Granted Patent B2

US 12,332,615 · App. 17/710,597 · Granted Jun 17, 2025

Inventors: Santle Camilus Kulandai Samy (Sunnyvale, CA); Young M. Lee (Old Westbury, NY); Michael J. Risbeck (Madison, WI)

Assignee: TYCO FIRE & SECURITY GMBH

G05B13/048F24F11/63G05B13/0265

Abstract

Claims (63)

1. A method for building equipment comprising at least one of heating, ventilation, or air conditioning (HVAC) equipment, lighting equipment, security equipment, or fire safety equipment, the method comprising:

automatically selecting a prediction horizon used by a predictive model for the building equipment by performing evaluations of model performance at successively narrower ranges of possible prediction horizons and choosing the successively narrower ranges by choosing higher ranges in response to successful results of the evaluations and choosing lower ranges in response to unsuccessful results of the evaluations until the prediction horizon is determined;

using the predictive model with the prediction horizon to generate predicted behavior of the building equipment based at least on current measurements of a set of variables related to the building equipment and perform an automated control action comprising generating control signals for the building equipment based on the predicted behavior of the building equipment generated using the predictive model; and

affecting the predicted behavior of the building equipment by applying the control signals to the building equipment.

2. The method of claim 1 , wherein using the predictive model comprises performing fault predictions, the method further comprising automatically affecting an operation of the building equipment based on whether a fault of the building equipment is predicted to occur within the prediction horizon using the predictive model.

3. The method of claim 1 , further comprising choosing the successively narrowe ranges by:

dividing an initial range at a midpoint of the initial range;

choosing a first narrower range of the successively narrower ranges as a portion of the initial range greater than the midpoint in response to a successful evaluation result based on the midpoint of the initial range; and

choosing the first narrower range of the successively narrower ranges as a portion of the initial range less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the initial range.

4. The method of claim 3 , further comprising:

dividing the first narrower range at the midpoint of the first narrower range;

choosing a second narrower range of the successively narrower ranges as a portion of the first narrower range greater than the midpoint in response to a successful evaluation result based on the midpoint of the first narrower range; and

choosing the second narrower range of the successively narrower ranges as a portion of the first narrower range less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the first narrower range.

5. The method of claim 4 , further comprising:

dividing the second narrower range at the midpoint of the second narrower range;

choosing a first end point greater than the midpoint in response to a successful evaluation result based on the midpoint of the first narrower range; and

choosing a second end point less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the first narrower range.

6. The method of claim 5 , further comprising using the first end point or the second end point as the prediction horizon in response to a successful evaluation result using the first end point or the second end point as the prediction horizon.

7. The method of claim 1 , wherein the evaluations comprise comparing a plurality of performance metrics to a plurality of thresholds.

8. The method of claim 7 , wherein the plurality of performance metrics comprise two or more of a precision metric, a recall metric, a false positive rate, a true positive rate, an accuracy metric, and an area under a receiver operating characteristic curve.

9. The method of claim 1 , wherein performing the evaluations of model performance comprises:

using midpoints of the successively narrower ranges as the prediction horizons for fault prediction models;

training the fault prediction models; and

executing tests of the fault prediction models.

10. A building system, comprising:

building equipment operable to affect a condition of a building, the building equipment comprising at least one of heating, ventilation, or air conditioning (HVAC) equipment, lighting equipment, security equipment, or fire safety equipment;

a first portion of circuitry programmed to:

automatically select a prediction horizon for a predictive model by performing evaluations of model performance at successively narrower ranges of possible prediction horizons until the prediction horizon is determined based on results of the evaluations;

train the predictive model based on the prediction horizon;

modify the predictive model to create a modified predictive model suitable for edge execution; and

provide the modified predictive model to a second portion of circuitry locally integrated with the building equipment; and

the second portion of circuitry programmed to:

use the modified predictive model with the prediction horizon to predict, based at least on current measurements of a set of variables related to the building equipment, whether a fault of the building equipment will occur over the prediction horizon and to perform an automated control action comprising generating control signals for the building equipment responsive to the modified predictive model predicting the fault will occur over the prediction horizon; and

mitigate the fault by applying the control signals to the building equipment.

11. The system of claim 10 , wherein using the predictive model comprises performing fault predictions for the building equipment, and wherein the circuitry is programmed to perform the automated control action by automatically influencing operation of the building equipment based on whether a fault is predicted to occur at the prediction horizon.

12. The system of claim 10 , wherein the circuitry is further programmed to choose the successively narrower ranges by choosing higher ranges in response to successful results of the evaluations and choosing lower ranges in response to unsuccessful results of the evaluations.

13. The system of claim 10 , wherein the circuitry is further programmed to choose the successively narrower ranges by:

dividing an initial range at a midpoint of the initial range;

choosing a first narrower range of the successively narrower ranges as a portion of the initial range greater than the midpoint in response to a successful evaluation result based on the midpoint of the initial range;

choosing the first narrower range of the successively narrower ranges as a portion of the initial range less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the initial range;

dividing the first narrower range at the midpoint of the first narrower range;

choosing a second narrower range of the successively narrower ranges as a portion of the first narrower range greater than the midpoint in response to a successful evaluation result based on the midpoint of the first narrower range;

choosing the second narrower range of the successively narrower ranges as a portion of the first narrower range less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the first narrower range;

dividing the second narrower range at the midpoint of the second narrower range;

choosing a first end point greater than the midpoint in response to a successful evaluation result based on the midpoint of the first narrower range;

choosing a second end point less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the first narrower range; and

using the first end point or the second end point as the prediction horizon in response to a successful evaluation result using the first end point or the second end point as the prediction horizon.

14. The system of claim 10 , wherein the evaluations comprise comparisons of a plurality of performance metrics to a plurality of thresholds.

15. The system of claim 14 , wherein the plurality of performance metrics comprise two or more of a precision metric, a recall metric, a false positive rate, a true positive rate, and accuracy metric, and an area under a receiver operating characteristic curve.

16. One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, perform operations comprising:

automatically selecting a prediction horizon used by a predictive model for building equipment comprising at least one of heating, ventilation, or air conditioning (HVAC) equipment, lighting equipment, security equipment, or fire safety equipment, the prediction horizon selected by performing evaluations of model performance at successively narrower ranges of possible prediction horizons and choosing the successively narrower ranges by choosing higher ranges in response to successful results of the evaluations and choosing lower ranges in response to unsuccessful results of the evaluations until the prediction horizon is determined;

using the predictive model with the prediction horizon to generate predicted behavior of the building equipment based at least on current measurements of a set of variables related to the building equipment and perform an automated control action comprising generating control signals for the building equipment based on the predicted behavior of the building equipment generated using the predictive model; and

affecting the predicted behavior of the building equipment by applying the control signals to the building equipment.

17. The one or more non-transitory computer-readable media of claim 16 , wherein the evaluations comprise comparisons of a plurality of performance metrics to a plurality of thresholds.

18. The one or more non-transitory computer-readable media of claim 17 , wherein the plurality of performance metrics comprise two or more of a precision metric, a recall metric, a false positive rate, a true positive rate, and accuracy metric, and an area under a receiver operating characteristic curve.

19. The one or more non-transitory computer-readable media of claim 16 , the operations further comprising choosing the successively narrower ranges by:

dividing an initial range at a midpoint of the initial range;

choosing a first narrower range of the successively narrower ranges as a portion of the initial range greater than the midpoint in response to a successful evaluation result based on the midpoint of the initial range; and

choosing the first narrower range of the successively narrower ranges as a portion of the initial range less than the midpoint in response to an unsuccessful evaluation result based on the midpoint of the initial range.

20. The one or more non-transitory computer-readable media of claim 16 , wherein performing the evaluations of model performance comprises:

using midpoints of the successively narrower ranges as the prediction horizons for fault prediction models;

training the fault prediction models; and

executing tests of the fault prediction models.

Assignments (2)

Continuity (1)

References Cited (126)