Self-learning fault detection for HVAC systems
Systems and methods for self-learning fault detection and diagnosis in an HVAC system include a server identifying a fault and one or more predicted causes of the fault based on measurements of operational parameters received from sensors associated with the HVAC system. The operational parameters are compared to evaluation criteria, such as predetermined thresholds, to identify a potential fault. Parameters may be weighted, and optionally scaled to a standardized range to facilitate the diagnosis of HVAC systems of disparate configurations and capacities. Evaluation criteria for each fault are periodically analyzed in view of operational parameter history to identify new criteria having a lower probability of misdiagnosis. Fault detection criteria which are determined to have an unacceptable error rate may be deactivated or flagged for review.
1. A heating, ventilation and air conditioning (HVAC) system fault detection method for optimizing fault detection and diagnosis, comprising:
providing an electronic data gathering device configured to acquire data from one or more components associated with the heating, ventilation and air conditioning system;
receiving, at a server, a plurality of signals indicative of sensed HVAC system operating parameters based on data acquired by the electronic data gathering device;
identifying, at the server coupled to a database, a sensed operating parameter of the plurality of the signals indicative of sensed HVAC system operating parameters that exceeds a parameter threshold to determine a set of error parameters;
normalizing, at the server, the sensed operating parameter according to a defined scale;
determining, at the server, from the set of error parameters, a potential fault and a corresponding fault threshold;
multiplying, at the server, each error parameter by a predetermined weighting factor to generate a set of weighted error parameters;
summing, at the server, the set of weighted error parameters to generate a summed value;
confirming, at the server, that the potential fault is a detected fault in response to a determination that the summed value exceeds the corresponding fault threshold;
storing, in the database coupled to the server, a dataset including a set of optimization parameters comprising the parameter threshold, the predetermined weighting factor, and the corresponding fault threshold;
applying an adjustment to the server to improve accuracy of the fault detection and diagnosis; and
performing the applying step periodically based on a predetermined time interval or on a predetermined number of instances of receiving information related to the potential fault.
2. The HVAC fault detection method in accordance with claim 1 , wherein the set of optimization parameters further comprises parameters selected from the group consisting of sensed HVAC system operating parameters, the set of error parameters, and the detected fault.
3. The HVAC fault detection method in accordance with claim 1 , further comprising transmitting a fault message indicative of the identified fault to a user device.
4. The HVAC fault detection method in accordance with claim 1 , further comprising providing an initial set of parameter thresholds.
5. The HVAC fault detection method in accordance with claim 1 , further comprising performing noise reduction on at least one of the received signals.
6. The HVAC fault detection method in accordance with claim 1 , further comprising:
receiving, at the server, feedback data indicative of whether the detected fault is an actual fault; and
storing, in the dataset, the feedback data.
7. The HVAC fault detection method in accordance with claim 1 , further comprising:
selecting, from the database, a plurality of datasets having a common detected fault;
identifying, within the selected plurality of datasets, each unique set of optimization parameters;
obtaining, for each of the selected plurality of datasets, a weighted total sum of the optimization parameters;
determining a z-score for each set of weighted total sums of the optimization parameters;
identifying the set of optimization parameters having the most negative z-score; and
utilizing the set of optimization parameters having the most negative z-score as predetermined weighting factors.
8. The HVAC fault detection method in accordance with claim 7 , further comprising comparing the z-score of the set of optimization parameters having the most negative z-score to a threshold.
9. The HVAC fault detection method in accordance with claim 8 , further comprising transmitting an alert in response to the comparing.
10. The HVAC fault detection method in accordance with claim 8 , further comprising inhibiting detection of the common detected fault in response to the comparing.
11. The HVAC fault detection method in accordance with claim 7 , further comprising:
determining a mean of each set of weighted total sums of the optimization parameters; and
determining a standard deviation of each set of weighted total sums of the optimization parameters.
12. The HVAC fault detection method in accordance with claim 11 , wherein the z-score is computed in accordance with the formula z-score=(fault threshold—mean of each set of weighted total sums)/the standard deviation mean of each set of weighted total sums.
13. A heating, ventilation and air conditioning (HVAC) fault detection system for optimizing fault detection and diagnosis, comprising:
an electronic data gathering device configured to acquire data from one or more components associated with the HVAC system;
a server configured for receiving and analyzing a plurality of signals indicative of sensed HVAC system operating parameters from one or more sensors of an HVAC system and for transmitting the received plurality of signals indicative of sensed HVAC system operating parameters to a user device, said user device is operably connected to the server, wherein the server comprises:
a database configured for storing the received plurality of signals indicative of sensed HVAC system operating parameters;
a processor operatively coupled to the database;
a memory operatively coupled to the processor and including a set of executable instructions which, when executed by the processor, cause the processor to:
identify a sensed operating parameter of the plurality of the signals indicative of sensed HVAC system operating parameters that exceeds a parameter threshold to determine a set of error parameters;
normalize the sensed operating parameter according to a defined scale;
determine, from the set of error parameters, a potential fault and a corresponding fault threshold;
multiply each error parameter by a predetermined weighting factor to generate a set of weighted error parameters;
sum the set of weighted error parameters to generate a summed value;
confirm that the potential fault is a detected fault in response to a determination that the summed value exceeds the corresponding fault threshold;
store, in the database, a dataset including a set of optimization parameters comprising the parameter threshold, the predetermined weighting factor, and the corresponding fault threshold;
apply an adjustment to the server to improve accuracy of the fault detection and diagnosis; and
perform the applying periodically based on a predetermined time interval or on a predetermined number of instances of receiving information related to the potential fault.
14. The HVAC fault detection system in accordance with claim 13 , wherein the memory includes executable instructions that further cause the processor to transmit a fault message indicative of the identified fault.
15. The HVAC fault detection system in accordance with claim 13 , wherein the memory includes executable instructions that further cause the processor to:
identify, within the selected plurality of datasets, each unique set of optimization parameters;
obtain, for each of the selected plurality of datasets, a weighted total sum of the optimization parameters;
determine a z-score for each set of weighted total sums of the optimization parameters;
identify the set of optimization parameters having the most negative z-score; and
utilize the identified set of optimization parameters having the most negative z-score as predetermined weighting factors.
16. The HVAC fault detection system in accordance with claim 15 , wherein the memory includes executable instructions that further cause the processor to compare the z-score of the set of optimization parameters having the most negative z-score to a threshold.
17. The HVAC fault detection system in accordance with claim 15 , wherein the memory includes executable instructions that further cause the processor to transmit an alert in response to the comparing.
18. The HVAC fault detection system in accordance with claim 15 , wherein the memory includes executable instructions that further cause the processor to compute the z-score in accordance with the formula z-score=(fault threshold—mean of each set of weighted total sums)/the standard deviation mean of each set of weighted total sums.