Skip to main content

Plant Onboarding Process

To provide a high level of data quality through its portal, Novasense pays the highest attention to metadata quality during the onboarding process. The onboarding of a new plant in the Novasense Portal involves multiple steps.

Plant Documentation Transfer

The first onboarding step is to provide Novasense with access to the plant documentation. The following documents are required:

  • Filled Onboarding Form Template
  • String wiring plans
  • Inverter datasheets
  • Solar module datasheets
  • Information about data sources used
warning

Ensure that you provide accurate, correct, and precise plans and metadata. Errors in the documentation might lead to additional effort in the subsequent onboarding steps.

Initial Plant Onboarding

The initial plant onboarding is performed by Novasense in three steps:

  1. Document Plausibility Check: Verify if the provided documents are clear and coherent.
  2. Time Series Check: Check normalized raw data to ensure that the electrical measurements are coherent with the provided plant layout information.
  3. Novasense Portal Setup: Set up the Novasense Portal with all metadata, obtain weather, and inverter data history.

Any obviously incorrect plant data provided to Novasense will lead to an interruption of the process for further clarification with the customer. In case of uncertainty, Novasense might request a final validation of the plant layout from the customer.

note

At this stage, the plant will be visible on the Novasense Portal, and the anomaly detection will have standard thresholds defined for perfectly functioning plants. The fine-tuning of the anomaly detection takes place in the next step.

Anomaly Detection Tuning

After 6-12 months of operation, the historical data will be evaluated, and the anomaly detection will be tuned to provide an optimal balance between sensitivity and selectivity. If needed, machine learning models will be trained to improve the quality of the target values calculations. More information about power modeling is provided on the KPI Calculation Page.

Tuning Status Overview

The anomaly detection tuning process follows different statuses throughout a plant's lifecycle:

StatusDescription
Not StartedNew plant - tuning has not yet begun. Standard thresholds are applied.
Waiting for DataPlant is collecting the required 6-12 months of normally performing data, covering both winter and summer months.
In ProgressTuning has started and is currently being performed by Novasense staff. This process might require multiple days to complete. This status also indicates that clarifications or additional information from the customer may be required.
Completed (Physical Model)Tuning finished using only the physical model. No machine learning was necessary for this plant.
Completed (Machine Learning Model)Tuning finished with machine learning models trained to improve calculation quality.
In Progress (After Change)Plant configuration changed, requiring re-tuning of the anomaly detection (status returns to "In Progress").
ArchivedPlant is no longer actively monitored. Anomaly detection is suspended.

Tuning Lifecycle

The following diagram illustrates the anomaly detection tuning lifecycle:

Status Transitions

  • Not Started → Waiting for Data: Once the plant is onboarded and data collection begins.
  • Waiting for Data → In Progress: After 6-12 months of quality data covering seasonal variations.
  • In Progress → Completed: When threshold tuning and model training (if needed) are finished.
  • Completed → In Progress: When significant plant configuration changes occur (e.g., inverter replacement, module additions).
  • Any Status → Archived: When the plant is decommissioned or no longer actively monitored.
note

Before the tuning status reaches "Completed", it is normal to observe performance-related anomalies, as the exact behavior of the plant is not yet fully characterized. After tuning completion, false positives may occur sporadically, particularly during periods of very low irradiance, but should remain exceptional.