基于VMD混合多尺度机器学习模型的碳排放权价格预测

Carbon Emission Price Prediction Based on VMD Hybrid Multi-Scale Machine Learning Model

构造了一种融合变分模态分解的多尺度混合碳价预测模型VMD-PSO-LSTM。结果显示,该模型能有效映射并拟合复杂多尺度的碳价时频信号,预测误差RMSE、MAE、MAPE仅为0.2109、0.176和0.0021,碳价预测精度和稳定性均优于基准模型。该模型的预测效果并不受随机样本预测期限差异的影响,并在较长随机区间的样本外预测上误差较小,展现出较强的预测鲁棒性和稳定性。

A multi-scale hybrid carbon price prediction model VMD-PSO-LSTM based on variational modal decomposition was constructed.The results show that the VMD-PSO-LSTM model could effectively map and fit complex multi-scale carbon price time-frequency signals,and the prediction errors RMSE,MAE and MAPE are only 0.2109,0.176 and 0.0021,and the accuracy and stability of carbon price prediction are better than those of the benchmark model.The prediction effect of the VMD-PSO-LSTM model is not affected by the difference of the prediction period of random samples,and the error in the out-of-sample prediction with a long random interval is small,showing strong prediction robustness and stability.