基于TensorFlow神经网络的MCR-WPT系统负载与互感识别方法

TensorFlow Neural Network Based Load and Mutual Inductance Identification Method for Magnetic Coupling Resonant Wireless Power Transfer System

针对磁耦合谐振式无线电能传输(MCR-WPT)系统负载与互感识别精度低、速度慢等问题,提出一种基于TensorFlow神经网络的双LCC型MCR-WPT系统负载与互感识别方法。该方法基于TensorFlow深度学习框架,采用神经网络模型,将MCR-WPT系统的负载与互感识别问题等效为非线性方程的求解问题,进而转化为深度学习非线性拟合问题,并给出模型的训练方法,最后得到基于TensorFlow神经网络的MCR-WPT系统负载与互感识别模型。通过离线方式训练负载与互感识别模型,并将训练完成的识别模型导入微型控制器,只需要采集系统输入电流值和传输距离就能够实现负载与互感在线同时识别,识别速度快、精度高,有利于系统的实时控制,且成本较低、易于实现,有利于工程推广应用。

Aiming at the problems of low accuracy and slow speed in identifying the load and mutual inductance of the magnetic coupling resonant wireless power transfer(MCR-WPT) system, a TensorFlow neural network based identification method for the load and mutual inductance of the double-side LCC type MCR-WPT system is proposed. This method is based on the TensorFlow deep learning framework and adopts a neural network model. The load and mutual inductance identification problem of the MCRWPT system is equivalent to the problem of solving nonlinear equations, which is then transformed into a deep learning nonlinear fitting problem. The training method of the model is given and the TensorFlow neural network based identification model for the load and mutual inductance of MCR-WPT system is finally obtained. The identification model for the load and mutual inductance is trained offline and then the trained model is imported into the micro controller. It only needs to collect the input current value and transmission distance of the system to realize the online simultaneous identification of the load and mutual inductance. The identification speed is fast and the accuracy is high, which is conducive to the real-time control of the system. And the model is low in cost and easy to implement, which is conducive to the engineering popularization and application.