INNOPOWERDIGITAL
CASE STUDY · SOLAR O&M

FusionSolar fleet analytics

Monthly yield reporting across a multi-site solar portfolio had become a manual Excel ritual — three days of copying data, chasing inverter serials, and reformatting charts for management. Inverter-level anomalies were routinely missed until the next month's review. We extended Huawei's FusionSolar portal into a custom analytics layer that does the whole job in minutes.

React 18ViteNode.jsExpressHuawei FusionSolar Northbound APIExcel export
KEY METRICS
Monthly reporting
3 days → 10 min
Anomaly detection
Inverter-level
Sites consolidated
Multi-site portfolio
/ 01

The problem

Huawei's FusionSolar portal is excellent for real-time monitoring of a single station, but it doesn't answer the questions a fleet owner actually cares about: how is yield tracking month-over-month across the portfolio, which inverters are underperforming against their siblings, and what's the gap between PV yield and billed energy across sites? Extracting that information meant a three-day Excel exercise every month, and the anomalies that mattered most — individual underperforming inverters — were buried until someone happened to look.

/ 02

What we built

A dashboard layered on top of FusionSolar's Northbound API. The frontend is a React application that consolidates station-level KPIs across the whole portfolio, with drill-down into inverter yield comparisons, monthly accumulated reports with variance columns, and one-click Excel export in the format management already reads.

The anomaly-detection view is the piece operators ask for first: inverters are benchmarked against their station-mates and flagged when their yield diverges beyond a configurable threshold. What used to be invisible until month-end is now surfaced within hours.

/ 03

The API dance

The Northbound API is permissioned in ways that aren't always obvious — we spent real time navigating failCode 20010 restrictions on device-level endpoints and ultimately built our aggregations primarily on station-level KPIs with targeted device serial number queries for the views that required inverter granularity. If you've worked with FusionSolar you'll recognise this; if you haven't, trust that the 'simple' solution the docs imply is rarely available in production. We know the workarounds.

/ 04

What we'd do differently

We'd invest earlier in a local caching layer. The Northbound API has rate limits that bite hard during month-end reporting, when multiple users are pulling full-portfolio views simultaneously. A nightly sync into a local time-series store would eliminate that class of pain entirely, and we've added it to the roadmap for the next iteration.

/ 05

Why it matters for outbound

Most of our enterprise prospects have already invested in monitoring tools — FusionSolar, SolarEdge, Envision, or custom SCADA. The gap isn't telemetry; it's the management-ready layer on top. If you're selling into solar asset owners, understanding that gap — and being able to show you've closed it before — is the difference between a generic pitch and a conversation the buyer actually wants to have.

NEXT CASE STUDY
AI document classification pipeline
READY TO BUILD SOMETHING LIKE THIS?
Start with a 30-minute pipeline audit.
Book your audit