管理工程学报2024，Vol.38Issue(3)：186-201,16.DOI:10.13587/j.cnki.jieem.2024.03.014
混频数据驱动的电子废弃物生成量时空演化预测
Prediction of the spatiotemporal evolution of electronic waste generation driven by mixed frequency data
摘要
Abstract
E-waste is one of the fastest growing and most hazardous waste streams in the world,accounting for approximately 5%of total solid waste.China is one of the world's largest producers and consumers of electrical and electronic products,and China's e-waste generation is expected to grow to 27.2 million tons by 2030.To strengthen the management of key e-waste,China published the"Waste Electrical and Electronic Equipment Disposal Catalogue(First Batch)"in 2010,which included five products:TV sets,refrigerators,washing machines,room air conditioners and microcomputers.In 2015,China expanded the product range from the initial five to 14,including electrical and electronic products such as range hoods,electric water heaters,gas water heaters and mobile phones.Considering that the estimation and prediction of e-waste generation is the basis on which the government to formulates circular economy development policies and enterprises and determines production capacity,many studies have focused on the original four targeted products and mobile phone,for which data are easily available and can be discussed across regions;However,e-waste investigations,and therefore data,regarding the 14 new categories are lacking.Hence,the spatial and temporal evolution patterns of e-waste in the 14 new categories have not been explored. In this study,we construct a framework for predicting the spatiotemporal evolution of e-waste generation driven by mixed-frequency data,based on the modeling idea of integrated methodology,to address the problem that e-waste generation prediction is easily affected by multiple factors such as seasonality,sudden change and regionality. On the one hand,GM(1,1),support vector regression,Holt-Winters and X12 models are integrated to predict annual sales values of electrical and electronic products at low frequencies and quarterly sales share values at high frequencies,and the predicted quarterly sales values of products are obtained by multiplicative operations.It was found that among the forecasting models,the mixed-frequency forecasting model could obtain higher forecasting accuracy than the classical models(Holt-Winters,gray waveform forecasting and X12 model)and significantly improve the short-term forecasting.The MAPE of the MF-X12 model for the gas water heater test set data was reduced by 33.4184%compared with the GWM model,which had the best performance among the classical models,and the MAPE of the solar water heater test set data was reduced by 24.9437%and 26.5015%,respectively,compared to the best performing HWA and HWM models in the classical model. On the other hand,the life span of specific electrical and electronic products is investigated and fitted by means of the questionnaire method or expert consultation method,and the overall generation of e-waste nationwide is estimated by using the classical market supply model.On this basis,considering the average product ownership per 100 households and the total number of households and product structure data in each place,a regional quota estimation model is constructed to map the national total amount of e-waste to each region in each place,which provides a basis for the analysis of the spatial and temporal evolution pattern of e-waste.It is found that the generation of different categories of e-waste(gas water heaters and mobile phones)shows different spatiotemporal evolution characteristics,with gas water heaters showing a"south to north"trend and mobile phones showing a"north to south"trend;these trends are related to product attributes and their influencing sales factors. To conclude,Chinese e-waste dismantling enterprises apply for subsidies on a quarterly basis from the national WEEE disposal fund based on product types and quantities;this study provides a more accurate estimation and prediction of the quarterly generation of e-waste,that accounts for the seasonal and sudden change data characteristics of the time series.This study can provide data and decision support for the state to better plan the collection and use of funds,and for enterprises to better optimize production lines and dismantling capacity.At the same time,due to the uneven economic and population development in China,there are significant differences in the regional distribution of different electrical and electronic products.This study proposes a regional quota estimation model to address the imperfect statistics and publication systems of regional sales of electrical and electronic products,thereby estimating e-waste generation and its evolution characteristics for different product categories and regions.This data can help the country,from a national perspective,layout regional e-waste dismantling centers for these products.关键词
混频数据/电子废弃物/季节性预测/时空演化Key words
Mixed frequency data/E-waste/Seasonal prediction/Temporal and spatial evolution分类
管理科学引用本文复制引用
王方,余乐安,何昌华,刘启明..混频数据驱动的电子废弃物生成量时空演化预测[J].管理工程学报,2024,38(3):186-201,16.基金项目
国家自然科学基金项目(72001165) （72001165）
陕西省创新能力支撑计划(2022SR5016) （2022SR5016）
陕西省哲学社会科学重大理论与现实问题研究项目(2022ND0389) The National Natural Science Foundation of China(72001165) （2022ND0389）
The Shaanxi Province Innovation Capacity Support Program(2022SR5016) （2022SR5016）
The Major Theoretical and Practical Research Project of Philosophy and Social Sciences in Shaanxi Province(2022ND0389) （2022ND0389）
