Adaptive selection of machine learning/deep learning model with optimal hyper-parameters for anomaly detection of connected chillers

1. A chiller model management system for a building, 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:

determine whether chiller fault data exists in chiller data used to generate a plurality of chiller shutdown prediction models;

generate a first performance evaluation value for each of the plurality of chiller shutdown prediction models using a first evaluation technique in response to a determination that chiller fault data exists in the chiller data;

generate a second performance evaluation value for each of the plurality of chiller shutdown prediction models using a second evaluation technique in response to a determination that chiller fault data does not exist in the chiller data; and

select one of the plurality of chiller shutdown prediction models based on the first performance evaluation value and the second performance evaluation value.

2. The system of claim 1 , wherein the instructions cause the one or more processors to:

receive the chiller data for one or more chillers, the chiller data indicating performance of one or more chillers; and

generate the plurality of chiller shutdown prediction models based on the chiller data.

3. The system of claim 1 , wherein the instructions cause the one or more processors to:

in response to identifying chiller fault data in the chiller data, calculate receiver operating characteristics for the chiller data; and

determine an area under a curve of the receiver operating characteristics.

4. The system of claim 1 , wherein the instructions cause the one or more processors to:

in response to identifying that the chiller data is deficient of any chiller fault data, calculate a standard deviation normalized by mean for a measure of a probability density function for the chiller data; and

determine an inverse of the standard deviation normalized by mean for the measure of the probability density function.

5. The system of claim 1 , wherein the instructions cause the one or more processors to

receive the chiller data for a plurality of epochs, wherein the plurality of chiller shutdown prediction models is generated for each of the plurality of epochs.

6. The system of claim 1 , wherein the instructions cause the one or more processors to select a chiller prediction model with a greatest value resulting from a calculated area under a curve of receiver operating characteristics or a calculated inverse of a standard deviation normalized by mean for each epoch.

7. The system of claim 1 , wherein the instructions cause the one or more processors to:

train each of the plurality of chiller shutdown prediction models with different sets of hyper-parameters; and

select the set of hyper-parameters for each of the plurality of chiller shutdown prediction models that produces a most accurate prediction.

8. The system of claim 1 , wherein the instructions cause the one or more processors to generate a fault prediction using the selected chiller prediction model.

9. The system of claim 1 , wherein the instructions cause the one or more processors to update a fault prediction in response to the selected chiller prediction model for each epoch of a plurality of epochs.

10. A method for managing chiller models for a building, the method comprising:

determining whether chiller fault data exists in chiller data used to generate a plurality of chiller shutdown prediction models;

generating a first performance evaluation value for each of the plurality of chiller shutdown prediction models using a first evaluation technique in response to a determination that chiller fault data exists in the chiller data;

generating a second performance evaluation value for each of the plurality of chiller shutdown prediction models using a second evaluation technique in response to a determination that chiller fault data does not exist in the chiller data; and

selecting one of the plurality of chiller shutdown prediction models based on the first performance evaluation value and the second performance evaluation value.

11. The method of claim 10 , further comprising:

receiving the chiller data for one or more chillers, the chiller data indicating performance of one or more chillers; and

generating the plurality of chiller shutdown prediction models based on the chiller data.

12. The method of claim 10 , further comprising:

in response to identifying chiller fault data in the chiller data, calculating receiver operating characteristics for the chiller data; and

determining an area under a curve of the receiver operating characteristics.

13. The method of claim 10 , further comprising:

in response to identifying that the chiller data is deficient of any chiller fault data, calculating a standard deviation normalized by mean for a measure of a probability density function for the chiller data; and

determining an inverse of the standard deviation normalized by mean for the measure of the probability density function.

14. The method of claim 10 , further comprising receiving the chiller data for a plurality of epochs, wherein the plurality of chiller shutdown prediction models is generated for each of the plurality of epochs.

15. The method of claim 10 , further comprising selecting a chiller prediction model with a greatest value resulting from a calculated area under a curve of receiver operating characteristics or a calculated inverse of a standard deviation normalized by mean for each epoch.

16. The method of claim 10 , further comprising:

training each of the plurality of chiller shutdown prediction models with different sets of hyper-parameters; and

selecting the set of hyper-parameters for each of the plurality of chiller shutdown prediction models that produces a most accurate prediction.

17. The method of claim 10 , further comprising generating a fault prediction using the selected chiller prediction model.

18. The method of claim 10 , further comprising updating a fault prediction in response to the selected chiller prediction model for each epoch of a plurality of epochs.

19. A chiller model management system for a building, the system comprising:

one or more processors configured to:

determine whether chiller fault data exists in chiller data used to generate a plurality of chiller shutdown prediction models;

generate a first performance evaluation value for each of the plurality of chiller shutdown prediction models using a first evaluation technique in response to a determination that chiller fault data exists in the chiller data;

generate a second performance evaluation value for each of the plurality of chiller shutdown prediction models using a second evaluation technique in response to a determination that chiller fault data does not exist in the chiller data; and

select one of the plurality of chiller shutdown prediction models based on the first performance evaluation value and the second performance evaluation value.

20. The system of claim 19 , wherein the one or more processors are configured to:

receive the chiller data for one or more chillers, the chiller data indicating performance of one or more chillers; and

generate the plurality of chiller shutdown prediction models based on the chiller data.