In development · pilot programme forming

Energy that thinks ahead.

Enertia is an energy management system for solar, battery and grid sites. It plans against tomorrow’s sun, load and tariff, and it is built to control whichever inverter brand you already have, from a controller on site.

Plan · next 48 h · 30-min intervals Illustrative data
Illustrative 48-hour energy plan A synthetic example of a forward plan: forecast solar generation and site load over two days, shaded by peak, standard and off-peak tariff periods. The planned battery schedule charges from surplus solar and off-peak grid, and discharges through peak periods so that grid import stays below a demand cap. Not real site data. DEMAND CAP BATTERY SCHEDULE ▲ DISCHARGE ▼ CHARGE NOW 06:00 +6H 12:00 +12H 18:00 +18H 00:00 +24H 06:00 +30H 12:00 +36H 18:00 +42H 00:00 +48H
Solar forecast Site load forecast Planned grid import Battery Peak tariff Synthetic example. Tariff windows approximate a weekday time-of-use schedule.
The problem

Most sites are run by rules written for yesterday.

  1. A

    Locked to one brand

    The capable controllers usually come from the inverter maker and only drive that maker’s kit. Mixed sites, expansions and brand changes end up with several apps and no single plan.

  2. B

    Reacting to the present

    “If battery above 80%, do this.” Fixed rules cannot see a cloudy afternoon coming, an evening peak tariff, or the demand peak you already set in winter.

  3. C

    Controlled from somewhere else

    Cloud-only control is at the mercy of latency, rate limits and outages. When the connection drops, so does the strategy.

The approach

Four ideas, one controller.

We mark each one with where it stands today. Enertia is in development and we would rather you knew exactly what exists.

01 Abstraction built · first vendor adapter in progress

Speaks to everything

One device abstraction sits between the intelligence and the hardware. Each inverter, battery and meter is described by what it can actually do: its capabilities, its levers and its limits. The optimiser never needs to know the brand.

02 Planned

Learns the site

No commissioning spreadsheets. Enertia is designed to find out what it is connected to and which levers really work, then keep fitted models of load, solar yield, usable battery capacity and round-trip efficiency, and re-fit them as the site changes.

03 Planned · rules-based strategies come first

Looks ahead

Control against forecasts, not the current instant: solar yield, learned load, the tariff calendar, load-shedding schedules and the demand you have already registered this billing year. The plan is re-solved every few minutes over a 24 to 48 hour horizon.

04 Hardware in design

Lives on site

A fanless, DIN-rail controller that runs the control loop locally and keeps working when the internet or the cloud does not. The cloud handles the fleet, forecast feeds and reporting. It does not hold the site hostage.

Why forecasting matters in South Africa

One bad half-hour. Twelve months of cost.

On many commercial tariffs, the demand you pay for is the higher of your notified maximum demand or the highest half-hour you registered over a rolling twelve months. Miss a single interval and that number sticks.

So the cost is lopsided. Being a little conservative costs a little, and it costs it once. Overshooting can set the charge for a year. Enertia is designed around that asymmetry: reserves, caps and battery plans are sized knowing that one excursion is far more expensive than caution.

How one peak sets twelve months of demand charges Illustrative bar chart of monthly maximum demand over a year. One month spikes above the rest, and a horizontal line from that spike shows the chargeable demand held at that level for the following twelve months. J F M A M J J A S O N D CHARGED AT THIS PEAK FOR 12 MONTHS MONTHLY MAX DEMAND · ILLUSTRATIVE
Engineering stance

Shadow first. Armed later.

An EMS writes to equipment that carries real money and real people. We build it so it can prove itself before it is trusted.

  • 01

    Shadow by default

    A new controller starts in shadow mode. It reads, plans and records what it would have done, and writes nothing to the plant until someone explicitly arms it.

  • 02

    Every cycle explained

    Each control cycle leaves a decision record: the inputs it saw, the constraint that bound, and the action it took or would have taken, with the reason.

  • 03

    Actuation verified, not assumed

    A vendor cloud can accept a command and quietly apply something else. Enertia reads back what the equipment actually did, and treats a mismatch as a fault to alarm on. It does not re-send and hope.

  • 04

    Stale data is refused

    If telemetry is too old to trust, the loop declines to act on it rather than steering a site on yesterday’s numbers.

decision-log · shadow
cycle 14:30:00  mode=SHADOW
inputs     fresh (age 18 s)
site_load  71.2 kW   pv 12.4 kW
binding    demand_cap 46.0 kW
would_do   battery discharge 12.8 kW
wrote      nothing  (not armed)
reason     import 58.8 kW would breach cap
cycle 14:35:00  mode=SHADOW
inputs     stale (age 412 s)
would_do   refuse cycle, hold last safe state
Illustrative only: shows the kind of record each cycle leaves. Not output from a real site.
Where we are

Built in order, one proven stage at a time.

The control core is being built in deliberate order. Everything risky is retired against a simulator and recorded site data before live equipment is touched.

  1. Built

    Canonical energy vocabulary

    Units, sign convention and constraints that every part of the system shares.

  2. Built

    Telemetry foundation

    Half-hour, billing-aligned rollups with 13 months of history for rolling-peak demand.

  3. Built

    Capability and lever model

    The vendor-neutral core, property-tested before any real vendor was connected.

  4. Built

    Simulator with fault injection

    A simulated vendor with lopsided levers and realistic failures, as a zero-risk test bed.

  5. Built

    Control loop and shadow mode

    A fixed-cadence loop that refuses stale data and explains every cycle.

  6. Built

    Peak-shave and night-charge strategies

    Replayed against recorded days from a real commercial site.

  7. In progress

    First vendor adapter

    Sigenergy, with actuation read-back and governed-command detection.

  8. Planned

    Failsafe and arming gate, then a pilot shadow run

    A full billing month in shadow before any staged, reversible cutover.

  9. Planned

    Second vendor · edge hardware · forecasting and optimisation · fleet platform

    The route from one proven site to a product.

Pilot programme

Put a site in the pilot.

We are looking for a small number of South African sites with solar, batteries or both, where tariffs and demand charges carry real money. Pilot sites run Enertia in shadow mode first: it plans and explains, and it controls nothing until you agree to switch it on.