Comparative Analysis of the Noise Elimination Based on the Wtbmra with Various Decomposition/Reconstruction Levels Level for a PEMFC

Abstract

Recently, in order to fulfill the requirements for decreasing air pollution caused by the use of fossil fuels and achieving low or zero emission, some renewable sources such as solar, hydrogen fuel cells, wind, geothermal, biomass, and tidal energy have been taken center stage under the international spotlight. Among them, the fuel cell has been broadly considered as a renewable and environmental energy-conversion solution for settling aforementioned problem. Specifically, the polymer electrolyte membrane fuel cell (PEMFC) has numerous advantages including low-operating temperature, fast start-up ability, high current density, and high potential for low cost when compared to other sources. Then, in order to utilize this PEMFC efficiently, it is surely necessary to have an optimal fuel cell management system. For this purpose, an equivalent electrical-circuit model (ECM) should be elaborately designed to describe the dynamic characteristic of the fuel cell terminal voltage in a PEMFC. Therefore, the most important thing in the ECM-based fuel cell management is to achieve an accurate experimental PEMFC voltage signal (EPVS). Unfortunately, there may be at risk for instantaneous and unwanted sensing of noisy. This absolutely results in erroneous fuel cell management system caused by noise-including EPVS loss. As a result, there should be an imperative to investigate a sophisticated approach to eliminate the noise from this inevitable loss. So far, no definitive solution has been provided to settle this problem.This approach newly gives insight to the design and implementation of the wavelet transform-based multi resolution analysis (WTBMRA) for noise elimination and ultimately an efficient PEMFC operation accomplishment. As is well known, the WTBMRA is more remarkable for processing of signal that having non-stationary and transient phenomena characteristics. Fortunately, because of various fuel cell voltage forms by different time intervals and magnitudes, the EPVS can sufficiently considered as an original signal in the WTBMRA. The WTBMRA has the decomposition and reconstruction abilities with a vigorous function of both time and frequency localization of the EPVS. Therefore, it is capable of obtaining low- and high-frequency components An and Dn . Like other studies dealing with noise elimination, there are three steps for noise elimination in this approach. The noticeable difference between previous approaches and the proposed approach is to compare the analytic results of noise elimination with various decomposition levels based on the WTBMRA. According to this analysis, it can be expected to select an optimal decomposition level that clearly shows the high performance on noise elimination under the same condition of mother wavelet. Basically, this approach finally considered the Daubechies as a mother wavelet which belongs to the family of orthogonal wavelet filtering the WTBMRA. For reference, the Daubechies wavelet (db) is extensively used in solving a board rage of problems due to its orthogonal property. In this approach, the decomposition and reconstruction scale in the WTBMRA is selected as 3. With regard to the signal-to-noise (SNR) ration, all comparative analyses are elaborately evaluated for determining the optimal decomposition and reconstruction levels. Two techniques on noise elimination such as hard-thresholding and soft-thresholding with threshold value by VisuShrink calculation are implemented. Consequently, our analytic results sufficiently the suggest the clear comparison by showing the SNR difference using two techniques. From this approach, there are two conclusions. The first conclusion is that the performance on noise elimination using soft-thresholding is always inferior to that of hard-thresholding. The second conclusion is that there are the highest SNR values at each specified decomposition and reconstruction levels. Our experimental apparatus basically designed for obtaining the noise-including EPVS from the PEMFC by the ‘Materials and Electro-Chemistry Laboratory’ in Inha University. Figure 1