Setup
Your park is onboarded and data is flowing — now turn it into an accurate digital twin. The good news: you do not type in every string's configuration by hand. You configure only what you know for sure, let the Deep Analysis detect the rest, and then review where the two disagree. This page is the short version of that workflow; each step links to the detailed guide.
The Core Idea
Configure what you are certain of. The analysis fills everything else. A wrong manual value is worse than an empty one — empty is honest, and the analysis fills it for you.
The Workflow
Step 0 — Onboarding must be complete
The Onboarding wizard must be finished, with two values set correctly: the plant's DC peak power (modules) and its AC peak power (inverters). Every later comparison — detected power, performance, curtailment — is anchored to these two numbers.
Step 1 — Assign datasheets
Panels and inverters need a datasheet so the analysis knows their ratings. Datasheets are extracted from the documents you uploaded during Onboarding (Step 1) — and the one you need may already exist in the platform's library, in which case you simply assign it on the Components page.
Why datasheets matter so much
A module datasheet is the electrical fingerprint of your panels. With it, the analysis can translate what it measures into real quantities: how many panels are wired in a row, how many rows run in parallel, and what each string can deliver at its peak. Without datasheets the analysis still recognises orientation and the wiring structure — but it cannot state your capacity in watts. Assigning datasheets is the single highest-value minute of this whole setup. The inverter datasheet matters too: its AC limit lets the analysis recognise clipping instead of misreading it as a plant problem.
Step 2 — Collect data for a couple of days
The analysis works from real production data, so let the park report for a few days first. More days — ideally with some clear weather — give it more to work with. Nothing to do here but wait.
Step 3 — Configure what you know for sure
For each measurement point (string), enter only the values you are certain about: azimuth, tilt, panels in a row, parallel rows.
Where this information actually lives
String orientation and layout are usually not in the solar yield report — they come from the cable pulling list (Kabelzugliste) or the installation documentation. If you do not have it, you probably do not know these values for sure — and that is fine, the analysis will detect them.
Step 4 — Leave unknowns empty and run the Deep Analysis
Leave every uncertain field empty and start the Deep Analysis from the Components page. The analysis detects the missing values string by string — and it keeps your manual values untouched: the priority is always manual before detected before default. You can follow the run live in the deep-analysis protocol.
Step 5 — Review the result
The Review Status card (top right of the Components page) shows how much peak power was detected versus configured, how many strings are auto-configured, and how many need attention. Click Review to see exactly those: conflicting values are marked orange directly in the table. Fix the configuration where the detection is right, keep your value where it is not, and verify strings once they are confirmed.
Repeat until All Good
Re-run the analysis as often as you like — each string keeps its individually best detection across runs, so a better day only improves the picture. You are done when the card reads All Good: power matching, everything verified, nothing flagged.
What You Get — and What Comes Next
This workflow is not data entry for its own sake. Its purpose is a faithful digital copy of your park — built from what the plant actually measures, corrected only where you know better. The analysis tries to understand the park as well as the available information allows: with nothing but discovered components it already maps which inputs carry strings and how they are wired; with datasheets it quantifies capacity; with your verified values it becomes trustworthy end to end.
That digital twin is the foundation for what follows: once enough of your strings are verified (about 80% of the visible ones), the platform switches on the automatic health watchdog — every night it compares each string's real behaviour against its twin, so losses, degradation, and creeping faults surface on their own, without anyone staring at charts. The better the twin, the earlier and more precisely problems are caught. See Digital Twin, Loss Detection, and Component States.
Where to Go Deeper
Each part of this workflow has its own focused guide:
- How to operate the Components page — every dialog, the bulk editing, the review controls, and how value priorities work in detail: Configuring Components
- How the analysis works — what it detects, how it handles curtailment and clouds, and where its limits are: Digital Twin
- What the results feed into — Loss Detection and Component States