Using non-parametric statistics to identify the best pathway for supplying hydrogen as a road transport fuel

摘要

The wealth of estimates quantifying the well-to-tank (WTT) impacts of hydrogen vary significantly. This variation is due to both methodology and the chosen production pathway (gasification, electrolysis, or steam reforming). The statistical distribution of the WTT estimates is non-Gaussian and this work demonstrates the adaptive kernel density estimator as a robust, non-parametric statistical method for determining the underlying probability density function. The approach is flexible, expandable and can be used to investigate the development of hydrogen supply pathways through time. The adaptive kernel density estimator outperforms the first generation (oversmoothed and least squares cross-validation), second generation Sheather and Jones Plug-In and the median. In particular, it represents the multimodal features of the data set better than both the first and second generation methods with less variability than the least squares cross-validation approach. The peak of the distribution represents the most likely pathway (best estimate) for supplying hydrogen. This work suggests that the overall best estimate for supplying hydrogen is by natural gas from Europe via central reforming, subject to a trade-off between the energy impacts and the resultant emissions. Through time, the overall hydrogen production process has become more energy efficient at the expense of greater emissions per MJ delivered to the tank. The best-in-class pathway is that with the lowest greenhouse gas emissions per MJ hydrogen delivered and represents the state-of-the-art. Overall, the best-in-class pathway combination for providing hydrogen is by electricity from renewables via electrolysis.