Ultra-Low Power Electronic Current Transducer Based on 4-Level Encoding and Weak-Light Commutation

The power supply solutions for the high-voltage side of active electronic current transducers (ECTs) suffer from dead zones in low currents, low energy conversion efficiency and so on. This paper proposes an ECT with a power consumption less than 4 milliwatts (mW) on the high-voltage side, which employs a battery-powered scheme to solve the problems. This scheme utilizes Light-emitting Diodes (LEDs) operating in a weak light mode with 1.8 V driving voltage and effective duty cycle of 0.16. A low-power driver, composed of double half-bridge, is adopted to merge two digital signals into a single 4-level one, halving the power consumption of the LEDs. A 4-level anti-interference decoding method integrating support vector machine and recurrent neural network is employed in the low voltage side. The experimental results demonstrate that the power consumption on the high-voltage side of ECT has a maximum value of 3.14 mW, with accuracy class of 0.2 S. The combined algorithm can correctly decode all the 4-level experimental datasets, as well as the expanded dataset generated by autoencoder and generative adversarial networks. Furthermore, it also exhibits excellent generalization capabilities and resistance to noise interference.