摘要
A reliable, efficient and economical power supply for dielectric barrier discharge (DBD) is essential for its industrial applications. However, the equivalent load parameters complicare the design of power supply as they are variable and varied nonlinearly in response to varied voltage and power. In this paper the equivalent electrical parameters of DBD are predicted using a neural network, which is beneficial for the design of power supply and helps to investigate how the electrical parameters influence the equivalent load parameters. The electrical parameters includ- ing voltage and power are determined to be the inputs of the neural network model, as these two parameters greatly influence the discharge type and the equivalent DBD load parameters which are the outputs of the model. The voltage and power are decoupled with pulse density modula- tion (PDM) and hence the impact of the two electrical parameters is discussed individually. The neural network model is trained with the back-propagation (BP) algorithm. The obtained neural network model is evaluated by the relative error, and the prediction has a good agreement with the practical values obtained in experiments.
A reliable, efficient and economical power supply for dielectric barrier discharge (DBD) is essential for its industrial applications. However, the equivalent load parameters complicare the design of power supply as they are variable and varied nonlinearly in response to varied voltage and power. In this paper the equivalent electrical parameters of DBD are predicted using a neural network, which is beneficial for the design of power supply and helps to investigate how the electrical parameters influence the equivalent load parameters. The electrical parameters includ- ing voltage and power are determined to be the inputs of the neural network model, as these two parameters greatly influence the discharge type and the equivalent DBD load parameters which are the outputs of the model. The voltage and power are decoupled with pulse density modula- tion (PDM) and hence the impact of the two electrical parameters is discussed individually. The neural network model is trained with the back-propagation (BP) algorithm. The obtained neural network model is evaluated by the relative error, and the prediction has a good agreement with the practical values obtained in experiments.
基金
supported by National Natural Science Foundation of China(Nos.51107115,11347125,51407156)
China Postdoctoral Science Foundation(Nos.20110491766,2014M551735)