Abstract: An example method for controlling a DC/DC converter or a standalone DC microgrid comprises an artificial neural network (ANN) based control method integrated with droop control. The ANN is trained to implement optimal control based on approximate dynamic programming. In one example, Levenberg-Marquardt (LM) algorithm is used to train the ANN, where the Jacobian matrix needed by LM algorithm is calculated via a Forward Accumulation Through Time algorithm. The ANN performance is evaluated by using power converter average and switching models. Performance evaluation shows that a well-trained ANN controller has a strong ability to maintain voltage stability of a standalone DC microgrid and manage the power sharing among the parallel distributed generation units. Even in dynamic and power converter switching environments, the ANN controller shows an ability to trace rapidly changing reference commands and tolerate system disturbances, and operate the DC/DC converter or the microgrid in standalone conditions.

Abstract: A dc/dc buck converter controller comprises an artificial neural network (ANN) controller comprising an input layer and an output layer. The input layer receives an error value of a dc/dc buck converter and an integral of the error value. The output layer produces an output error voltage. A pulse-width-modulation (PWM) sliding mode controller (SMC) is configured to receive the output error voltage and produce a control action voltage by multiplying the output error voltage with a PWM gain. A drive circuit is configured to receive the control action voltage and provide a drive voltage to an input switch of the dc/dc buck converter. A PI control block is configured to modify a reference output voltage based on a maximum current constraint input. A locking circuit is configured to maintains the output error voltage at the saturation limit of the PWM SMC to manage a maximum duty cycle constraint.

Abstract: Described herein is a neural network-based vector control method for the induction motor. The disclosure includes an approach to implement optimal vector control for an induction motor by using an NN; a NN controller to substitute two decoupled proportional-integral (PI) controllers in current loop; and, a mechanism to train the NN controller by using a Levenberg-Marquardt (LM)+forward accumulation through time (FATT) algorithm.

Abstract: A dc/dc buck converter controller comprises an artificial neural network (ANN) controller comprising an input layer and an output layer. The input layer receives an error value of a dc/dc buck converter and an integral of the error value. The output layer produces an output error voltage. A pulse-width-modulation (PWM) sliding mode controller (SMC) is configured to receive the output error voltage and produce a control action voltage by multiplying the output error voltage with a PWM gain. A drive circuit is configured to receive the control action voltage and provide a drive voltage to an input switch of the dc/dc buck converter. A PI control block is configured to modify a reference output voltage based on a maximum current constraint input. A locking circuit is configured to maintains the output error voltage at the saturation limit of the PWM SMC to manage a maximum duty cycle constraint.

Abstract: Described herein is a neural network-based vector control method for the induction motor. The disclosure includes an approach to implement optimal vector control for an induction motor by using an NN; a NN controller to substitute two decoupled proportional-integral (PI) controllers in current loop; and, a mechanism to train the NN controller by using a Levenberg-Marquardt (LM)+forward accumulation through time (FATT) algorithm.

Abstract: Described herein is a neural network-based vector control method for the induction motor. The disclosure includes an approach to implement optimal vector control for an induction motor by using an NN; a NN controller to substitute two decoupled proportional-integral (PI) controllers in current loop; and, a mechanism to train the NN controller by using a Levenberg-Marquardt (LM)+forward accumulation through time (FATT) algorithm.

Abstract: An example method for controlling a DC/DC converter or a standalone DC microgrid comprises an artificial neural network (ANN) based control method integrated with droop control. The ANN is trained to implement optimal control based on approximate dynamic programming. In one example, Levenberg-Marquardt (LM) algorithm is used to train the ANN, where the Jacobian matrix needed by LM algorithm is calculated via a Forward Accumulation Through Time algorithm. The ANN performance is evaluated by using power converter average and switching models. Performance evaluation shows that a well-trained ANN controller has a strong ability to maintain voltage stability of a standalone DC microgrid and manage the power sharing among the parallel distributed generation units. Even in dynamic and power converter switching environments, the ANN controller shows an ability to trace rapidly changing reference commands and tolerate system disturbances, and operate the DC/DC converter or the microgrid in standalone conditions.

Abstract: Described herein is an approximate dynamic programming (ADP) vector controller for control of a permanent magnet (PM) motor. The ADP controller is developed using the full dynamic equation of a PM motor and implemented using an artificial neural network (ANN). A feedforward control strategy is integrated with the ANN-based ADP controller to enhance the stability and transient performance of the ADP controller in both linear and over modulation regions. Simulation and hardware experiments demonstrate that the proposed ANN-based ADP controller can track large reference changes with high efficiency and reliability for PM motor operation in linear and over modulation regions.

Abstract: An example system for controlling a grid-connected energy source using a neural network is described herein. The example system can include a grid-connected converter (“GCC”) operably coupled between an electrical grid and an energy source, a n-order grid filter (e.g., where n is an integer greater than or equal to 2) operably coupled between the electrical grid and the GCC, and a nested-loop controller. The nested-loop controller can have inner and outer control loops and can be operably coupled to the GCC. A d-axis loop can control real power, and a q-axis loop can control reactive power. Additionally, the inner control loop can include a neural network that is configured to optimize dq-control voltages for controlling the GCC. The neural network can account for circuit dynamics of the n-order grid filter while optimizing the dq-control voltages.

Abstract: Described herein is an approximate dynamic programming (ADP) vector controller for control of a permanent magnet (PM) motor. The ADP controller is developed using the full dynamic equation of a PM motor and implemented using an artificial neural network (ANN). A feedforward control strategy is integrated with the ANN-based ADP controller to enhance the stability and transient performance of the ADP controller in both linear and over modulation regions. Simulation and hardware experiments demonstrate that the proposed ANN-based ADP controller can track large reference changes with high efficiency and reliability for PM motor operation in linear and over modulation regions.

Abstract: An example method for controlling an AC electrical machine can include providing a PWM converter operably connected between an electrical power source and the AC electrical machine and providing a neural network vector control system operably connected to the PWM converter. The control system can include a current-loop neural network configured to receive a plurality of inputs. The current-loop neural network can be configured to optimize the compensating dq-control voltage. The inputs can be d- and q-axis currents, d- and q-axis error signals, predicted d- and q-axis current signals, and a feedback compensating dq-control voltage. The d- and q-axis error signals can be a difference between the d- and q-axis currents and reference d- and q-axis currents, respectively. The method can further include outputting a compensating dq-control voltage from the current-loop neural network and controlling the PWM converter using the compensating dq-control voltage.

Abstract: An example system for controlling a grid-connected energy source using a neural network is described herein. The example system can include a grid-connected converter (“GCC”) operably coupled between an electrical grid and an energy source, a n-order grid filter (e.g., where n is an integer greater than or equal to 2) operably coupled between the electrical grid and the GCC, and a nested-loop controller. The nested-loop controller can have inner and outer control loops and can be operably coupled to the GCC. A d-axis loop can control real power, and a q-axis loop can control reactive power. Additionally, the inner control loop can include a neural network that is configured to optimize dq-control voltages for controlling the GCC. The neural network can account for circuit dynamics of the n-order grid filter while optimizing the dq-control voltages.

Abstract: Described herein is a neural network-based vector control method for the induction motor. The disclosure includes an approach to implement optimal vector control for an induction motor by using an NN; a NN controller to substitute two decoupled proportional-integral (PI) controllers in current loop; and, a mechanism to train the NN controller by using a Levenberg-Marquardt (LM)+forward accumulation through time (FATT) algorithm.

Abstract: An example method for controlling an AC electrical machine can include providing a PWM converter operably connected between an electrical power source and the AC electrical machine and providing a neural network vector control system operably connected to the PWM converter. The control system can include a current-loop neural network configured to receive a plurality of inputs. The current-loop neural network can be configured to optimize the compensating dq-control voltage. The inputs can be d- and q-axis currents, d- and q-axis error signals, predicted d- and q-axis current signals, and a feedback compensating dq-control voltage. The d- and q-axis error signals can be a difference between the d- and q-axis currents and reference d- and q-axis currents, respectively. The method can further include outputting a compensating dq-control voltage from the current-loop neural network and controlling the PWM converter using the compensating dq-control voltage.