Systems and methods for predicting occupancy for one building using a model trained at another building

Patent metadata

Field	Value
Publication number	US-12455091-B2
Application number	US-202217949331-A
Country	US
Kind code	B2
Filing date	Sep 21, 2022
Priority date	Sep 21, 2022
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

A Building Management System (BMS) may be controlled in accordance with predicted occupancy using a trained model. A model is trained by providing the model with time stamped environmental data and corresponding time stamped occupancy data pertaining to a training building, wherein the time stamped environmental data is derived from one or more environmental sensors of the training building and the corresponding time stamped occupancy data is derived from one or more occupancy sensors of the training building. Once trained, the trained model is provided with time stamped environmental data for a use building that is derived from one or more environmental sensors of the use building. The trained model outputs a predicted occupancy value that represents a predicted occupancy count in the use building, and the BMS of the use building is controlled based at least in part on the predicted occupancy value.

First claim

What is claimed is:

1 . A method for controlling one or more components of a Building Management System (BMS) of a use building in accordance with predicted occupancy of the use building, the predicted occupancy based upon a trained model, the method comprising:
training a model by:
providing the model with time stamped environmental data and corresponding time stamped occupancy data pertaining to a training building that is different from the use building, wherein the time stamped environmental data is derived from one or more environmental sensors of the training building and the corresponding time stamped occupancy data is derived from one or more occupancy sensors of the training building;
training the model over time using the time stamped environmental data and the corresponding time stamped occupancy data, resulting in a trained model;
employing the trained model in the use building, where employing the trained model in the use building comprises:
providing the trained model with time stamped environmental data pertaining to the use building, wherein the time stamped environmental data is derived from one or more environmental sensors of the use building;
receiving a raw occupancy count value produced by the trained model;
applying one or more normalization factors to the raw occupancy count value to produce a predicted occupancy value, wherein the one or more normalization factors account for one or more differences between the training building and the use building including one or more of:
a volumetric difference between the training building and the use building; and
a maximum occupancy count difference between the training building and the use building;

the trained model outputting the predicted occupancy value that represents a predicted occupancy count in the use building; and
operating the BMS of the use building based at least in part on the predicted occupancy value.

2 . The method of claim 1 , wherein the BMS of the use building and a BMS of the training building each have one or more Air handler Units (AHUs) each servicing a corresponding space, the method comprising:
training the model for a particular AHU of the training building with the time stamped environmental data and the corresponding time stamped occupancy data pertaining to the space of the training building serviced by the particular AHU of the training building; applying the trained model to each of two or more AHUs of the use building, thereby outputting a predicted occupancy value for each of the spaces serviced by the two or more AHUs of the use building; and aggregating the predicted occupancy value for each of the spaces serviced by the two or more AHUs of the use building to derive the predicted occupancy value for the use building.

3 . The method of claim 2 , further comprising:
applying one or more normalization factors to the predicted occupancy value for each of the spaces serviced by the two or more AHUs of the use building before aggregating the predicted occupancy value for each of the spaces serviced by the two or more AHUs of the use building to derive the predicted occupancy value for the use building, wherein the one or more normalization factors account for one or more of:
volumetric differences between the space serviced by the particular AHU of the training building and the space serviced by the corresponding AHU of the use building;
maximum occupancy count differences between the space serviced by the particular AHU of the training building and the space serviced by the corresponding AHU of the use building; and
differences in load capacity between the particular AHU of the training building and the corresponding AHU of the use building.

4 . The method of claim 1 , wherein the model comprises a regression machine learning model.

5 . The method of claim 1 , wherein the time stamped environmental data comprises two or more of:
an indoor CO 2 concentration; an indoor temperature; a fresh air intake flow rate of an AHU of the BMS; an indoor humidity; an indoor Total Volatile Organic compounds (TVOC); and an indoor Particulate Matter 2.5 (PM2.5).

6 . The method of claim 5 , wherein the time stamped environmental data comprises one or more of:
an outdoor CO 2 concentration; an outdoor temperature; an outdoor humidity; an outdoor Total Volatile Organic compounds (TVOC); and an outdoor Particulate Matter 2.5 (PM2.5).

7 . The method of claim 1 , wherein the time stamped environmental data comprises two or more of outside air temperature, outside humidity, return air temperature, return air humidity, supply air temperature, supply air humidity, supply air static pressure and supply temperature setpoint.

8 . A method for controlling one or more components of a Building Management System (BMS) of a use building in accordance with predicted occupancy of the use building, wherein the BMS of the use building and a BMS of the training building each have one or more Air handler Units (AHUs), the predicted occupancy based upon a trained model, the method comprising:
training a model by:
providing the model with time stamped environmental data and corresponding time stamped occupancy data pertaining to a training building that is different from the use building, wherein the time stamped environmental data is derived from one or more environmental sensors of the training building and the corresponding time stamped occupancy data is derived from one or more occupancy sensors of the training building;
training the model over time using the time stamped environmental data and the corresponding time stamped occupancy data, resulting in a trained model;
employing the trained model in the use building, where employing the trained model in the use building comprises:
providing the trained model with time stamped environmental data pertaining to the use building, wherein the time stamped environmental data is derived from one or more environmental sensors of the use building;
receiving a raw occupancy count value produced by the trained model;
applying one or more normalization factors to the raw occupancy count value to produce a predicted occupancy value, wherein the one or more normalization factors account for one or more differences between the training building and the use building including a difference in a load capacity between one or more of the AHUs of the training building and the use building;
the trained model outputting the predicted occupancy value that represents a predicted occupancy count in the use building; and
operating the BMS of the use building based at least in part on the predicted occupancy value.

9 . A method comprising:
storing a trained model that is trained to predict an occupancy count of a training building space in a training building using time stamped environmental data and corresponding time stamped occupancy data pertaining to the training building space in the training building, the training building space in the training building serviced by an Air Handling Unit (AHU) of the training building, and wherein the training building space in the training building has a training building space volume and the AHU of the training building has a load capacity;
predicting an occupancy count of a use building space in a use building, wherein the use building is different from the training building and the use building space in the use building is serviced by an AHU of the use building, and wherein the use building space in the use building has a use building space volume and the AHU of the use building has a load capacity, wherein predicting the occupancy count of the use building space includes:
providing to the trained model time stamped environmental data p

Assignees

• Honeywell Int Inc

Inventors

• Pal Rohil
• Kumar Navneet
• Sandeep Deepika
• Ranjan Prabhat
• Gupta Bhavesh

Classifications

• F24F2110/10

  Temperature · CPC title

• F24F2110/66

  Volatile organic compounds [VOC] · CPC title

• F24F2110/70

  Carbon dioxide · CPC title

• F24F2120/10

  Occupancy · CPC title

• F24F2110/64

  Airborne particle content · CPC title