6 December 2022 in IEEE Transactions on Industrial Electronics
In this article, a multiscenario parameter optimization method for an active disturbance rejection controller (ADRC) of a permanent magnet synchronous motor (PMSM) based on deep reinforcement learning (DRL) is proposed as MSPO-DRL. The parameter setting of the nonlinear ADRC has always been one of the difficulties affecting the optimal performance of ADRC, and there will be different optimal parameters under different control requirements. In this article, an artificial intelligence algorithm is introduced into the parameter optimization process of ADRC, and the DRL parameter optimization model that can automatically optimize and adjust the ADRC parameters in different application scenarios is constructed so that the ADRC can achieve the best control effect conveniently and the limitation of current methods has been solved. ADRC is applied to the speed loop of flux weakening control of PMSM for more electric aircraft, and the mathematical model of ADRC in this environment is built above all. The Markov decision process is integrated into ADRC. The interface module and reward function between ADRC and MSPO-DRL are designed. The concept of the ADRC control scenario is defined and integrated into the concept of the Markov decision process to improve the generalization of DRL. Then, the MSPO-DRL model is established, and the deep deterministic gradient strategy is used as the gradient descent strategy to converge the parameters optimization. After the model learning is completed, different environmental conditions are randomly selected for simulation and experiments to verify the optimization effect and generalization performance of the algorithm. Optimizations that are carried out by the heuristic algorithms are used for comparisons, and the superiority and feasibility of the proposed algorithm are verified.
25 May 2023 in IEEE Transactions on Power Electronics
This article proposes an improved model-free active disturbance rejection deadbeat predictive current control(ADRDPCC) method for permanent magnet synchronous motor (PMSM) used in more electric aircraft (MEA) based on the data driven method, which is used to solve the parameter mismatch problem of deadbeat predictive current control (DPCC) and improve the performance of PMSM control system. DPCC model applied to the current loop of the MEA motor is established as the main control strategy of the system. The principle of active disturbance rejection control is combined with DPCC, and the ADRDPCC structure is formed to optimize the control strategy. A specific extended state observer (ESO) of ADRDPCC is designed to track the internal disturbance caused by parameter mismatch and the external disturbance in real time. DPCC is used as the control law of the ADRDPCC structure to predict the current and output the reference voltage based on the observation of ESO. A deep reinforcement learning model based on ADRDPCC is designed and trained based on the data-driven method. The trained model can compensate and optimize ADRDPCC based on the disturbance observed by ESO and the observed control state of PMSM. The simulated and experimental results show the superiority of the proposed method.
10 January 2023 in IEEE Transactions on Energy Conversion
In this article, a novel active disturbance rejection control (ADRC) based on deep reinforcement learning (DRL) is proposed to improve the performance of permanent magnet synchronous motor (PMSM) for more electric aircraft (MEA). MEA motors have the requirements of safety and stability so that a new ADRC method is put forward based on the limitation of nonlinear error attenuation function of traditional ADRC. The flux weakening control model of PMSM is firstly established. Then the ADRC model is built and applied to the speed loop of the control system. In order to reduce the number of control parameters and the jitter of the control law, deep neural network is employed to replace the traditional control law. Markov decision process is integrated into the novel ADRC to establish DRL model. A method based on twin delayed deep deterministic (TD3) policy gradient algorithm is proposed to train the neural network and optimize the DRL model. Model predictive control and traditional ADRC are used as comparison algorithms. Simulation and experiments show the effectiveness and superiority of the proposed method.
17 August 2023 in IEEE Transactions on Power Electronics
The article proposes an enhanced deadbeat predictive current control for the permanent-magnet synchronous motor using an iterative sliding mode observer to improve the current control accuracy. The inevitable deadtime and distortion of flux linkage will lead to large periodic current ripples related to the speed, which may degrade the performance of the motor control. To overcome these problems and improve the robustness of the controller, the iterative learning control (ILC) and sliding mode observer are combined, and feedforward to the deadbeat controller for prediction accuracy improvement. Moreover, in order to satisfy the requirements of the variable speed operation of the motor, the ILC is enhanced to achieve spatial learning ability. Experimental results are finally implemented on STM32-based platform to validate the effectiveness of the proposed method for current ripples suppression and adaptability to variable speed.
23 October 2023 in IEEE Journal of Emerging and Selected Topics in Power Electronics
The operation of permanent magnet synchronous motors is subject to many disturbances due to structural characteristics and operating situations. Particularly at low speeds, periodic disturbances, for instance, offset current error, cogging torque, etc., have a greater impact on the stable operation of the motor. For periodic disturbances, the conventional disturbance observer suffers from high delay and static error. Even though the observer bandwidth can be increased to enhance the tracking ability to perturbations, this will introduce high frequency noise at the same time. To solve these problems, an improved linear active disturbance rejection controller in this paper is designed. By adding an error proportional feedback term to the extended state observer, the capability of system to observe load perturbations is tremendously enhanced. A high-pass filter is employed at the state error feedback as a speed compensator. This method can reduce the speed ripples feasibly without complex structures and tedious calculations. The stability of the proposed method is demonstrated by theoretical analysis, and both simulation and experimental results are given to verify the effectiveness of the proposed method.
16 March 2022 in IEEE Transactions on Energy Conversion
The arc permanent magnet motor (arc motor) is widely used in large telescope for its high efficiency and high power density. Due to the unique structure, the periodic end torque, cogging torque and flux harmonics will cause certain speed ripples which can affect the performance of the drive system. To solve these problems, the paper proposes a novel model predictive control (MPC) based on extended state observer and iterative learning control, where the MPC is equipped in speed loop and the disturbances estimated by the conventional extended state observer (ESO) are fed forward to the controller to improve the ability of disturbance rejection. A position dependent iterative learning control (ILC) is used to estimate the periodic disturbances at different speeds. The MPC calculates the q axis current reference with the compensation of ESO and ILC, which can have a fast response and disturbance rejection ability for periodic and nonperiodic disturbances. The stability of the proposed method is demonstrated by theoretical analysis. The experimental results validate the effectiveness of the proposed method at different speeds.