The development of artificial neural network space vector PWM and diagnostic controller for voltage source inverter

Data Mining Applications to Fault Diagnosis in Power Electronic Systems: A Systematic Review

“…In [23] and [24], the ANN applications are employed to detect OC faults and hard faults (which refer to the analogue part of the circuit) in a delta-sigma converter, respectively. Transistor switch faults in a voltage source inverter have been identified in [25] by a new model of ANN based on the controller of space vector modulation.…”

“…The encoding and decoding operations via AE for the input variable x is performed as follow: 𝑓 = 𝑔(𝜔 𝑥 * 𝑥 + 𝑏 𝑥 ) (25) 𝑦 = 𝑔(𝜔 𝑦 * 𝑥 + 𝑏 𝑦 ) (26) where 𝑦 is the output features and 𝑓 denote the network features. 𝜔 𝑥 , 𝜔 𝑦 , 𝑏 𝑥 , and 𝑏 𝑦 represent the input-to-hidden weights, hidden-to-output weights, bias of hidden units, and bias of output units, respectively.…”

A Neural Network Based Space Vector PWM for Permanent Magnet Synchronous Motor Drive

“…The input of the network is the phase angle of the reference voltage vector * and the outputs are fifteen turn-on functions g A ( *), g B ( *), g Fig. 3 Complete ANN-SVPWM controller g 21B ( *), g 11C ( *), g 21C ( *), g 11A ( *), g 21A ( *), g 11B ( *), g 21B ( *), g 11C ( *), g 21C ( *) for the three phases A, B, and C. The equations (3,4,5) are used for generating neural network training data. This subnet use tansig neurons in the hidden layers and purelin neurons in the output layer, and it is obtained by training with trainlm function which is implemented by Levenberg-Marquardt algorithm.…”

“…The performance of the inverter is compared for the different architectures of NN based SVM. A kind of artificial neural network based space vector modulation and diagnostic controller scheme for a voltage source inverter is proposed in [5] that the network architecture is too complicated with 8 subnets and about 122 neurons. The paper proposes a more optimal back-propagation type multi-layer feed forward ANN-based SVM that fully covers the undermodulation and overmodulation regions by using method of linear modulation between two limit trajectories.…”