Adaptive training and deployment of single device and clustered device fault detection models for connected equipment
A fault prediction system for building equipment includes one or more memory devices configured to store instructions that, when executed on one or more processors, cause the one or more processors to receive device data for a plurality of devices of the building equipment, the device data indicating performance of the plurality of devices; generate, based on the received device data, a plurality of prediction models comprising at least one of single device prediction models generated for each of the plurality of devices or cluster prediction models generated for device clusters of the plurality of devices; label each of the plurality of prediction models as an accurately predicting model or an inaccurately predicting model based on a performance of each of the plurality of prediction models; and predict a device fault with each of the plurality of prediction models labeled as an accurately predicting model.
1. A fault prediction system for building equipment, the system comprising:
one or more memory devices configured to store instructions that, when executed on one or more processors, cause the one or more processors to:
receive device data for a plurality of devices of the building equipment, the device data indicating performance of the plurality of devices;
generate, based on the received device data, a plurality of prediction models comprising at least one of single device prediction models generated for each of the plurality of devices or cluster prediction models generated for device clusters of the plurality of devices;
label each of the plurality of prediction models as an accurately predicting model or an inaccurately predicting model based on a performance of each of the plurality of prediction models; and
predict a device fault with each of the plurality of prediction models labeled as an accurately predicting model.
2. The system of claim 1 , wherein the plurality of devices of the building equipment comprise at least one of chillers or air handling units.
3. The system of claim 1 , wherein the instructions cause the one or more processors to:
receive an indication of a new device, wherein the new device is not associated with historical device data;
assign the new device to a cluster prediction model of the plurality of cluster prediction models labeled as an accurately predicting model;
receive new device data for the new device, the new device data indicating operation of the new device; and
predict a device fault for the new device based on the new device data and the cluster prediction model to which the new device is assigned.
4. The system of claim 1 , wherein the instructions cause the one or more processors to:
re-evaluate a prediction model of the plurality of prediction models to determine whether the prediction model is to retain a previously assigned label designating the prediction model as an accurately predicting model or an inaccurately predicting model;
if the prediction model is determined to operate consistent with the previously assigned label, retain the previously assigned label designating the prediction model as an accurately predicting model or an inaccurately predicting model; and
if the prediction model is determined to operate inconsistent with the previously assigned label, switch the previously assigned label from designating the prediction model as an accurately predicting model to a new label designating the prediction model as an inaccurately predicting model or switch the previously assigned label from designating the prediction model as an inaccurately predicting model to a new label designating the prediction model as an accurately predicting model.
5. The system of claim 1 , wherein the instructions cause the one or more processors to:
train the plurality of single device prediction models, resulting in one or more model parameters; and
train the plurality of cluster prediction models with the one or more model parameters.
6. The system of claim 1 , wherein the instructions cause the one or more processors to form the plurality of device clusters based on cluster criteria comprising at least one of:
new devices without historic device data;
existing devices without proper historic device data; or
devices for which existing single device prediction models are insufficient.
7. The system of claim 1 , wherein the instructions cause the one or more processors to receive device data for a plurality of epochs, wherein the plurality of single device prediction models and the plurality of cluster prediction models are generated for each of the plurality of epochs.
8. The system of claim 1 , wherein the instructions cause the one or more processors to assign existing devices that are not associated with historic device data to one of the device clusters.
9. The system of claim 1 , wherein the instructions cause the one or more processors to generate a fault prediction indicating that a device of the plurality of devices of building equipment will shut down at a future time.
10. A method for predicting faults for building equipment, the method comprising:
receiving device data for a plurality of devices of the building equipment, the device data indicating performance of the plurality of devices;
generating, based on the received device data, a plurality of prediction models comprising at least one of single device prediction models generated for each of the plurality of devices or cluster prediction models generated for device clusters of the plurality of devices;
labeling each of the plurality of prediction models as an accurately predicting model or an inaccurately predicting model based on a performance of each of the plurality of prediction models; and
predicting a device fault with each of the plurality of prediction models labeled as an accurately predicting model.
11. The method of claim 10 , wherein the plurality of devices of the building equipment comprise at least one of chillers or air handling units.
12. The method of claim 10 , comprising:
receiving an indication of a new device, wherein the new device is not associated with historical device data;
assigning the new device to a cluster prediction model of the plurality of cluster prediction models labeled as an accurately predicting model;
receiving new device data for the new device, the new device data indicating operation of the new device; and
predicting a device fault for the new device based on the new device data and the cluster prediction model to which the new device is assigned.
13. The method of claim 10 , comprising:
re-evaluating a prediction model of the plurality of prediction models to determine whether the prediction model is to retain a previously assigned label designating the prediction model as an accurately predicting model or an inaccurately predicting model;
if the prediction model is determined to operate consistent with the previously assigned label, retaining the previously assigned label designating the prediction model as an accurately predicting model or an inaccurately predicting model; and
if the prediction model is determined to operate inconsistent with the previously assigned label, switching the previously assigned label from designating the prediction model as an accurately predicting model to a new label designating the prediction model as an inaccurately predicting model or switching the previously assigned label from designating the prediction model as an inaccurately predicting model to a new label designating the prediction model as an accurately predicting model.
14. The method of claim 10 , comprising:
training the plurality of single device prediction models, resulting in one or more model parameters; and
training the plurality of cluster prediction models with the one or more model parameters.
15. The method of claim 10 , comprising forming the plurality of device clusters based on cluster criteria comprising at least one of:
new devices without historic device data;
existing devices without proper historic device data; or
devices for which existing single device prediction models are insufficient.
16. The method of claim 10 , comprising receiving device data for a plurality of epochs, wherein the plurality of single device prediction models and the plurality of cluster prediction models are generated for each of the plurality of epochs.
17. The method of claim 10 , comprising assigning existing devices that are not associated with historic device data to one of the device clusters.
18. The method of claim 10 , comprising generating a fault prediction indicating that a device of the plurality of devices of building equipment will shut down at a future time.
19. One or more non-transitory computer-readable storage media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving device data for a plurality of devices of building equipment, the device data indicating performance of the plurality of devices;
generating, based on the received device data, a plurality of prediction models comprising at least one of single device prediction models generated for each of the plurality of devices or cluster prediction models generated for device clusters of the plurality of devices;
labeling each of the plurality of prediction models as an accurately predicting model or an inaccurately predicting model based on a performance of each of the plurality of prediction models; and
predicting a device fault with each of the plurality of prediction models labeled as an accurately predicting model.
20. The non-transitory computer-readable storage media of claim 19 , the operations comprising:
training the plurality of single device prediction models, resulting in one or more model parameters; and
training the plurality of cluster prediction models with the one or more model parameters.