Site-wide operations management optimization for manufacturing and processing control

1. A computer-implemented method comprising:

applying, by a processor, an optimization algorithm to a control system of a processing plant based on whether the control system is guided by a linear-based predictive model or a non-linear-based predictive model in which a gradient is available, wherein applying the optimization algorithm comprises:

responsive to the control system being guided by a linear-based predictive model, applying a mixed-integer linear program to the linear-based predictive model, wherein the mixed-integer linear program comprises one or more individual decision trees, wherein a topology of each decision tree is fixed such that, for each feature vector x in the respective decision tree, there is a unique path leading to a leaf node of the tree from the root of the tree; and

responsive to the control system being guided by a non-linear-based predictive model, applying a two-level augmented Lagrangian method to the non-linear-based predictive model, wherein an augmented Lagrangian method is applied to non-linear coupling constraints in an outer level of the two-level augmented Lagrangian method and a multi-block alternating direction method of multipliers (ADMM) is applied to linear coupling constraints in an inner level of the two-level augmented Lagrangian method;

calculating, by the processor, a set variable using the optimization algorithm;

predicting, by the processor, an output based on the calculated set variable;

comparing, by the processor, an actual output at the processing plant to the predicted output; and

suspending, by the processor, a physical process at the processing plant in response to the actual output being a threshold value apart from the predicted output.

2. The computer-implemented method of claim 1 , wherein calculation of the set variable is based at least in part on a measured volume of fluid processed by a predecessor processing plant.

3. The computer-implemented method of claim 1 , wherein the suspending a physical process comprises disabling an actuator.

4. The computer-implemented method of claim 1 , wherein the non-linear based model includes a deep neural network (DNN).

5. The computer-implemented method of claim 1 , wherein the processing plant is member of a network of processing plants.

6. A system comprising:

a memory having computer readable instructions; and

one or more processors configured to execute the computer readable instructions, the computer readable instructions configured to control the one or more processors to perform operations comprising:

applying an optimization algorithm to a control system of a processing plant based on whether the control system is guided by a linear-based predictive model or a non-linear-based predictive model in which a gradient is available, wherein applying the optimization algorithm comprises:

responsive to the control system being guided by a linear-based predictive model, applying a mixed-integer linear program to the linear-based predictive model, wherein the mixed-integer linear program comprises one or more individual decision trees, wherein a topology of each decision tree is fixed such that, for each feature vector x in the respective decision tree, there is a unique path leading to a leaf node of the tree from the root of the tree; and

responsive to the control system being guided by a non-linear-based predictive model, applying a two-level augmented Lagrangian method to the non-linear-based predictive model, wherein an augmented Lagrangian method is applied to non-linear coupling constraints in an outer level of the two-level augmented Lagrangian method and a multi-block alternating direction method of multipliers (ADMM) is applied to linear coupling constraints in an inner level of the two-level augmented Lagrangian method;

calculating a set variable using the optimization algorithm;

predicting an output based on the calculated set variable;

comparing an actual output at the processing plant to the predicted output; and

suspending a physical process at the processing plant in response to the actual output being a threshold value apart from the predicted output.

7. The system of claim 6 , wherein calculation of the set variable is based at least in part on a measured volume of fluid processed by a predecessor processing plant.

8. The system of claim 6 , wherein the suspending a physical process comprises disabling an actuator.

9. The system of claim 6 , wherein the non-linear based model includes a deep neural network (DNN).

10. The system of claim 6 , wherein the processing plant is member of a network of processing plants.

11. A non-transitory computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform operations comprising:

applying an optimization algorithm to a control system of a processing plant based on whether the control system is guided by a linear-based predictive model or a non-linear-based predictive model in which a gradient is available, wherein applying the optimization algorithm comprises:

responsive to the control system being guided by a linear-based predictive model, applying a mixed-integer linear program to the linear-based predictive model, wherein the mixed-integer linear program comprises one or more individual decision trees, wherein a topology of each decision tree is fixed such that, for each feature vector x in the respective decision tree, there is a unique path leading to a leaf node of the tree from the root of the tree; and

responsive to the control system being guided by a non-linear-based predictive model, applying a two-level augmented Lagrangian method to the non-linear-based predictive model, wherein an augmented Lagrangian method is applied to non-linear coupling constraints in an outer level of the two-level augmented Lagrangian method and a multi-block alternating direction method of multipliers (ADMM) is applied to linear coupling constraints in an inner level of the two-level augmented Lagrangian method;

calculating a set variable using the optimization algorithm;

predicting an output based on the calculated set variable;

comparing an actual output at the processing plant to the predicted output; and

suspending a physical process at the processing plant in response to the actual output being a threshold value apart from the predicted output.

12. The computer program product of claim 11 , wherein calculation of the set variable is based at least in part on a measured volume of fluid processed by a predecessor processing plant.

13. The computer program product of claim 11 , wherein the suspending a physical process comprises disabling an actuator.

14. The computer program product of claim 11 , wherein the non-linear based model includes a deep neural network (DNN).