HVAC system with self-optimizing control from normal operating data

1. A building management system, comprising:

building equipment configured to operate in accordance with an input to alter a variable state or condition of a building as a process function of the input while incurring a cost of operating the equipment as a cost function of the input; and

a controller configured to:

calculate a value of a self-optimizing control variable as a function of a measured state of the building equipment; and

drive the value of the self-optimizing control variable towards a setpoint value by generating the input based on the self-optimizing control variable and providing the input to the building equipment;

wherein the function comprises multiplying the measured state by a matrix and adding an offset vector; and

wherein values of elements of the matrix and the offset vector are determined using a non-optimal reference.

2. The building management system of claim 1 , wherein the function is influenced by a disturbance and wherein the cost function is influenced by the disturbance.

3. The building management system of claim 1 , wherein the self-optimizing control variable represents a derivative of the cost function with respect to the input.

4. The building management system of claim 1 , wherein the controller is configured to:

conduct a test by:

defining a nominal input vector having non-optimal values;

altering one element of the nominal input vector to create a modified input vector;

controlling the building equipment with the modified input vector; and

measuring and recording a steady state output; and

repeat the test a number of times to generate a set of non-optimal data.

5. The building management system of claim 4 , wherein the controller is configured to:

build, for each test, a first vector of changes in the modified input vector relative to the nominal input vector;

build, for each test, a second vector of changes in the steady state output relative to a nominal output, the nominal output having non-optimal values;

build a third vector of a change in each test in the cost relative to a nominal cost; and

calculate the values of elements of the matrix and the offset vector using the first vectors, the second vectors, and the third vector.

6. The building management system of claim 1 , wherein the building equipment comprises a VAV system; and

wherein the measured state comprises a supply air temperature.

7. The building management system of claim 6 , wherein the VAV system comprises a fan, a compressor, and an evaporator; and

wherein the cost is a ratio of heat removed in the evaporator divided by a combined amount of power exerted by the compressor and the fan.

8. A method comprising:

operating building equipment in accordance with an input to alter a variable state or condition of a building as a process function of the input while incurring a cost of operating the equipment as a cost function of the input;

calculating, by a controller, a value of a self-optimizing control variable as a function of a measured state of the building equipment; and

driving, by the controller, the value of the self-optimizing control variable towards a setpoint value by generating the input based on the self-optimizing control variable and providing the input to the building equipment;

wherein the function comprises multiplying the measured state by a matrix and adding an offset vector; and

wherein values of elements of the matrix and the offset vector are determined using a non-optimal reference.

9. The method of claim 8 , wherein the process function is influenced by a disturbance and wherein the cost function is influenced by the disturbance.

10. The method of claim 8 , wherein the self-optimizing control variable represents a derivative of the cost function with respect to the measured state.

11. The method of claim 8 , comprising:

conducting a test by:

defining a nominal input vector having non-optimal values;

altering one element of the nominal input vector to create a modified input vector;

controlling the building equipment with the modified input vector; and

measuring and recording a steady state output; and

repeating the test a number of times to generate a set of non-optimal data.

12. The method of claim 11 , comprising:

building, for each test, a first vector of changes in the modified input vector relative to the nominal input vector;

building, for each test, a second vector of changes in the steady state output relative to a nominal output, the nominal output having non-optimal values;

building a third vector of changes in the cost relative to a nominal cost in the tests; and

calculating the values of the elements of the matrix and the offset vector using the first vectors, the second vectors, and the third vector.

13. The method of claim 8 , wherein the building equipment comprises a VAV system; and wherein the measured state comprises a supply air temperature.

14. The method of claim 13 , wherein the VAV system comprises a fan, a compressor, and an evaporator; and

wherein the cost is a ratio of heat removed in the evaporator divided by a combined amount of power exerted by the compressor and the fan.

15. A method for operating building equipment and identifying a function for a self-optimizing control structure, comprising:

selecting a non-optimal reference input vector;

varying a first input of the non-optimal reference vector to generate a modified input vector;

controlling building equipment using the modified input vector to affect a variable state or condition of a building;

measuring an output of the building equipment;

determining an output difference between the output of the building equipment and a non-optimal reference output of the building equipment corresponding to the non-optimal reference input vector;

determining an input difference between the non-optimal reference input vector and the modified input vector;

determining a cost difference between a cost associated with controlling building equipment using the modified input vector and a reference cost associated with controlling building equipment using the non-optimal reference vector;

repeating a plurality of times to conduct a plurality of step tests and generate a dataset of output differences, input differences, and cost differences;

identifying an offset vector and a matrix using the dataset; and

identifying the function as a sum of the offset vector and the matrix multiplied by the output.

16. The method of claim 15 , wherein identifying the offset vector and the matrix using the dataset comprises:

building a first matrix comprising the output difference for each of the step tests;

building a second matrix comprising the input difference for each of the step tests;

building a third vector comprising the cost difference for each of the step tests; and

calculating the offset vector and the matrix using the first matrix, the second matrix, and the third vector.

17. The method of claim 16 , wherein calculating the matrix and the offset vector using the first matrix, the second matrix, and the third vector comprises:

calculating a coefficient vector using the third vector and the output difference;

filling a first derivative vector and a second derivative matrix with elements of the coefficient vector, wherein the first derivative vector corresponds to a first derivative of a cost function with respect to the output and the second derivative matrix corresponds to a second derivative of the cost function with respect to the output; and

calculating the offset vector and the matrix using the first derivative vector and a second derivative vector.

18. The method of claim 17 , wherein calculating the offset vector and the matrix using the first derivative vector and the second derivative vector comprises:

multiplying the first matrix by a process matrix to determine the offset vector; and

multiplying a transpose of the process matrix by the second derivative matrix to determine the matrix.

19. The method of claim 18 , comprising calculating the process matrix using the first matrix and the second matrix, wherein the process matrix corresponds to a physical behavior of the building equipment and a building served by the building equipment.

20. The method of claim 15 , comprising installing the building equipment at a building before identifying the function.