自适应结构优化灰色预测模型及其在新能源汽车产业演进趋势预测中的应用

New-energy automotive industry is the key content of China’s manufacturing strategy called “Made in China 2025”. In recent years, this industry is undergoing explosive growth owing to the supporting policies from central and local governments. However, China’s new-energy automotive industry is facing an increasing challenge for sustainable development, when a series of restrictions are coming, such as the “fall off” subsidy policies, constraints of battery techniques, and intensifying competition from petrol vehicles. Therefore, how to use the scientific methods to analyze and predict the evolution trend of China’s new-energy automotive industry and formulate effective policies during the transformation from policy-driven to policy-market-driven periods, is becoming an important problem waiting to be solved. To this end, much significant work has been conducted in this project. Firstly, based on the characteristics of China’s new-energy automobile industry, two novel kinds of self-adaptive buffer operators are proposed by incorporating the policy variable and considering the complex influences of other factors, so as to improve the capability of processing the original data. Secondly, three self-adaptive grey prediction models having optimized structure are designed by considering the driving effects of subsidy policies, nolinear influences of various factors, and accumulative effects of input variables. Thirdly, in order to solve the problems of parameter estimation and test, new methods for parameter estimation and test are put forward, which contribute to the applications of grey models. Finally, by using these above newly proposed models, the evolution trends of China’s new-energy automobile industry are predicted under different development scenarios of the dominated factors. In addition, several policies are formulated by using system simulation tools after removing the subsidy policies. These proposed policies can provide support for decision makers.

新能源汽车产业是我国制造强国战略的重点发展领域，近些年在政府支持下得到快速增长。然而，在补贴政策退坡、电池技术制约、燃油车竞争加剧等背景下，产业可持续发展面临挑战日益突出。因此，如何用科学的方法分析和预测新能源汽车产业由政策驱动到政策市场双轮驱动转型的演进趋势并形成政策工具组合，是迫切需要解决的科学问题。本项目立足于新能源汽车产业发展典型特征，构建含有政策变量和其他多因素交叉影响的自适应缓冲算子，提升对典型特征数据处理能力；针对补贴政策驱动、多因素非线性作用、时滞累积效应等数据特征构建自适应结构优化多变量灰色预测模型群；针对模型参数估计和检验问题，提出新型灰色模型的参数估计和参数可靠性检验方法，增强灰色模型的拓展能力；最后，利用上述前沿理论方法，结合主导因素的不同发展情景，研究我国新能源汽车产业演进趋势，并利用系统仿真工具提出后补贴时代政策工具组合，为促进产业可持续发展提供决策支持。

培育和发展新能源汽车产业符合绿色低碳发展方向，是推动经济高质量发展和实现“碳达峰碳中和”目标的重要战略举措。鉴于当前补贴政策退坡、国内外市场竞争激烈等大背景，如何准确把握和预测我国新能源汽车产业未来发展趋势是迫切需要解决的科学问题。本项目在自然科学基金的支持下，针对新能源汽车产业中的多种特征时变参量预测建模进行了系统深入的探索和研究。首先，针对基于各类复杂冲击扰动序列提出了面向动态数据特征的自适应数据预处理技术，涉及基于新信息的自适应缓冲算子构造、季节特征数据预处理技术等；然后，提出了面向多种时间序列特点的自适应结构优化灰色预测模型群，包括初始条件自适应优化的灰色预测模型、结构自适应优化的单变量和多变量灰色模型；随后，针对参数估计和检验问题，提出了自适应寻优算法的选择框架，以及基于蒙特卡洛模拟和概率密度分析的参数稳健性检验方法，增强灰色模型的拓展能力；最后，利用国内外新能源汽车发展的实际案例数据，验证了上述模型方法的实用性、可靠性、以及稳健性。