7.7.22. PMSM Rotor Forecasting

Forecast PMSM rotor winding temperature.

7.7.22.1. Overview

This example demonstrates temperature forecasting for PMSM motor rotor windings. Predicting temperature rise helps prevent overheating and enables proactive thermal management in motor drive applications.

Application: Motor thermal management, electric vehicles, industrial drives

Task Type: Time Series Forecasting

Data Type: Multivariate (motor operating parameters)

7.7.22.2. Configuration

common:
  target_module: 'timeseries'
  task_type: 'generic_timeseries_forecasting'
  target_device: 'F28P55'

dataset:
  dataset_name: 'forecasting_pmsm_rotor'

training:
  model_name: 'FCST_4k_NPU'
  training_epochs: 50
  batch_size: 32

testing: {}
compilation: {}

7.7.22.3. Running the Example

cd tinyml-modelzoo
./run_tinyml_modelzoo.sh examples/forecasting_pmsm_rotor/config.yaml

cd tinyml-modelzoo
run_tinyml_modelzoo.bat examples\forecasting_pmsm_rotor\config.yaml

7.7.22.4. Dataset Details

Input Variables:

Motor current
Motor speed
Ambient temperature
Coolant temperature
Historical rotor temperature

Forecast Target:

Future rotor winding temperature

7.7.22.5. Recommended Models

Model	Parameters	Use Case
`FCST_2k_NPU`	~2,000	Basic forecasting
`FCST_4k_NPU`	~4,000	Balanced accuracy
`FCST_LSTM8`	Varies	LSTM-based (non-NPU)

7.7.22.6. Results

Prediction Plot:

PMSM Temperature Predictions — Forecast vs actual rotor temperature showing prediction accuracy

On-Device Results:

7.7.22.7. See Also

HVAC Indoor Temp Forecast - HVAC temperature forecasting
Forecasting Example - Forecasting tutorial
Example on GitHub