| Popis: |
In this paper, a novel digital soft-start control method which employs a reference modification with a neural network predictor is presented to improve the transient characteristics in parallel operated dc-dc converters. To reduce the power consumption in parallel operated dc-dc converters, the mode of them is changed frequently from stand-by mode to operating mode, or vice versa. In the transient state where the parallel operation starts, a bad voltage drop occurs and it affects to the stable operation of power supplies. To avoid the voltage drop, the neural network predictor is adopted to predict the output voltages of each parallel operated converters and it modifies the reference value of a PID control in the transient state. The neural networks are trained to predict output voltages of each converter its former sensed data. In coordination with a conventional PID control, it operates to reduce the differences between output voltages and desired ones of each converter. From evaluation results, it is confirmed that the presented method has a superior transient characteristic compared to the conventional method. |
