEMS finds the meters in the SunSpec model chain; with several meters (e.g. export+import and consumption) you choose the one at the grid connection point, the current power of each helps to recognise it.
• The sign is found automatically by comparing a few values with the grid power entity; this needs some import or export (at least 100 W). Otherwise choose it by hand.
• The entity stays required: it is the history, the basis of the forecasts and the fallback. While no Modbus value is newer than 3 intervals (at least 5 seconds), SLEMS uses the entity; after 5 minutes without Modbus a repair issue appears. The sensor Smart meter update interval shows the source as attribute source.
• When the source is switched, the learned battery response times start again, since they contain the delay of the old source.

With a tariff SLEMS works in money: it checks your electricity bill, shows every month what another tariff would have cost and can control the batteries by the prices.

• Enter a tariff as on the bill, line by line, from templates with public list prices (Austria, Germany, Switzerland through ElCom) or as YAML, also created by an AI from a bill.

• Dynamic tariffs with day-ahead prices per quarter hour (fetched only when you switch it on) and power prices for the monthly peak.

• Price aware control (optional): keeps the batteries' energy for expensive hours, on request charges cheaply from the grid or feeds in when it pays well, and measures what that saved.

  

All settings and details: Tariffs and prices.

Every integration that provides a solar forecast for the Home Assistant energy dashboard can be selected, e.g. Forecast.Solar or Solcast. Several entries (e.g. one Forecast.Solar entry per roof plane) are summed up. SLEMS uses the finest resolution the provider delivers (e.g. 15 or 30 minutes). Forecast.Solar marks each period by its end; SLEMS takes this into account.

Recommendation: Helios Forecast (open source, no account or API key, computed locally with weather data from Open-Meteo). It learns from the production of your own PV system, including shading by trees or mountains, soiling and an orientation that is a few degrees off, and uses the existing long-term statistics of your energy sensor right away. It provides a forecast for the energy dashboard in 15 minute steps, so it can be selected in SLEMS directly.

The weather entity improves the consumption forecast, especially for a heat pump. The forecast of the national weather service fits best:

Met.no, the Home Assistant default, works everywhere but is less detailed in alpine terrain.

SLEMS learns your house's consumption from the long-term statistics of Home Assistant and forecasts today and tomorrow by the hour: heat pumps by the outdoor temperature, public holidays like weekends. When the consumption rises suddenly (e.g. at the start of the heating season), SLEMS raises the forecast at once. The card Forecast accuracy shows how well the consumption and PV forecasts hit. When switching from another battery integration, choose its house consumption sensor as House consumption history, then SLEMS learns from the whole history at once.

Details: Consumption forecast.

SLEMS also works without any battery: add the consumers only.

• Works: directing the PV surplus into controlled consumers (heating rod, wallbox also with evcc, dehumidifier, heat pump through a switch input) with priorities, minimum runtime and pause, Avoid short runs and external blocking; daily targets with the sources surplus only and + grid (also in the cheapest window with price aware control); tariffs, bill check, market prices, tariff comparison and price chart; energy flow, consumption and PV forecast with their accuracy.
• Not without batteries: everything that charges, discharges or plans a battery – grid friendly charging, night discharge, import peak shaving, price hold, charging from the grid, feeding in from the batteries – and currently the feed-in cap (it needs a battery even with supporting consumers).

SLEMS controls one or more home batteries: it distributes the power between them, plans their use over the day (grid friendly charging, night discharge, prices, import peaks) and protects them along the way. A battery can be added in three ways:

• Marstek Venus E 3.0 directly over Modbus TCP: the fastest way, with the most data (cell voltages, counters, firmware). SLEMS searches the network for it. The Venus allows only one Modbus connection.
• Existing Home Assistant entities (control experimental): any battery that another integration brings in, e.g. through Omnibattery, also a balcony storage or hybrid inverter with its own PV modules.
• Battery from a Modbus profile (experimental): a battery with Modbus TCP, described by a profile an AI creates from its manual.

What SLEMS brings for every battery:

• A self-test when it is added, checking signs, values and response.
• Learned values: response time, phase and efficiency. With several batteries only as many run as is efficient, with a smooth switch.
• Limits and protection: state of charge, charge and discharge power, minimum power; gentle charging near full, cell delta and cell balancing, a pause for firmware updates.
• Optional wear costs, so price aware charging from the grid only happens when it earns more.

All settings and details: Batteries and Battery from a Modbus profile.

SLEMS steers the PV surplus into controllable consumers and plans them together with the batteries. Add the consumers SLEMS controls (heating rod, wallbox, dehumidifier, …) and heat pumps, whose consumption it forecasts from the weather. Each needs its own power and energy sensor; SLEMS suggests them from the device.

• Control: on/off, a power set point in W, a current set point in A (wallbox, also with evcc) or a release for an automation of your own, e.g. for the hot water charge of a heat pump.
• Priorities, minimum runtime and pause, avoid short runs for devices with a compressor and an external block; SLEMS recognises a device's own thermostat by itself.
• Battery support: per consumer, whether the batteries may supply it.
• Daily targets: runtime, energy, temperature or the car's state of charge by a time of day, from the surplus where possible and only if needed from the batteries or the grid.
• Feed-in cap: consumers help take the surplus above the limit.

All settings and details: Consumers, plus the guides to the hot water of a heat pump and the wallbox with evcc.

SLEMS adds the entry SLEMS to the Home Assistant sidebar:

• Overview: energy flow, key figures, the chart with the day's forecast, plan and measurements, with a tariff also the prices.
• Batteries and Consumers: one card per device with its state, plan, daily target and the most important settings.
• Simulation: try other settings and compare them with the current ones, without changing anything; also what other batteries or more PV would have changed in the recorded year.
• Settings: all settings grouped and directly editable, with the learned values.

Everything in detail: Dashboard.

The entity SLEMS Operating mode switches between:

• Off – nothing is planned or sent.
• Observe only – forecasts and plans are computed and shown, but no command is sent to batteries or consumers. This is the default.
• Active – plans are executed: SLEMS sends set points to the batteries and switches/sets the consumers. It reacts to every change of the smart meter. The entity Control status shows whether the control is active or paused because the smart meter did not report for a while (60 s or ten update intervals; the batteries then follow their own logic until the meter is back).

In the operating mode Active SLEMS controls on every new value of the smart meter: it brings the grid power to a target with the batteries, using the learned response times so no command counts twice. On top there is a plan for the day from the PV and consumption forecasts. The options can be combined:

• Grid friendly charging (on by default): the batteries charge with the surplus above a feed-in limit, so they absorb the midday peak and are still full in the afternoon.

  

• Night discharge: energy that would still be in the batteries in the morning is fed in at night, down to a reserve the PV refills the next day; useful e.g. in an energy community.

  

• Import peak shaving: the batteries keep energy for the peaks; with a power price SLEMS picks the limit and reserve with the lowest cost.

  

• Feed-in cap: if you may feed in only part of the PV power, SLEMS makes room in the batteries in time instead of curtailing PV.

  

• Bad weather mode: stores as much surplus as possible before bad weather and keeps it over the next night.

SLEMS learns many values itself (buffers, grid surplus targets, capacity, morning reserve, consumer power). Ongoing problems show up under Settings → Repairs.

Everything in detail: Operation and control, including Which option when? and Problems and notifications.

SLEMS is available in English and German. Home Assistant uses two different language settings:

• Entity names (e.g. House consumption, Operating mode, also the key figures in the dashboard) follow the server language under Settings → System → General. They are set when the integration loads, so reload SLEMS after changing it.
• Dialogs, menus, states (e.g. Observe only) and the texts of the dashboard follow the language in the user profile.

If names stay in the wrong language after an update of SLEMS, restart Home Assistant (translations are only read at startup) and reload the browser without cache. Entity IDs such as sensor.slems_house_consumption keep the language of their creation; only the displayed names change.

Possible extensions, grouped by topic. Wanted: feedback from users with a real wallbox on the evcc guide.

Electric car

• Targets that apply from plugging in the car ("80 % in 3 hours").
• Batteries with an availability (e.g. "car plugged in") and a reserve until a time of day: the basis for a car supplying the house (V2H).
• Learned presence of the car for the planning.

Planning and prices

• One planning for all battery goals: charging, discharging and feeding in per quarter hour by costs over the next 36–48 hours (prices, night discharge, feed-in cap, grid friendly charging), instead of separate rules each with its own horizon. Part of it:
  • the daily targets of the consumers planned together with the batteries,
  • a more stable plan of the price aware control (changed only when clearly better).
• Consumers preferred at very cheap or negative prices: a price threshold per consumer (e.g. the heating rod or a boiler), based on the total import price including grid fees, not only the exchange price.
• Estimated prices from the weather: the forecast of wind, solar and load of the bidding zone (Energy-Charts) moves the estimate; with them known exactly, the deviation of the last year would have dropped by about a third.

Tariffs

• Prices from a Home Assistant entity of the supplier (e.g. the Tibber integration, Octopus Germany, Ostrom): a line type with adapters for the common forecast formats, stating whether the price already includes the grid fees and levies.
• Index tariffs (e.g. on the Austrian electricity price index ÖSPI): a line type with the supplier's formula, the monthly index values fetched (with consent) from the Austrian Energy Agency; templates for index tariffs too.

Forecast

• Own forecast for vacations: a base load profile learned from whole vacation days (all vacations together) and a factor for the heat pump from the same days; until about three vacation days are learned the estimate of today.
• Program devices (washing machine, dishwasher, dryer): learn the course of a program and add the rest of a running program to the forecast of the next hours (helps the evening and night planning); later possibly starting them with surplus if the device allows a remote start.

Consumers

• SG Ready / smart grid of the heat pump: raise the set temperature with PV surplus or at cheap prices (building, buffer tank, hot water as heat storage), through the SG Ready states or an offset entity; heating and cooling ahead as a release with limits (highest temperature, offset, duration), not as a daily target. The forecast knows that the heat drawn ahead is missing later.

Batteries and installations

• Battery groups with a shared fuse: several batteries on one circuit (e.g. two Venus behind one circuit breaker) get a highest power together, which the active batteries of the group share, instead of a fixed half for each.
• Batteries with their own PV: the first real devices (Marstek B2500, Zendure SolarFlow) and hybrid inverters through their integrations or through Modbus profiles (candidates to test are available), and a shared highest output for several units at one inverter.
• Further kinds of installations: batteries with their own meter or a backup output with loads behind it, and batteries serving several inverters.
• Further battery models: shipped Modbus profiles from own profiles and candidates that passed their trial with users of these batteries; a driver of its own only for models that do not fit a profile.

See developers.md.

GPL-3.0, see LICENSE. The Marstek Venus register map and control sequence are based on Omnibattery (GPL-3.0).