基于PCC-LSTM模型的全国碳排放权交易价格预测研究

Research on national carbon emission trading price prediction based on PCC-LSTM model

为了有效提升碳价的预测精度和保证全国碳交易市场的稳定运行,在利用皮尔逊相关系数法(PCC)提取关键碳价影响要素的基础上,分别运用ARIMA模型、多项式回归算法和LSTM模型,构建了全国碳交易市场的碳价预测模型。对比结果显示:相较于其它两类模型,PCC-LSTM模型的预测精度最高,平均精度达到99.26%,标准偏差最小,仅为0.2630,很好地显示了深度学习算法在样本数据充足条件下的优越性和实用性。相较于上半年的平均碳价,预测得到的11月碳价渐趋平稳,且有所提高,一定程度反映了全国碳交易市场整体运行平稳和年底交易活跃的特性。PCC-LSTM模型在碳价预测领域的成功应用,有助于企业熟悉碳市场的运行机制和变化规律,对扩大碳市场的覆盖范围和确保碳市场的稳定、健康发展具有重要的推动作用。

To improve the prediction accuracy of the carbon price and ensure the stable operation of the national carbon market,the Pearson correlation coefficient(PCC)method was first used to determine the key influence factors of carbon price;3 carbon price prediction models in the national carbon market were established based on 3 types of prediction approaches,named as ARIMA model,polynomial regression algorithm and LSTM model,respectively.The comparison results among 3 models demonstrated that,compared with the other 2 types of models,the PCC-LSTM model had the highest prediction accuracy,with an average accuracy of 99.26%,and the smallest standard deviation of 0.2630.It reflected the superiority and practicability of the deep learning algorithm under the condition of sufficient sample data.Compared with the average carbon price for the first half of 2022,the predicted carbon price in November would gradually stabilize and increase,which reflected overall stable operation and active trading characteristics at the end of the year of the national carbon trading market.The successful application of the PCC-LSTM model in carbon price prediction was capable of helping enterprises to familiarize themselves with the operation mechanism and variation rule of the carbon market,expanding the coverage range of the carbon market and ensuring its stable and healthy development.