Systems, methods and devices for approximate dynamic programming vector controllers for operation of IPM motors in linear and over modulation regions

1. A method for controlling an interior-mounted permanent magnet (IPM) alternating-current (AC) electrical machine by:

providing a pulse-width modulated (PWM) converter operably connected between an electrical power source and the IPM AC electrical machine;

providing a neural network vector control system operably connected to the PWM converter, the neural network vector control system comprising a current-loop neural network configured to implement an approximate dynamic programming (ADP) algorithm, wherein the current-loop neural network is trained to minimize a cost function of the ADP algorithm using a forward accumulation through time (“FATT”) algorithm;

receiving a plurality of inputs at the current-loop neural network;

outputting a first compensating dq-control voltage from the current-loop neural network, wherein the current-loop neural network is configured to optimize the first compensating dq-control voltage based on the plurality of inputs;

providing a feedforward controller operably connected to the PWM converter;

receiving reference current inputs at the feedforward controller, wherein the inputs at the feedforward controller comprise a reference d-axis current, isd*, and a reference q-axis current, isq*;

outputting a second compensating dq-control voltage from the feedforward controller, wherein the feedforward controller is configured to regulate the second compensating dq-control voltage based on the reference current inputs; and

controlling the PWM converter using the first compensating dq-control voltage and the second compensating dq-control voltage,

wherein the feedforward controller is designed based on default parameters of the IPM AC electrical machine.

2. The method of claim 1 , wherein the compensating dq-control voltages from the current-loop neural network and feedforward controller are further added together to form a resultant compensating dq-control voltage; and

controlling the PWM converter using the resultant compensating dq-control voltage.

3. The method of claim 2 , wherein the resultant compensating dq-control voltage is normalized based on the maximum possible dq control voltage under six-step modulation condition of the PWM converter; and

controlling the PWM converter using the normalized resultant compensating dq-control voltage.

4. The method of claim 3 , wherein the normalized resultant compensating dq-control voltage is further processed through a saturation function;

limiting the amplitude of the normalized dq-control voltage to 1 or less, which is the six-step modulation index; and

controlling the PWM converter using the processed compensating dq-control voltage.

5. The method of claim 4 , wherein the normalized resultant compensating dq-control voltage is allowed to be larger than 1 , which is the six-step modulation index, for controlling the IPM AC electrical machine in over modulation and six-step modulation regions.

6. The method of claim 5 , wherein the processing through the saturation function captures transient information in terms of amplitude and angle of the normalized resultant compensating dq-control voltage for controlling the IPM AC electrical machine in over modulation and six-step modulation regions.

7. The method of claim 4 , wherein the normalized resultant compensating dq-control voltage is allowed to change rapidly from smaller than 1 to larger than 1 before the saturation process; and

providing control capability when the IPM AC electrical machine is operating between over modulation and six-step modulation regions.

8. The method of claim 1 , wherein changes to the plurality of inputs to the current-loop neural network are blocked when the IPM AC electrical machine completely operates in a six-step modulation region; and

controlling the IPM AC electrical machine to operate on a boundary of the six-step modulation region.

9. The method of claim 8 , wherein the blocking applied to the plurality of inputs to the current-loop neural network is released when the operation of an IPM machine returns from the six-step modulation region to an over modulation region or a linear modulation region; and

recovering the current-loop neural network to its full control capability.

10. A system for controlling an interior-mounted permanent magnet (IPM) alternating-current (AC) electrical machine comprising:

a pulse-width modulated (PWM) converter operably connected between an electrical power source and the IPM AC electrical machine;

a neural network vector control system operably connected to the PWM converter, the neural network vector control system comprising a processor and a current-loop neural network configured to implement an approximate dynamic programming (ADP) algorithm, wherein the current-loop neural network is trained to minimize a cost function of the ADP algorithm using a forward accumulation through time (“FATT”) algorithm, said neural network vector control system configured to:

receive a plurality of inputs at the current-loop neural network; and

output a first compensating dq-control voltage from the current-loop neural network, wherein the current-loop neural network is configured to optimize the first compensating dq-control voltage based on the plurality of inputs; and

a feedforward controller operably connected to the PWM converter, wherein the feedforward controller:

receives reference current inputs, wherein the inputs at the feedforward controller comprise a reference d-axis current, i sd * , and a reference q-axis current, i sq *;

outputs a second compensating dq-control voltage from the feedforward controller, wherein the feedforward controller regulates the second compensating dq-control voltage based on the reference current inputs; and

controls the PWM converter using the first compensating dq-control voltage and the second compensating dq-control voltage.

11. The system of claim 10 , wherein the compensating dq-control voltages from the current-loop neural network and the feedforward controller are further added together to form a resultant compensating dq-control voltage; and

the PWM converter is controlled using the resultant compensating dq-control voltage.

12. The system of claim 11 , wherein the resultant compensating dq-control voltage is normalized based on the maximum possible dq control voltage under six-step modulation condition of the PWM converter; and

the PWM converter is controlled using the normalized resultant compensating dq-control voltage.

13. The system of claim 12 , wherein the normalized resultant compensating dq-control voltage is further processed through a saturation function;

the amplitude of the normalized dq-control voltage is limited to 1 or less, which is the six-step modulation index; and

the PWM converter is controlled using the processed compensating dq-control voltage.

14. The system of claim 13 , wherein the normalized resultant compensating dq-control voltage is allowed to be larger than 1, which is the six-step modulation index, for controlling the IPM AC electrical machine in over modulation and six-step modulation regions.

15. The system of claim 14 , wherein the processing through the saturation function captures transient information in terms of amplitude and angle of the normalized resultant compensating dq-control voltage for controlling the IPM AC electrical machine in over modulation and six-step modulation regions.

16. The system of claim 13 , wherein the normalized resultant compensating dq-control voltage is allowed to change rapidly from smaller than 1 to larger than 1 before the saturation process; and

control capability is provided when the IPM AC electrical machine is operating between over modulation and six-step modulation regions.

17. The system of claim 10 , wherein changes to the plurality of inputs to the current-loop neural network are blocked when the IPM AC electrical machine completely operates in a six-step modulation region; and

the IPM AC electrical machine is controlled to operate on a boundary of the six-step modulation region.

18. The system of claim 17 , wherein the blocking applied to the plurality of inputs to the current-loop neural network is released when the operation of an IPM machine returns from the six-step modulation region to an over modulation region or a linear modulation region; and the current-loop neural network is recovered to its full control capability.