Building control system with predictive maintenance based on time series analysis

1. A controller for an energy plant, the controller comprising:

a data collector coupled to a device of the energy plant, the data collector configured to:

obtain a time series of a performance variable associated with the device;

obtain an auto-correlation function data of the device based on the time series of the performance variable associated with the device;

obtain a partial auto-correlation function data of the device based on at least one of the time series of the performance variable or the auto-correlation function data, the partial auto-correlation function data excluding a contribution of a portion of the samples considered in the auto-correlation function;

a model generator coupled to the data collector, the model generator configured to generate an electronic model of the device using both the auto-correlation function data and the partial auto-correlation function data to select parameters of the electronic model and modify the parameters of the electronic model until the parameters are statistically significant and a slope of the auto-correlation function data or the partial auto-correlation function data satisfies a threshold criterion; and

a predictor coupled to the data collector, the predictor configured to predict a first time, at which a future event of the device is predicted to occur, based on the electronic model.

2. The controller of claim 1 , wherein the future event is the performance variable associated with the device reaching a predetermined threshold.

3. The controller of claim 1 , wherein the predictor is configured to predict, at a second time, the first time, at which the future event of the device is predicted to occur, the first time after the second time.

4. The controller of claim 1 , wherein:

the auto-correlation function data indicates a correlation between the time series of the device and a lagged time series of the device, and

the partial auto-correlation function data indicates a correlation between the time series of the device and the lagged time series without a contribution from intervening samples between the time series and the lagged time series.

5. The controller of claim 1 ,

wherein the model generator is configured to generate the electronic model of the device based on the auto-correlation function data and the partial auto-correlation function data of the device,

wherein the predictor is configured to predict the first time, at which the future event of the device is predicted to occur, based on the electronic model.

6. The controller of claim 5 , further comprising a stationary data processor coupled to the data collector, the stationary data processor configured to:

determine a stationarity of the auto-correlation function data of the device based on the partial auto-correlation function data of the device; and

in response to determining that the stationarity of the auto-correlation function data of the device is less than a predetermined threshold, perform differencing on the time series of the performance variable associated with the device to modify the auto-correlation function data and the partial auto-correlation function data of the device,

wherein the model generator is configured to obtain the electronic model based on the differenced operating on the auto-correlation function data of the device.

7. The controller of claim 6 , wherein the stationary data processor is further configured to:

detect a pattern of the time series of the performance variable associated with the device based on the auto-correlation function data and the partial auto-correlation function data; and

remove the detected pattern from the time series of the performance variable associated with the device to modify the auto-correlation function data and the partial auto- correlation function data of the device.

8. The controller of claim 5 , wherein the slope is at a cut-off point of the auto-correlation function data and the model generator is configured to determine whether the slope satisfies the threshold criterion by:

determining whether the slope is less than a predetermined threshold; and

in response to determining that the auto-correlation function data has the slope less than the predetermined threshold at the cut-off point, adding a moving average corresponding to the cut-off point of the auto-correlation function data to the electronic model.

9. The controller of claim 5 , wherein the slope is at a cut-off point of the partial auto-correlation function data and the model generator is configured to determine whether the slope satisfies the threshold criterion by:

determining whether the slope is less than a predetermined threshold; and

in response to determining that the partial auto-correlation function data has the slope less than the predetermined threshold at the cut-off point, adding an auto-regressive term corresponding to the cut-off point of the partial auto-correlation function data to the electronic model.

10. The controller of claim 1 , further comprising:

a model evaluator coupled to the predictor, the model evaluator configured to compare the future event and an actual event of the device to evaluate an accuracy of the model.

11. The controller of claim 1 , further comprising:

a report generator coupled to the predictor, the report generator configured to generate a report indicating the future event of the device and the first time, at which the future event is predicted to occur.

12. The controller of claim 1 , further comprising:

a device controller coupled to the predictor and the device, the device controller configured to automatically configure the device according to the predicted future event and the first time, at which the predicted future event is predicted to occur.

13. A method for operating an energy plant, the method comprising:

obtaining a time series of performance variable associated with a device in the energy plant;

obtaining an auto-correlation function data of the device based on the time series of the performance variable associated with the device;

obtaining a partial auto-correlation function data of the device based on at least one of the time series of the performance variable or the auto-correlation function data, the partial auto-correlation function data excluding a contribution of a portion of the samples considered in the auto-correlation function;

generating an electronic model of the device using both the auto-correlation function data and the partial auto-correlation function data to select parameters of the electronic model and modify the parameters of the electronic model until the parameters are statistically significant and a slope of the auto-correlation function data or the partial auto-correlation function data satisfies a threshold criterion; and

predicting a first time, at which a future event of the device is predicted to occur, based on the electronic model.

14. The method of claim 13 , wherein the future event is the performance variable associated with the device reaching a predetermined threshold.

15. The method of claim 13 , wherein predicting the first time, at which the future event of the device is predicted to occur, comprises:

predicting at a second time, the first time, at which the future event of the device is predicted to occur, the first time after the second time.

16. The method of claim 13 , further comprising:

wherein the auto-correlation function data indicates a correlation between the time series of the device and a lagged time series of the device, and

wherein the partial auto-correlation function data indicates a correlation between the time series of the device and the lagged time series without a contribution from intervening samples between the time series and the lagged time series.

17. The method of claim 16 , further comprising:

generating the electronic model of the device based on the auto-correlation function data and the partial auto-correlation function data of the device,

wherein the first time, at which the future event of the device is predicted to occur, is predicted based on the electronic model.

18. The method of claim 17 , further comprising:

determining a stationarity of the auto-correlation function data of the device based on the partial auto-correlation function data of the device; and

performing differencing on the time series of the performance variable associated with the device to modify the auto-correlation function data and the partial auto-correlation function data of the device, in response to determining that the stationarity of the auto-correlation function data of the device is less than a predetermined threshold,

wherein the electronic model is obtained based on the differenced operating on the auto-correlation function data of the device.