ChargeGrid
EV charging network operator console — twenty-six stations, eighty-four connectors, fault prediction before failure.
Overview
The console an operator buys, not the app a driver downloads. Live connector status across the whole network, sessions and revenue, and a fault model that predicts failures on connectors that still work — which is the only kind of maintenance that saves money rather than spending it. Predictions are correlated site-wide first, so a grid problem is never reported as four hardware faults.
Specifications
USE CASE
Network operations, predictive maintenance and dynamic pricing
WORKFLOW
Ingest, monitor, predict, dispatch, price
OUTPUT
Fault predictions with evidence and a window; suggested pricing tariffs
SECURITY
OCPP-authenticated telemetry, role-scoped dispatch approval
Features
- 01Eighty-four-cell live connector grid where the letter carries the state, not just the fill
- 02Fault prediction with the telemetry evidence and a days-to-failure window
- 03Technician dispatch approval with the revenue at risk stated
- 04Per-station utilisation forecast with suggested pricing windows
- 05Driver reviews with model-assigned sentiment and theme chips
Architecture
CHARGEGRID PIPELINE
────────────────────────────────────────────
┌────────────────┐
│ OCPP telemetry │
└────────────────┘
↓
┌────────────────────────┐
│ Status + session store │
└────────────────────────┘
↓
╭┈┈┈┈┈┈┈┈┈┈┈┈┈╮
┊ Fault model ┊
╰┈┈┈┈┈┈┈┈┈┈┈┈┈╯
↓
╭┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈╮
┊ Utilisation forecast ┊
╰┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈┈╯
↓
┌────────────────┐ ┌─────────────────┐ ┌─────────┐
│ Dispatch queue │ │ Pricing windows │ │ Reviews │
└────────────────┘ └─────────────────┘ └─────────┘
↓
┌──────────────────┐
│ Operator console │
└──────────────────┘
╭┈┈╮ dashed = model step ┌──┐ deterministic stepWhat it will not do
Every screen carries one honest failure record with its recovery path. This is the one buyers remember. A run predicted contactor failure on four connectors at one truck stop and recommended taking all four offline. The voltage variance was a grid-side problem across the whole industrial estate. Predictions are now correlated site-wide first. A second run declined to propose a standing price rise off a two-day public event — event days are excluded from the pricing baseline.
