An independent concept rebuild of apparent.com, unaffiliated with Apparent Inc. Built by Daniel Ramirez as an application for the Software Engineer role.

Read the letter

Technology

igOS

The brain of the Apparent platform.

Meet the Intelligent Grid Operating System — a cloud-based, machine-learning energy platform that senses, analyses and acts on real-time grid data. It is the intelligence behind REMI, our Realtime Energy Management Interface.

Not just monitoring. Continuous, active optimisation.

Most energy management platforms do load following: they observe and react on timescales of seconds to minutes. igOS operates on a fundamentally different architecture. It senses and dispatches bidirectionally multiple times per second, which gives it the ability to produce the exact waveform of energy demanded by any load, or by the grid itself.

That sub-second capability unlocks services no other DER platform can offer: true reactive power production, dynamic voltage support and frequency response — capabilities previously exclusive to spinning fossil-fuel generators.

The claim, running

Sub-second, in the sense that it is happening now.

The paragraph above makes a specific engineering claim. This is that claim as executable code: a dispatch state machine and an IEEE 1547 volt-var loop, stepping on a fixed timestep four times a second, independent of the display refresh rate.

Novato feeder · 2 MW PV · 2 MWh storage
11:16 · 0 cycles @ 4 Hz
PCC voltage · per unit 0.9969
P · real ±300.0 kW 327 W
Q · reactive ±1.00 Mvar -40.2 kvar
PV
1.76 MW
Building
1.48 MW
Battery
-277.8 kW
Inverter Q
592.3 kvar
PjQ
P
327 W
Q
-40.2 kvar
|S|
40.2 kVA
cos φ
0.008 lag
φ
-89.5°
full scale
100.0 kVA
Dispatch Charging from surplus
52.1% SoC

Why a fixed timestep

A loop pinned to requestAnimationFrame is an animation, not a controller: its interval changes with the monitor, the GPU and whether the laptop is on battery, and every integrator inside it silently changes gain along with it. Time is kept in an accumulator instead, so the physics steps identically at 60 Hz, at 120 Hz, and in a headless test with no display at all.

Droop first, integrator second

The volt-var law is the piecewise-linear characteristic from IEEE 1547-2018 clause 5.3.3, with a PI trim on top. Droop gives an instant, stable, autonomous response with nothing to wind up; the trim removes the steady-state offset that pure droop always leaves behind. The deadband between them is what stops a feeder full of inverters fighting each other over measurement noise.

The circle is not negotiable

An inverter is limited by its apparent-power rating, so its operating envelope is a circle in the P-Q plane and real power has already claimed part of it: Q_max = √(S² − P²). Every var command is clipped to that circle before it reaches the plant, which is exactly why oversizing the inverter is how you buy reactive headroom.

Architecture

Six layers of intelligence.

  1. 01

    Grid edge intelligence

    Autonomous, dynamic balancing of behind-the-meter generation and load. igOS reacts to real-time energy fluctuations, stabilising voltage and preventing grid disruption at the source.

  2. 02

    Real-time energy management

    Sub-second bidirectional control across all distributed resources. Storage, solar, EV charging and building loads are orchestrated simultaneously, not sequentially.

  3. 03

    Digitisation and grid visibility

    Every action igOS takes generates a data packet, giving utilities and grid operators live visibility into DER assets at the building level rather than at the substation.

  4. 04

    Flexible and scalable

    Plug-and-play integration of new resources — solar arrays, batteries, EV fleets — without additional infrastructure investment. It scales from a single building to a regional portfolio.

  5. 05

    Energy market participation

    Aggregation of DER portfolios into ISO ancillary service markets, wholesale energy trading and utility demand response programmes, generating new revenue from existing assets.

  6. 06

    Machine learning and forecasting

    ML applied to load profiles, weather patterns and pricing data to forecast demand and optimise use. Higher solar production becomes an opportunity rather than a stability concern.