Updated unit process data for coal-based energy in China including parameters for overall dispersions

Abstract

Chinese coal power generation is part of the life cycle of most products and the largest single source for many emissions. Reducing these emissions has been a priority for the Chinese government over the last decade, with improvements made by replacing older power plants, improving thermal efficiency and installing air pollution control devices. In the present research, we aim to acknowledge these improvements and present updated unit process data for Chinese coal power. In the course of doing so, we also explore the implementation and interpretation of overall dispersions related to a generically averaged process, such as Chinese coal power. In order to capture geographical and temporal dispersions, updated unit process data were calculated for Chinese coal power at both a national and a provincial level. The updated unit process dataset was also propagated into life cycle inventory (LCI) ranges using Monte Carlo simulations, allowing for discrepancies to be evaluated against the most commonly used inventory database (ecoinvent) and overall dispersions to be shown for some selected provinces. Compared to ecoinvent, the updated dataset resulted in reductions with between 8 and 67 % for all evaluated inventory flows except for dinitrogen monoxide (N2O). However, interprovincial differences in emissions diverged with up to 250 %. A random outcome in a few Monte Carlo runs was inverted operators, where positive values became negative or the other way around. This is a known possible outcome of matrix calculations that needs to be better evaluated when interpreting propagated outcomes. The present manuscript provides recommendations on how to implement and interpret dispersions propagated into LCI results. In addition, updated and easily accessible unit process data for coal power plants averaged across China and for individual provinces are presented, with clear distinctions of inherent uncertainties, spread (variance) and unrepresentativeness. Recommendations are also provided for future research and software developments.