Overview of simulation

Once the blueprint is correctly configured and the model is trained, the blueprint must be simulated using historical data to evaluate the model and blueprint output prior to its implementation.

Testing the blueprint

Once the model has been trained, the blueprint should be tested.
The text source should be reconfigured to point to another data file (other than the one used to train the model) so that the model may be evaluated using an independent data set.
To evaluate the model output:
- you can port the model output into a text sink for evaluation,
- alternatively, you can configure probes on the model output port (and possibly configure another probe on a port that contains the target data).

Check through the blueprint to ensure that it is configured correctly
A runnability check should be performed (click on Blueprint>Check Runnability or press F7) to ensure that
Click on Blueprint>Run>Simulated (or F5) to run the simulation in simulation mode.
Click on Blueprint>Run>Playback (or CTRL+F5) to run the simulation in playback mode.
Once the blueprint starts executing, the Analysis view tab will become available if analysis is enabled on a model block and the Trendviewer tab will become available if probes are configured in the blueprint.
The blueprint output should then be evaluated.

Enabling analysis

To enable analysis on a model block; right-click on the block and toggle Analysis on.
To edit the analysis options, right click on the block again and click on Analysis Properties.
Specify the name for the model output that should appear on the y-axis in the analysis view window when the blueprint is run in the Display Title :box
The analysis probe smooths the input data fields to reduce the effects of noise on the analysis calculations. Specify the moving average window size in the Smoothing Filter Length [mins]: box. This information is also specified in the model alarming block will be updated if changed.
The set point value may be specified in the Setpoint Value: window, this information is also specified in the model alarming block will be updated if changed.
The history buffer size represents the number of analysis results records that will be saved to memory by the blueprint. Thus if the input data is 1 minute apart and the history buffer size is 1440, the view will retain at most 1440 minutes (i.e. 24 hours) of analysis results. If the simulation data set is 3 days long, the user will only be able to scroll through the analysis results of the last day once the simulation is complete. Thus in order to have the full 3 days worth of data to view once the simulation is complete, the history buffer size should be increased to 4320.

Return to top