# US 2021/0173360 A1 · Augmented deep learning using combined regression and artificial neural network modeling

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0173360 A1
- Document ID: US20210173360A1
- Status: Published application
- Kind: A1
- Application number: 17/179,832
- Filed: 2021-02-19
- Published: 2021-06-10
- Earliest claimed: 2017-08-03
- Assignees: Johnson Controls Technology Company
- Applicants: Johnson Controls Technology Company
- Inventors: Kirk H. Drees
- Web page: https://patentgrep.com/patent/US20210173360A1

## CPC

- G05B 13/027: {using neural networks only}
- G05B 13/048: {using a predictor}
- G05B 15/02: electric
- G06F 18/22: Matching criteria, e.g. proximity measures
- G06F 18/23213: with fixed number of clusters, e.g. K-means clustering
- G06N 3/045: Combinations of networks
- G06N 3/0499: Feedforward networks
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06N 7/00: Computing arrangements based on specific mathematical models
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G05B 2219/25011: Domotique, I-O bus, home automation, building automation
- G06F 2218/00: Aspects of pattern recognition specially adapted for signal processing
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method for initiating and automatically improving model-driven operations in a low-data scenario includes creating a regression model using pre-operation data prior to initiating the model-driven operations, using the regression model to initiate and perform the model-driven operations during an operational stage, collecting operational data during the operational stage, creating a first artificial neural network model using the operational data, transitioning from using the regression model to perform the model-driven operations to using the first artificial neural network model to perform the model-driven operations responsive to the operational data satisfying a first sufficiency threshold.

## Claims

### Claim 1

A method for operating a building management system for a physical plant, the method comprising:
creating a regression model using pre-operation data during a pre-operational stage of physical plant;
identifying a plurality of data clusters generated by physical plant data during an operational stage of the physical plant;
determining whether the plurality of data clusters exceeds a first data sufficiency threshold;
in response to a determination that the plurality of data clusters exceeds the first data sufficiency threshold, creating a first artificial neural network model using the plurality of data clusters;
determining whether new physical plant data meets a first similarity criterion of at least one of the plurality of data clusters;
in response to a determination that the new physical plant data meets the first similarity criterion, making a first artificial neural network prediction using the first artificial neural network model; and
modifying a characteristic of the physical plant according to the first artificial neural network prediction.

### Claim 2 (depends on 1)

The method of claim 1, wherein the pre-operation data comprises at least one of manufacturing data and offsite data.

### Claim 3 (depends on 1)

The method of claim 1, wherein the physical plant data comprises at least one of plant input data or plant output data.

### Claim 4 (depends on 1)

The method of claim 1, wherein the first data sufficiency threshold is based on a quantity of physical plant data.

### Claim 5 (depends on 1)

The method of claim 1, wherein the method further comprises:
in response to a determination that the plurality of data clusters does not exceed the first data sufficiency threshold, making a regression model prediction using the regression model;
utilizing the regression model prediction to perform at least one of a fault detection task, a fault diagnosis task, or a control task.

### Claim 6 (depends on 1)

The method of claim 1, wherein the method further comprises utilizing the first artificial neural network prediction as an input to the regression model, the first artificial neural network prediction configured to improve a quality of the regression model.

### Claim 7

A method for initiating and automatically improving model-driven operations in a low-data scenario, the method comprising:
creating a regression model using pre-operation data prior to initiating the model-driven operations;
using the regression model to initiate and perform the model-driven operations during an operational stage;
collecting operational data during the operational stage;
creating a first artificial neural network model using the operational data;
transitioning from using the regression model to perform the model-driven operations to using the first artificial neural network model to perform the model-driven operations responsive to the operational data satisfying a first sufficiency threshold.

### Claim 8 (depends on 7)

The method of claim 7, further comprising creating a second artificial neural network model using the operational data; and
transitioning from using the first artificial neural network model to perform the model-driven operations to using the second artificial neural network model to perform the model-driven operations responsive to the operational data satisfying a second sufficiency threshold.

### Claim 9 (depends on 7)

The method of claim 7, wherein the first sufficiency threshold is satisfied when at least a threshold quantity of the operational data is collected.

### Claim 10 (depends on 7)

The method of claim 7, wherein transitioning from using the regression model to using the first artificial neural network model is further responsive to satisfying a criterion indicative of similarity between the operational data and new operational data.

### Claim 11 (depends on 10)

The method of claim 10, wherein the criterion indicative of similarity between the operational data and the new operational data is based on a distance between the new operational data and a cluster of the operational data.

### Claim 12 (depends on 7)

The method of claim 7, wherein transitioning from using the regression model and using the first artificial neural network model comprises calculating a combined output using both the regression model and the first artificial neural network model and using the combined output to perform the model-driven operations.

### Claim 13

A method for initiating and automatically improving model-driven operations in a low-data scenario, the method comprising:
creating a regression model using pre-operation data prior to initiating the model-driven operations;
using the regression model to initiate and perform the model-driven operations during an operational stage;
collecting operational data during the operational stage;
creating a first artificial neural network model using the operational data responsive to determining that a first amount of the operational data has been collected; and
using the first artificial neural network model to continue performing the model-driven operations during the operational stage.

### Claim 14 (depends on 13)

The method of claim 13, wherein using the first neural network model to continue performing the model-driven operations during the operational stage comprises combining outputs of the first artificial neural network model and the regression model to generate a combined output and using the combined output to perform the model-driven operations.

### Claim 15 (depends on 14)

The method of claim 14, wherein combining the outputs of the first artificial neural network model and the regression model comprises calculating a weighted average, wherein the weighted average is weighted based on a distance between a new data sample of the operational data and a cluster of previous data samples of the operational data.

### Claim 16 (depends on 13)

The method of claim 13, wherein using the first neural network model to continue performing the model-driven operations during the operational stage comprises determining whether to use the first artificial neural network model or the regression model based on a quantification of similarity between a new data sample of the operational data and one or more previous samples of the operational data.

### Claim 17 (depends on 13)

The method of claim 13, wherein performing the model-driven operations during the operational stage comprises collecting more of the operational data, the method further comprising:
creating a second artificial neural network model using the operational data responsive to determining that a second amount of the operational data has been collected; and
using the second artificial neural network model to continue performing the model-driven operations during the operational stage.

### Claim 18 (depends on 17)

The method of claim 17, wherein using the second artificial neural network model to continue performing the model-driven operations during the operational stage comprises combining outputs of the first artificial neural network model and the second artificial neural network model to generate a combined output and using the combined output to perform the model-driven operations.

### Claim 19 (depends on 13)

The method of claim 13, wherein performing the model-driven operations during the operational stage comprises controlling a system using a prediction of the regression model, the first artificial neural network model, or a combination of the regression model and the first artificial neural network model.

### Claim 20 (depends on 19)

The method of claim 19, wherein the system comprises HVAC equipment.
