| Popis: |
[Objective] The objective of this study was to develop a precise calculation method for fuel-powered private car travel carbon emissions, utilizing big data from car owners and employing machine learning techniques. This method aims to address the limitations of existing macro-level studies and provide a scientific basis for accurately estimating road transport carbon emissions. It will also contribute to achieving peak carbon dioxide emissions and promoting carbon neutrality in China. [Methods] Based on multidimensional data disclosed by over 130000 fuel vehicle owners in China from 2014 to 2023, this article utilizes methods such as linear regression, decision trees, and neural networks to propose a calculation model and prediction model for the average cumulative travel carbon emissions of vehicle groups with the same number of days traveled as well as individual vehicles. Subsequently, a calculation model for annual average travel carbon emissions of vehicles is constructed. [Results] (1) In terms of predicting the cumulative travel carbon emissions of vehicles, compared to the vehicle cumulative travel carbon emission prediction model based solely on vehicle age, the prediction accuracy (R2) of the model with 25 variables in five categories (vehicle type, manufacturer, region, city level, price range) and vehicle age increased from 0.666 to 0.821. Age was found to have the highest impact on predicting cumulative travel carbon emissions for vehicles. (2) In terms of predicting the average cumulative carbon emissions from vehicles in a group with the same number of days traveled, the results based on neural network methods show that the prediction accuracy for the average cumulative travel carbon emissions at a national level reaches 0.915. The prediction accuracy for average cumulative travel carbon emissions for vehicles in five categories including vehicle type, manufacturer, region, city level, and price range ranges from 0.875 to 0.925. (3) In terms of calculating the average annual carbon emissions from vehicle travel, the national level has an average of 2.99 t. Vehicles in three categories—imported manufacturers, prices between (500, 1000] thousand CNY, and prices over 1000 thousand CNY—have higher average annual carbon emissions exceeding 4 t. Vehicles in six categories—MPV and SUV models, South China region, super large cities, price range between (200, 300] thousand CNY and price range between (300, 500] thousand CNY—have average annual carbon emissions ranging from 3.00 to 3.99 t. The remaining fifteen subcategories have lower average annual carbon emissions than the national level's average level and fall within the range of 2.15 to 2.99 t. [Conclusion] The large-scale vehicle owner data of multiple dimensions such as vehicle age, vehicle type, manufacturer, region, city level, and price range, along with the neural network model, contribute to improving the scientific and comprehensive measurement of carbon emissions from private car usage cycles. The calculation of road traffic carbon emissions in different regions should take into account the aforementioned differences in features. This study provides a reference for calculating carbon emissions from new energy private car travel. |
