GridBrain

The platform

Everything You Need

A complete platform for microgrid operators and energy communities: collection, control, compliance, and billing in one system rather than four.

Real-Time Monitoring

Live battery state of energy, solar production, grid power, and demand streamed to your dashboard as it happens. Per-circuit breakdowns come straight from revenue-grade metering, so what you see is what the meter saw.

SSE over Redis pub/sub

AI-Powered Control

A K-nearest-neighbors model learns how your specific array behaves, then combines that with a 48-hour weather forecast to predict tomorrow's generation. GridBrain uses that curve to decide when to lean on the grid and when to coast on sun.

KNN + 48h forecast + SOE trajectory

Fleet Management

One dashboard across every site you operate, with health, state of energy, and open issues at a glance. Role-based access means a site manager sees their site and an operator sees the whole portfolio.

Multi-tenant, role-based access

Billing & Analytics

Per-unit energy billing for residential and commercial tenants, with PPA rate modeling, historical consumption, and exportable reports. The numbers you bill on are the same readings the control loop ran on.

Residential, commercial, and PPA rates

Zero-Export Control

Automatic PV curtailment over Modbus with dead-band control keeps you inside your interconnection agreement and stops reverse power flow before it starts. If the meter stops answering, the site falls back to a safe state rather than guessing.

Modbus curtailment, dead-band, safe-state

Alerts & Notifications

Push notifications through the installable web app, SMS for the events that matter at 2am, and a full audit log for everything else. Alerts carry the reading that triggered them, so you start diagnosing instead of hunting.

PWA push, SMS, audit logging

The product

What Operators Actually See

Screens from the running platform. The fleet shown is sample data — real client sites stay private.

Fleet overview: a map of microgrid sites above a table of live solar, grid, battery, load and state-of-energy readings — Every site on one screen

Solar, grid, battery, load and state of energy for the whole fleet, each row live and each site one click away. Status is derived from the last reading, not from a checkbox someone forgot to tick.

Historical chart showing a day of solar production, site demand, grid power and battery state of energy — A day, reconstructed

Production, demand, grid draw and state of energy on one axis pair, down to five-minute resolution. Scrub any window from six hours to a year, and take the CSV when you need to argue with a utility.

Single-site dashboard: power flow cards for solar, grid, battery, generator and load, above the raw telemetry stream — Power flow, then the raw numbers

Where the energy is going right now, with the unedited telemetry underneath it — pack watt-hours, per-phase volts, frequencies, inverter counts. The summary is for the glance; the stream is for the argument.

No vendor lock-in

Hardware Agnostic

GridBrain isn't locked to a single vendor. Connect the equipment you already have, or mix and match as your site evolves. Adding a device family means writing a collector, not replacing a switchboard.

🔋

Battery Storage

Tesla Powerwall, Powerpack, and Megapack through the local gateway API: state of energy, battery, grid and solar power, island mode, and grid codes. Extensible to any REST or Modbus-accessible BESS.

☀️

Solar Inverters

SMA Sunny Boy and Tripower over authenticated HTTPS plus Modbus TCP, Fronius through the Solar API, and SolarEdge over SunSpec Modbus with scale-factor decoding. AC power, value codes, reset, and curtailment.

⚡

Meters & Sensors

eGauge revenue-grade meters with instant, cumulative, and per-circuit readings, generic Modbus energy meters, ControlByWeb relay I/O, and any HTTP or MQTT data source you can point at us.

🖥️

Edge Compute

The agent runs on ordinary Linux or Windows hardware: Raspberry Pi, Intel NUC, industrial PC, or a virtual machine. No proprietary gateway, no dedicated rack, no truck roll to swap a box you already own.

🔌

Open Protocols

Modbus TCP, SunSpec, REST, MQTT, and HTTP. New hardware is a collector plugin against a documented interface, so support arrives as a software release rather than a firmware campaign.

🔄

Retrofit Ready

Layer GridBrain onto an existing installation without replacing equipment. It runs alongside legacy SCADA, building management systems, and utility interconnects instead of demanding you rip them out.

Getting running

How It Works

From installation to autonomous operation in four steps.

Deploy Edge Agent

Install a lightweight Python agent on-site. It connects to your Tesla gateway, inverters, meters, and relays over the local network, and authenticates to the cloud with a per-site API key. Modbus inverters are rediscovered automatically if their addresses change.

Configure Site

Set battery thresholds, relay schedules, curtailment limits, and control parameters from the fleet settings dashboard. No config files on the box.

AI Takes Over

Once commissioned, the forecaster learns your array, projects state of energy forward, and optimizes grid import continuously. You set the limits it operates within, and you can take manual control at any time.

Review & Adjust

Track battery state of energy, solar production, grid import, and site demand across the fleet. Alerts fire when a site drifts, and you adjust thresholds, relay schedules, and curtailment limits as seasons and tenants change.

Who installs it X10 Engineers commissions the first site alongside your electrician. Later sites can be done by your own team.

What you need A Linux or Windows machine on the same network as your equipment, plus outbound HTTPS. No proprietary gateway to buy.

Downtime None to install. The agent reads before it writes, and autonomous control stays off until you switch it on.

Who stays in charge You set the limits it works within. Manual override is always available, and the site keeps running if the cloud is unreachable.

Where it runs

Built for Real Sites

The same platform covers a single home battery and a multi-tenant commercial campus. What changes is the configuration, not the software.

Multi-tenant property

Bill tenants from the meter, not a spreadsheet

Shared solar and storage across many units means somebody has to answer who used what. GridBrain reads per-circuit consumption and produces per-unit bills on residential, commercial, or PPA rates.

Per-unit consumption history and exportable statements
Separate residential and commercial rate structures
PPA modeling for third-party owned generation

Commercial and industrial

Stay inside the interconnection agreement

Zero-export sites fail expensively and quietly. Continuous metering with Modbus curtailment and dead-band control holds export at zero, and a meter that stops responding drives the site to a safe state instead of a guess.

Automatic PV curtailment with dead-band hysteresis
Meter failure detection with automatic safe-state
Relay scheduling for demand management

Remote and island sites

Keep running when the link drops

The edge agent is designed for bad connectivity. Control decisions are made locally, readings queue to an on-disk buffer while the uplink is down, and the queue drains when it returns. Nothing is lost to a flaky tunnel.

Local control loop independent of the cloud
On-disk retry queue for telemetry during outages
Watchdog recovery and automatic failover

Operators and fleets

Run twenty sites like one

Fleet view ranks sites by what needs attention rather than alphabetically. Drill from portfolio health into a single inverter's value codes without changing tools or logging in again.

Portfolio health, state of energy, and open issues at a glance
Per-site edge agents with centralized configuration
Full audit trail of every control action and login

Under the hood

Built for the Field

Production-grade infrastructure for real-world microgrid operations, from the cloud down to the Modbus register.

The part that matters to you is the middle row. The edge agent keeps controlling your site whether or not the cloud is reachable, buffers readings locally while the link is down, and sends them when it returns. Losing your internet connection does not mean losing control of your batteries.

Cloud Platform

Multi-tenant fleet management with role-based access
Real-time telemetry streaming over Server-Sent Events, backed by Redis pub/sub
Automated billing runs and historical data export
Installable progressive web app with push notifications
Django and PostgreSQL behind a Cloudflare Tunnel, no inbound ports

Edge Intelligence

Lightweight Python agent runs on ordinary Linux or Windows hardware
Local Modbus TCP control of SMA, Fronius, and SunSpec inverters
Tesla Gateway API integration, Powerwall through Megapack
On-disk SQLite retry queue so telemetry survives an outage
Automatic failover, watchdog recovery, and single-instance locking

AI Engine

K-nearest-neighbors solar generation prediction trained on your site
48-hour weather forecast integration with cloud cover factors
State-of-energy trajectory simulation before a decision is taken
Autonomous islanding and grid import optimization
Scheduled forecast refresh through Celery Beat

Safety & Compliance

Zero-export PV curtailment with dead-band control
Meter failure detection with automatic safe-state fallback
Controlled shutdown, with control resumed automatically once the site is back
Comprehensive audit logging of control actions and access
Per-site API keys, revocable independently

Questions

Common Questions

The things operators ask before they deploy.

Do I have to replace my existing equipment?

No. GridBrain is a retrofit layer. It reads and controls the Tesla gateways, inverters, meters, and relays already on site over their own protocols, and runs alongside existing SCADA or building management systems. If your hardware speaks Modbus TCP, SunSpec, REST, MQTT, or HTTP, it is already in scope.

What happens when the internet goes down?

The site keeps running. Control decisions are made by the edge agent locally, not in the cloud, so curtailment and battery management continue through an outage. Telemetry queues to an on-disk SQLite buffer and drains to the cloud when the link returns, so you do not lose the history either.

What hardware does the edge agent need?

Very little. It is a Python service that runs on ordinary Linux or Windows hardware with network access to your equipment: a Raspberry Pi, an Intel NUC, an industrial PC, or a virtual machine on hardware you already have. There is no proprietary gateway to buy.

How does the solar forecast actually work?

A K-nearest-neighbors model is trained on your site's own generation history, so it learns your array, your shading, and your local weather pattern rather than a generic curve. That model is combined with a 48-hour forecast and cloud cover factors to project generation, and a state-of-energy trajectory is simulated forward before any control decision is committed.

How is zero-export enforced safely?

Continuously, with hysteresis. The agent watches the meter and curtails PV over Modbus using dead-band control, which avoids the oscillation a naive threshold produces. Critically, if the meter stops answering, the site is driven to a defined safe state rather than continuing on a stale reading.

Can I bill individual tenants?

Yes. Per-circuit readings from revenue-grade eGauge metering feed per-unit billing with separate residential and commercial rate structures, plus PPA modeling for third-party owned generation. Statements and historical consumption export for your accounting system.

Ready to Go Autonomous?

Tell us what hardware is on your site and we will show you GridBrain running against it. Existing customers can sign in to their fleet.

Request a demo Sign in to your fleet

AI Energy Management System