The Problem
PV plant operators were managing solar installations with limited data visibility. Monitoring systems reported aggregate output metrics but could not identify which specific modules were underperforming or why. When a plant underperformed its expected output, the diagnostic process was manual: technicians travelled to site, conducted module-by-module inspections, and worked through faults without data to guide where to start.
For Sun Sniffer — a German solar technology company with proprietary hardware for module-level PV monitoring — this was the gap the product needed to close. They had hardware collecting granular efficiency data from individual panels; what they needed was a “Monitoring 2.0” platform that could turn that raw data into actionable diagnostic intelligence, not just time-series charts.
The Constraints
Dispatcher 2.0 regulatory compliance (Germany). The German energy regulatory framework requires grid-connected PV installations to meet Dispatcher 2.0 communication standards. Compliance was non-negotiable for the German market and had to be a designed-in output of the platform — not a retrofit.
Multi-tenant data isolation across installation types. The platform serves both commercial farm operators managing multiple large installations and residential customers with single installations. These use cases share a data processing pipeline but require strict tenant isolation — one customer must never have visibility into another’s installation data, regardless of account tier.
Hardware variability across deployments. Sun Sniffer’s monitoring hardware is installed across a wide range of PV configurations with different panel types, inverter models, and telemetry transmission behaviours. The data ingestion layer had to accommodate this variability without requiring custom integration code per installation.
Our Approach
We built SolarWatch as a cloud-based SaaS and mobile platform, with an AI simulation engine at its core. The simulation engine models expected output for each individual module under current environmental conditions and compares it against actual measured output — identifying deviations that indicate a fault category rather than normal variation.
The key architectural decision was detecting error patterns rather than individual anomalies. Flagging individual module deviations produces alert fatigue; identifying that a cluster of affected modules shares a fault signature gives operators a diagnosis. This distinction drove the design of the AI layer.
Data ingestion runs on Apache Kafka to handle continuous module-level telemetry at production volume, with a dynamic message schema layer that accommodates variability across different hardware configurations without custom per-installation code. Multi-tenancy is enforced at the data schema level before data reaches the simulation engine — tenant isolation is structural, not a permission gate.
Dispatcher 2.0 compliance is a first-class output of the platform’s grid communication layer, not a separate operational mode.
The Outcome
- Real-time module-level visibility across multiple PV installations from a single web portal or mobile app
- AI-driven error pattern detection significantly reducing costly on-site troubleshooting
- Full Dispatcher 2.0 compliance for the German energy market
- Multi-tenant SaaS architecture supporting commercial and residential installations at scale
“We are delighted to acknowledge that Insoftex skillfully programmed our frontend using React, meticulously bringing our design to life. Their adherence to our timelines and effective communication ensured a seamless and productive collaboration.”
— Ingmar Kruse, CEO, Sun Sniffer, Germany
Team
Engagement: 6 months, 4 engineers (1 AI/ML, 1 backend/data, 1 frontend, 1 DevOps).
Stack: TypeScript, React, Node.js, MongoDB, Apache Kafka, Docker, Kubernetes