Transfer learning of deep neural network for HVAC heat transfer dynamics

1. A method, comprising:

collecting a first dataset of input-output data for a first building;

training a deep learning model using the first dataset;

initializing parameters of a target model for a second building using parameters of the deep learning model;

collecting a second dataset of input-output data for a second building;

training the target model for the second building using the initialized parameters of the target model and the second dataset; and

controlling building equipment using the target model, wherein controlling the building equipment affects a variable state or condition of the building.

2. The method of claim 1 , wherein the target model is configured to predict future values of variable states or conditions of the second building.

3. The method of claim 1 , wherein the deep learning model is a Long Short Term Memory Network based Sequence to Sequence model.

4. The method of claim 1 , wherein the first dataset is larger than the second dataset.

5. The method of claim 1 , wherein training a deep learning model using the first dataset comprises performing a supervised learning step on a specific task related to a thermal dynamic feature of the first building.

6. The method of claim 5 , wherein controlling the building equipment using the target model comprises controlling the building equipment to perform a task similar to the specific task related to the dynamic feature of the first building.

7. The method of claim 1 , wherein the initializing parameters of the target model for the second building using parameters of the deep learning model comprises performing a task adaptation process.

8. The method of claim 1 , further comprising periodically retraining the target model for the second building as more data is collected over time for the second building.

9. The method of claim 8 , wherein periodically retraining the target model comprises adjusting the target model to adapt to changes in thermal dynamic feature variables affecting the second building.

10. A system, comprising:

a first building management system configured to provide a first dataset relating to a first building;

a second building management system configured to provide a second dataset relating to a second building;

a training system configured to:

train a deep learning model using the first dataset;

generate a target model by adapting the deep learning model to generate predictions relating to the second dataset by re-training the deep learning model using the second dataset; and

provide the target model to the second building management system, the second building management system configured to control building equipment to affect an environmental attribute of the second building using the target model.

11. The system of claim 10 , wherein the training system comprises one or more cloud-based computing resources.

12. The system of claim 10 , wherein the environmental attribute comprises indoor air temperature.

13. The system of claim 10 , wherein the first building is older than the second building.

14. The system of claim 10 , wherein the first dataset is larger than the second dataset.

15. The system of claim 10 , wherein the training system is configured to apply a task adaptation procedure.

16. The system of claim 10 , wherein the deep learning model comprises a Long Short Term Memory Network based Sequence to Sequence model.

17. The system of claim 10 , wherein the target model is configured to generate predictions relating to an indoor air temperature of the second building.

18. The system of claim 10 , wherein the target model is configured to generate predictions relating to a level of energy consumption of the second building.

19. The system of claim 10 , wherein the deep learning model is a thermal model.

20. The system of claim 10 , wherein the training system is configured to execute pre-training of the deep learning model using a supervised learning process on a specific task relating to the first building in a domain relevant to the second building.