Neural network predictive control cost function designer
First Claim
1. A method for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase response to each of a range of known signals, the neural network control system including a neural network model and a cost function, the method comprising:
- selecting parameters used in a cost function;
selecting an input weight to be applied to a control output by the cost function;
selectively incorporating predicted future states generated by a neural network model;
iteratively applying a control input from a range of known signals;
calculating a control output in response to the control input;
determining a control system phase and a control system amplitude of the control output in response to the control input; and
combining a known plant phase with regards to a known signal equivalent to the control input and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
1 Assignment
0 Petitions
Accused Products
Abstract
A method, a computer-readable medium, and a system for tuning a cost function to control an operational plant are provided. A plurality of cost function parameters is selected. Predicted future states generated by the neural network model are selectively incorporated into the cost function, and an input weight is applied to a control input signal. A series of known signals are iteratively applied as control input inputs, and the cost output is calculated. A phase is taken of the control and plant outputs in response to each of the known signals and combined, thereby allowing effective combinations of the cost function parameters, the input weight, and the predicted future states to be identified.
-
Citations
72 Claims
-
1. A method for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase response to each of a range of known signals, the neural network control system including a neural network model and a cost function, the method comprising:
-
selecting parameters used in a cost function;
selecting an input weight to be applied to a control output by the cost function;
selectively incorporating predicted future states generated by a neural network model;
iteratively applying a control input from a range of known signals;
calculating a control output in response to the control input;
determining a control system phase and a control system amplitude of the control output in response to the control input; and
combining a known plant phase with regards to a known signal equivalent to the control input and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22)
-
-
23. A method for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase response to each of a range of known signals, the neural network control system including a neural network model and a cost function, the method comprising:
-
selecting parameters used in a cost function;
selecting an input weight to be applied to a control output signal by the cost function;
selectively incorporating predicted future states generated by a neural network model by incorporating at least two of the predicted future states generated by the neural network model;
iteratively applying a control input signal from a range of known signals;
calculating a control output in response to the control input signal;
determining a control system phase and a control system amplitude of the control output in response to the control input signal; and
combining a known plant phase with regards to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
-
24. A method for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase response to each of a range of known signals, the neural network control system including a neural network model and a cost function, the method comprising:
-
selecting parameters used in a cost function;
selecting a plant input weight to be applied to a control output signal by the cost function;
incorporating predicted future states generated by a neural network model by incorporating all of the predicted future states and combining each of the predicted future states with a forget factor such that a proportional weight is accorded each of the predicted future states;
iteratively applying a control input signal from a range of known signals;
calculating a control output in response to the control input signal;
determining a control system phase and a control system amplitude of the control output in response to the control input signal; and
combining a known plant phase with regards to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
-
25. A computer-readable medium for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the computer-readable medium comprising:
-
first computer program code means for selecting parameters used in a cost function;
second computer program code means for selecting an input weight to be applied to a control output signal by the cost function;
third computer program code means for selectively incorporating predicted future states generated by a neural network model;
fourth computer program code means for iteratively applying a control input signal from a range of known signals;
fifth computer program code means for calculating a control output in response to the control input signal;
sixth computer program code means for determining a control system phase and a control system amplitude of the control output in response to the control input signal; and
seventh computer program code means for combining a known plant phase responsive to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable. - View Dependent Claims (26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46)
-
-
47. A computer-readable medium for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the computer-readable medium comprising:
-
first computer program code means for selecting parameters used in a cost function;
second computer program code means for selecting an input weight to be applied to a control output signal by the cost function;
third computer program code means for selectively incorporating predicted future states generated by a neural network model by incorporating at least two of the predicted future states generated by the neural network model;
fourth computer program code means for iteratively applying a control input signal from a range of known signals;
fifth computer program code means for calculating a control output in response to the control input signal;
sixth computer program code means for determining a control system phase and a control system amplitude of the control output in response to the control input signal; and
seventh computer program code means for combining a known plant phase with regards to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
-
48. A computer-readable medium for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the computer-readable medium comprising:
-
first computer program code means for selecting parameters used in a cost function;
second computer program code means for selecting an input weight to be applied to a control output signal by the cost function;
third computer program code means for incorporating predicted future states generated by a neural network model and combining each of the predicted future states with a forget factor such that a proportional weight is accorded each of the predicted future states;
fourth computer program code means for iteratively applying a control input signal from a range of known signals;
fifth computer program code means for calculating a control output in response to the control input signal;
sixth computer program code means for determining a control system phase of the control output in response to the control input signal; and
seventh computer program code means for combining a known plant phase responsive to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
-
49. A system for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the system comprising:
-
a cost function parameter selector configured to select parameters used in a cost function;
an input weight selector configured to select an input weight to be applied to a control output signal by the cost function;
a predicted future state selector configured to selectively incorporate predicted future states generated by a neural network model;
an iterator configured to iteratively apply a control input signal from a range of known signals;
a cost function calculator configured to calculate a control output in response to the control input signal;
a control system phase determiner configured to determine a control system phase and a control system amplitude of the control output in response to the control input signal; and
a combiner configured to combine a known plant phase responsive to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable. - View Dependent Claims (50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70)
-
-
71. A system for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the system comprising:
-
a cost function parameter selector configured to select parameters used in a cost function;
an input weight selector configured to select an input weight to be applied to a control output signal by the cost function;
a predicted future state selector configured to selectively incorporate at least two of the predicted future states generated by a neural network model;
an iterator configured to iteratively apply a control input signal from a range of known signals;
a cost function calculator configured to calculate a control output in response to the control input signal;
a control system phase determiner configured to determine a control system phase of the control output in response to the control input signal; and
a combiner configured to combine a known plant phase responsive to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
-
72. A system for tuning a cost function used by a neural network control system configured to control an operational plant having a known plant phase responsive to each of a range of known signals, the neural network control system including a neural network model and a cost function, the system comprising:
-
a cost function parameter selector configured to select parameters used in a cost function;
an input weight selector configured to select an input weight to be applied to a control output signal by the cost function;
a predicted future state selector configured to select each of the predicted future states and combine each of the predicted future states with a forget factor such that a proportional weight is accorded each of the predicted future states;
an iterator configured to iteratively apply a control input signal from a range of known signals;
a cost function calculator configured to calculate a control output in response to the control input signal;
a control system phase determiner configured to determine a control system phase of the control output in response to the control input signal; and
a combiner configured to combine a known plant phase responsive to a known signal equivalent to the control input signal and the control system phase such that effectiveness of the cost function parameters, the input weight, and the selectively incorporated predicted future states is determinable.
-
Specification