碳排放是一个受多因素交互作用的动态过程,准确预测碳排放量有利于碳减排措施的制定。由于碳排放本身模型具有动态变化性、非线性、社会性等特点,传统预测方法不能满足实际情况的需要。人工神经网络模型能够较好地描述碳排放时间系列数据的非线性特性,被广泛应用于预测国家、区域、行业等层面的碳排放量变化。其中,误差反向传播(Back Propagation,BP)神经网络模型和长短期记忆(Long Short-Term Memory,LSTM)神经网络模型备受关注。在模型预测过程中,通过识别目标模型的碳排放影响因素类型、提高输入层数据的准确性、构建适宜的线性—非线性耦合的组合模型等途径,进一步提高模型预测的准确性。研究结果对人工神经网络模型在碳排放预测中的应用情况进行梳理,为碳排放预测技术的进一步发展提供参考。

Carbon emission is a dynamic process influenced by various factors and accurately forecasting these emissions is conducive in developing reduction strategies. Traditional forecasting methods often fall short of actual situations due to the dynamic, nonlinear, and social characteristic of carbon emissions. The artificial neural network model, capable of capturing the nonlinear patterns in time-series data, is widely used to predict changes in carbon emissions at national, regional, and industrial levels. Among them, BP (Back Propagation) neural network model and the LSTM (Long Short-Term Memory) neural network model are particularly favored by researchers for carbon emission forecast. The prediction accuracy of these models can be enhanced by systematically categorizing the types of factors influencing carbon emissions, enhancing the accuracy of input data, and developing appropriate models that couple linear and nonlinear components. The research reviews the application of artificial neural network models in carbon emission forecast, offering guidance for the future development of carbon emission forecast technologies.