¶ Install QASSANDRA
¶ Create and Run a Project
¶ Create a New Project
- Start QASSANDRA
- On the Welcome page, click New Project
- QASSANDRA opens the configuration wizard
- Complete the three tabs in order:
- Plant
- Weather
- Model
A new project starts in a temporary working location. Save it as a .qas file when the project is ready.
¶ Configure the Plant
The Plant tab tells QASSANDRA where your measured data comes from and which value you want to predict.
¶ Enter the Plant Name
- Enter a clear name that identifies the plant
¶ Select the Plant Input
Choose one of the available sources:
- Google Drive
- Local Files
- OPC UA
¶ Google Drive Plant Data
If you selected Google Drive:
- Browse to the Google service credential file
- Enter the Google Drive Folder ID
- Click Connect
- Check the preview
- Use Download history data when you need to import the available history
- Confirm the Time and target columns
¶ Local Plant Data
If you selected Local Files:
- Browse to the folder containing the plant CSV files
- Load the history
- Check the preview
- Confirm the Time and target columns
The original files stay in the selected folder.
¶ OPC UA Plant Data
If you selected OPC UA:
- Browse to the OPC UA configuration supplied for the plant
- Click Connect
- Check that a valid time and target value are returned
- Continue with Weather and Model configuration
OPC projects may need a warm up period while QASSANDRA collects enough hourly history for training.
¶ Select the Target
QASSANDRA includes Flow as the default target for this version.
For the active target:
- Select the source column
- Select the target unit
- Make sure the correct target is active
¶ Select the Time Column
- Select the source column that contains timestamps and choose the matching time format.
¶ Load the Plant History
- Historical plant data is used to train and evaluate the model.
¶ Preview the Data
Use the plant preview to check:
- timestamps
- target values
- units
- missing periods
Figure X - Plant configuration
¶ Review a model before using it
QASSANDRA provides a Preview Training step before production use. The training result can include:
- measured values
- predicted values
- a time chart
- an XY comparison chart
- the evaluation result
- the selected model parameters
- comparison results when several algorithms are tested
The user can inspect the result and accept the selected model.