Heuristic Method of Automated and Learning Control, and Building Automation Systems Thereof
Apparatuses, systems, and methods of physical-model based building automation using in-situ regression to optimize control systems are presented. A simulation engine is configured to simulate a behavior or a controlled system using a physical model for the controlled system. A data stream comprises data from a controlled system. A training loop is configured to compare an output of a simulation engine to a data stream using a heuristic so that a physical model is regressed in a manner that the output of the simulation engine approaches the data stream.
1 . A controller for adjusting a model of a controlled system, the controller comprising:
a memory; and
a processor in communication with the memory and configured to:
receive a data stream from the controlled system;
simulate a behavior of the controlled system using a physical model of the controlled system to produce a model output;
compare the model output to the data stream producing a prediction of physical model behavior, and
use the prediction of the physical model behavior to adjust the physical model.
2 . The controller of claim 1 , further comprising at least one of: finding a reduction in error between values of the data stream and values of the model outputs to within an arbitrary threshold; finding a reduction in uncertainty between values of the data stream and values of the model output; or finding a reduction in uncertainty of values of the model output to within an arbitrary threshold.
3 . The controller of claim 1 , further comprising at least one of: reaching an arbitrary threshold on number of erroneous values of the model output; reaching an arbitrary threshold on number of accurate values of the model output; reaching an arbitrary threshold on number of values in the model output; or reaching an arbitrary threshold on computational time spent.
4 . The controller of claim 1 , wherein the data stream comprises sensor measurements, equipment state, environmental data, occupant input, or occupant behavior.
5 . The controller of claim 1 , wherein adjusting the physical model comprises adjusting parameters of the physical model or modifying inputs of the physical model.
6 . The controller of claim 1 , wherein the compare the model output to the data stream comprises using a cost function.
7 . The controller of claim 6 , wherein the cost function is time variant.
8 . The controller of claim 1 , wherein the use the prediction of the physical model behavior to adjust the physical model comprises regressing the physical model using a difference of the model output and the data stream to determine at least one adjustment to the physical model.
9 . The controller of claim 8 , wherein regressing the physical model comprises regressing the physical model using differential comparison.
10 . A method executed by at least one processor for modifying a physical model of a controlled system, the method comprising:
receiving, by the at least one processor, a data stream from the controlled system;
simulating a behavior of the controlled system using a physical model of the controlled system to produce a model output,
comparing the model output to the data stream producing a prediction of physical model behavior, and
using the prediction of the physical model behavior to adjust the physical model.
11 . The method of claim 10 , further comprising at least one of: finding a reduction in error between values of the data stream and values of the model output to within an arbitrary threshold; or finding a reduction in uncertainty between values of the data stream and values of the model output to within an arbitrary threshold.
12 . The method of claim 10 , further comprising at least one of finding a reduction in uncertainty of values of the model output to within an arbitrary threshold; reaching an arbitrary threshold on number of erroneous values of the model output; reaching an arbitrary threshold on number of accurate values of the model output; or reaching an arbitrary threshold on number of values in the model output; and reaching an arbitrary threshold on computational time spent.
13 . The method of claim 10 , wherein the data stream comprises sensor measurements, equipment state, environmental data, occupant input, or occupant behavior.
14 . The method of claim 10 , further comprising using a cost function to evaluate difference between the data stream and the model output.
15 . The method of claim 14 , wherein the cost function is time variant.
16 . The method of claim 10 wherein adjusting the physical model comprises making a modification to model inputs or making a modification to model parameters.
17 . A non-transitory machine-readable medium encoded with instructions for execution by a processor for modifying a physical model of a controlled system, the non-transitory machine-readable medium comprising:
instructions for receiving, by the processor, a data stream from the controlled system;
instructions for simulating a behavior of the controlled system using a physical model of the controlled system to produce a model output,
instructions for comparing the model output to the data stream producing a prediction of physical model behavior, and
instructions for using the prediction of the physical model behavior to adjust the physical model.
18 . The non-transitory machine-readable medium of claim 17 , wherein adjusting the physical model comprises adjusting parameters of the physical model.
19 . The non-transitory machine-readable medium of claim 17 , wherein the data stream comprises sensor measurements, equipment state, environmental data, occupant input, or occupant behavior.
20 . The non-transitory machine-readable medium of claim 17 , further comprising instructions for using a cost function to evaluate difference between the data stream and the model output.