Adaptively learning surrogate model for predicting building system dynamics from system identification model

Patent metadata

Field	Value
Publication number	US-2021191348-A1
Application number	US-201916726038-A
Country	US
Kind code	A1
Filing date	Dec 23, 2019
Priority date	Dec 23, 2019
Publication date	Jun 24, 2021
Grant date	—

Abstract

Systems and methods for training a surrogate model for predicting system states for a building management system based on generated data from a system identification model are disclosed herein. The system identification model is used to generate predicted system parameters of a zone of the building based on historic data from operation of the building equipment. The surrogate model is trained based on the predicted system parameters from the system identification model. Predicted future parameters of the variable state of the building are generated using the surrogate model. The surrogate model is re-trained based on new operational data from the building equipment. An updated series of predicted future parameters is generated using the re-trained surrogate model.

First claim

What is claimed is:

1 . A building management system comprising:
building equipment operable to control a variable state of a building; a processing circuit configured to:
use a system identification model to generate predicted system parameters of a zone of the building based on historic data from operation of the building equipment;
train a surrogate model based on the predicted system parameters from the system identification model;
generate predicted future parameters of the variable state of the building using the surrogate model;
re-train the surrogate model based on new operational data from the building equipment; and
generate an updated series of predicted future parameters using the re-trained surrogate model.

2 . The building management system of claim 1 , wherein a first series of predicted future parameters is less accurate than the updated series of predicted future parameters, and wherein the first series of predicted future parameters is available more quickly than the updated series of predicted future parameters generated using the re-trained surrogate model.

3 . The building management system of claim 1 , wherein the processing circuit is further configured to control the building equipment based on the predicted future parameters to generate the new operational data from the building equipment.

4 . The building management system of claim 1 , wherein the processing circuit is further configured to receive a threshold amount of new operational data from operation of the building equipment and to begin re-training of the surrogate model in response to receiving the threshold amount of new operational data.

5 . The building management system of claim 1 , wherein the processing circuit is further configured to continue to retrain the surrogate model and generate updated series of predicted future parameters as threshold amounts of new operational data is received from operation of the building equipment.

6 . The building management system of claim 1 , wherein the surrogate model is a long-short term memory model.

7 . The building management system of claim 1 , wherein the system identification model is a linear model.

8 . The building management system of claim 1 , wherein the surrogate model is used to model non-linear effects in an HVAC system of the building.

9 . The building management system of claim 1 , wherein the system parameters of the zone comprise one or more of a temperature of the zone, an air handling unit (AHU) electrical fan power, an AHU heating coil load, or an AHU cooling coil load.

10 . The building management system of claim 1 , wherein the processing circuit is configured to predict the system parameters by providing as input to the system identification model one or more of an ambient outdoor temperature, a temperature setpoint of the zone of the building, time data, a binary occupancy flag, or one or more auxiliary input variables.

11 . A method for training a surrogate model for predicting parameters for a building management system based on simulated data from a system identification model, the method comprising:
using the system identification model to generate predicted system parameters based on historic data from operation of building equipment; training the surrogate model based on the predicted system parameters from the system identification model; generating predicted future parameters of variable states for a building of interest using the surrogate model; re-training the surrogate model based on new operational data from the building equipment; and generating an updated series of predicted future parameters using the re-trained surrogate model.

12 . The method of claim 11 , wherein a first series of predicted future parameters is less accurate than the updated series of predicted future parameters, and wherein the first series of predicted future parameters is available more quickly than the updated series of predicted future parameters generated using the re-trained surrogate model.

13 . The method of claim 11 , the method further comprising controlling the building equipment based on the updated series of predicted future parameters.

14 . The method of claim 11 , the method further comprising receiving a threshold amount of new operational data from operation of the building equipment and continuing to re-train the surrogate model in response to receiving threshold amounts of new operational data.

15 . The method of claim 11 , wherein the surrogate model is a long-short term memory model.

16 . The method of claim 11 , wherein the system identification model is a linear model.

17 . The method of claim 11 , wherein the surrogate model is used to model non-linear effects in an HVAC system of the building.

18 . One or more computer-readable storage media having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
using a system identification model to generate predicted system parameters based on historic data; training a surrogate model based on the predicted system parameters from the system identification model; generating predicted future parameters using the surrogate model; re-training the surrogate model based on new data; and generating an updated series of predicted future parameters using the re-trained surrogate model.

19 . The computer-readable storage media of claim 18 , wherein a first series of predicted future parameters is less accurate than the updated series of predicted future parameters, and wherein the first series of predicted future parameters is available more quickly than the updated series of predicted future parameters generated using the re-trained surrogate model.

20 . The computer-readable storage media of claim 18 , wherein the operations further comprise controlling equipment based on the updated series of predicted future parameters.

21 . The computer-readable storage media of claim 18 , wherein the operations further comprise receiving a threshold amount of new operational data from operation of equipment and continuing to re-train the surrogate model in response to receiving threshold amounts of new operational data.

22 . The computer-readable storage media of claim 18 , wherein the surrogate model is a long-short term memory model.

23 . The computer-readable storage media of claim 18 , wherein the system identification model is a linear model.

Assignees

• Johnson Controls Tech Co

Inventors

• Lee Young M
• Jiang Zhanhong
• Drees Kirk
• Risbeck Michael

Classifications

• F24F11/64

  using pre-stored data · CPC title

• G05B2219/25011

  Domotique, I-O bus, home automation, building automation · CPC title

• G05B13/048 Primary

  using a predictor · CPC title

• G05B15/02 Primary

  electric · CPC title

• G05B13/047

  the criterion being a time optimal performance criterion · CPC title