Building control system with heat load estimation using deterministic and stochastic models
An environmental control system for a building including building equipment operable to affect a variable state or condition of the building. The system includes a controller including a processing circuit. The processing circuit can obtain training data relating to operation of the building equipment and can perform a system identification process to identify parameters of a system model using the training data. The processing circuit can augment the system model with a disturbance model and estimate values of a historical heat disturbance in the training data based on the augmented system model. The processing circuit can train one or more heat disturbance models based on the training data and the estimated values. The processing circuit can predict a heat disturbance using the augmented system model along with the one or more heat disturbance models and can control the building equipment based on the predicted heat disturbance.
1. An environmental control system for a building, the system comprising:
building equipment operable to affect a variable state or condition of the building; and
a controller comprising a processing circuit configured to:
identify values for unknown parameters of a non-integrating disturbance model;
estimate historical values of a heat disturbance using a system model augmented with the non-integrating disturbance model;
predict a future value of the heat disturbance using one or more heat disturbance models, wherein the one or more heat disturbance models are trained on estimated historical values of the heat disturbance; and
control the building equipment based on the predicted future value of the heat disturbance.
2. The system of claim 1 , wherein the processing circuit is further configured to train the one or more heat disturbance models by:
training a deterministic heat disturbance model using the estimated historical values of the heat disturbance; and
training a stochastic heat disturbance model using the estimated historical values of the heat disturbance and the deterministic heat disturbance model;
wherein the one or more heat disturbance models comprise the deterministic heat disturbance model and the stochastic heat disturbance model.
3. The system of claim 1 , wherein the non-integrating disturbance model is an oscillator function with a frequency of one day.
4. The system of claim 1 , wherein predicting the future value of the heat disturbance comprises:
estimating a deterministic heat disturbance for a future time step;
predicting a stochastic heat disturbance for the future time step; and
predicting the future value of the heat disturbance for the future time step based on the stochastic heat disturbance for the future time step and the deterministic heat disturbance for the future time step.
5. The system of claim 4 , wherein predicting the stochastic heat disturbance for the future time step comprises:
performing an online state estimation;
calculating a current residual based on the online state estimation and an estimated deterministic heat disturbance for a current time step; and
predicting the stochastic heat disturbance for the future time step based on a trained autoregressive model and the current residual.
6. The system of claim 1 , wherein predicting the future value of the heat disturbance comprises:
applying one or more Kalman gain parameters to a second system model;
performing a system identification process to determine values of the second system model and the one or more Kalman gain parameters; and
estimating a current stochastic heat disturbance based on the second system model with the one or more Kalman gain parameters and one or more current environmental conditions.
7. A method for operating building equipment in a building, the method comprising:
identifying values for unknown parameters of a non-integrating disturbance model:
estimating historical values of a heat disturbance using a system model augmented with the non-integrating disturbance model;
training one or more heat disturbance models on the estimated historical values of the heat disturbance affecting the building;
predicting a future value of the heat disturbance using the one or more heat disturbance models; and
controlling the building equipment using the predicted future value of the heat disturbance.
8. The method of claim 7 , wherein the non-integrating disturbance model is a second order model.
9. The method of claim 7 , wherein the non-integrating disturbance model is an oscillator function.
10. The method of claim 7 , wherein predicting the future value of the heat disturbance comprises:
estimating a deterministic heat disturbance for a future time step;
predicting a stochastic heat disturbance for the future time step; and
predicting the future value of the heat disturbance for the future time step based on the stochastic heat disturbance for the future time step and the deterministic heat disturbance for the future time step.
11. The method of claim 10 , wherein predicting the stochastic heat disturbance for the future time step comprises:
performing an online state estimation;
calculating a current residual based on the online state estimation and an estimated deterministic heat disturbance for a current time step; and
predicting the stochastic heat disturbance for the future time step based on a trained autoregressive model and the current residual.
12. The method of claim 7 , wherein predicting the future value of the heat disturbance comprises estimating a current stochastic heat disturbance based on a second augmented system model and one or more current environmental conditions.
13. One or more non-transitory computer-readable media containing program instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
obtaining estimated historical values of a heat disturbance using a second-order oscillating disturbance model;
training one or more heat disturbance models based on the estimated historical values of the heat disturbance affecting a building;
predicting a future value of the heat disturbance using the one or more heat disturbance models; and
controlling building equipment based on the predicted future value of the heat disturbance.
14. The non-transitory computer-readable media of claim 13 , wherein the second-order oscillating disturbance model has known parameters.
15. The non-transitory computer-readable media of claim 13 , wherein the second-order oscillating disturbance model has a frequency of one day.
16. The non-transitory computer-readable media of claim 13 , wherein predicting the future value of the heat disturbance comprises:
estimating a deterministic heat disturbance for a future time step;
predicting a stochastic heat disturbance for the future time step; and
predicting a forecasted heat disturbance for one or more subsequent time steps based on the stochastic heat disturbance for the future time step and the deterministic heat disturbance for the future time step.
17. The non-transitory computer-readable media of claim 13 , wherein predicting the future value of the heat disturbance comprises:
applying one or more Kalman gain parameters to an augmented system model;
performing a system identification process to determine values of the one or more Kalman gain parameters; and
estimating a current stochastic heat disturbance based on the augmented system model with the one or more Kalman gain parameters and one or more current environmental conditions.
18. The system of claim 1 , wherein the processing circuit is further configured to:
obtain training data relating to operation of the building equipment at a plurality of historical time steps; and
perform a system identification process to identify parameters of the system model using the training data.
19. The method of claim 7 , wherein estimating the historical values of the heat disturbance affecting the building is based on a set of training data for a plurality of historical time steps, the set of training data consisting of data other than the historical values of the heat disturbance.
20. The method of claim 7 , wherein the estimated historical values are obtained using a first augmented system model and controlling the building equipment using the predicted future value of the heat disturbance comprises using a second augmented system model different than the first augmented system model.
21. The non-transitory computer-readable media of claim 13 , wherein obtaining the historical values of the heat disturbance affecting the building for a plurality of historical time steps is based on a set of training data for the plurality of historical time steps, the set of training data consisting of data other than the historical values of the heat disturbance.
22. The non-transitory computer-readable media of claim 13 , wherein the operations further comprise:
performing a system identification process to identify parameters of a system model for the building;
augmenting the system model with the second-order oscillating disturbance model; and
estimating the historical values of the heat disturbance using the augmented system model.