Building management system with augmented deep learning using combined regression and artificial neural network modeling

1. A building management system comprising:

a database configured to store at least one of plant input data, plant output data, or identified trust region data;

a trust region identifier configured to perform a cluster analysis technique to identify trust regions;

a regression model predictor configured to utilize a regression model technique to calculate a regression model prediction based at least in part on plant input data and plant output data;

a distance metric calculator configured to calculate a distance metric;

an artificial neural network model predictor configured to utilize an artificial neural network model technique to calculate an artificial neural network model prediction based at least in part on the plant input data, the plant output data, and the distance metric; and

a combined prediction calculator configured to determine a combined prediction based on the distance metric and at least one of the regression model prediction or the artificial neural network model prediction; and

a controller configured to modify a characteristic of a physical plant based on the combined prediction.

2. The building management system of claim 1 , wherein the combined prediction calculator uses at least one of a weighted average or a Kalman filter to determine the combined prediction.

3. The building management system of claim 1 , wherein the distance metric calculator is configured to calculate the distance metric using plant input data and at least one of a cluster distribution mean or a cluster centroid.

4. The building management system of claim 3 , wherein the cluster distribution mean is identified using a Gaussian mixture model technique.

5. The building management system of claim 3 , wherein the cluster centroid is identified using a k-means technique.

6. The building management system of claim 1 , wherein the plant input data comprises at least one of manufacturing data and offsite data.

7. A method of making an augmented deep learning model prediction comprising:

receiving plant input data and plant output data from a physical plant;

performing a cluster analysis technique to identify trust regions;

calculating a regression model prediction using a regression model technique based at least in part on the plant input data and the plant output data;

calculating a distance metric;

calculating an artificial neural network prediction using an artificial neural network technique based at least in part on plant input data, plant output data, and the distance metric;

determining a combined prediction based on the distance metric and at least one of the regression model prediction or the artificial neural network prediction; and

modifying a characteristic of the physical plant according to the combined prediction.

8. The method of claim 7 , wherein determining the combined prediction comprises use of at least one of a weighted average or a Kalman filter.

9. The method of claim 7 , wherein calculating the distance metric comprises use of the plant input data and at least one of a cluster distribution mean or a cluster centroid.

10. The method of claim 9 , wherein the cluster distribution mean is identified using a Gaussian mixture model technique.

11. The method of claim 9 , wherein the cluster centroid is identified using a k-means technique.

12. The method of claim 7 , wherein the plant input data comprises at least one of manufacturing data and offsite data.