基于CNN-LSTM组合模型的碳价预测方法

Carbon Price Prediction Method by CNN-LSTM Combination Model

摘要:

随着全国碳排放权交易市场的正式启动，准确的碳价预测有助于市场管理机构实现对碳价的有效调控及控排企业的高效履约。本文首先对碳价波动的特征进行分析，然后基于卷积神经网络（Convolutional Neural Network, CNN）与长短期记忆网络（Long Short-Term Memory, LSTM），提出了一种基于CNN-LSTM组合模型的碳价预测方法，能够充分考虑碳价的时序特性，有效改善传统模型难以从时序数据中提取有效特征的问题。最后，本文以欧洲能源交易所及我国广州碳市场的碳价实例验证，将本文所提方法与其他常用预测模型进行对比，结果表明本文所提出的碳价预测方法在国内外碳价预测中均具有更高的预测准确性，为后续碳价预测选取研究模型提供了一定的参考。

Abstract:

With the official launch of the national carbon emission trading market, accurate carbon price forecasting will help market management institutions to achieve effective regulation of carbon prices and efficient performance of emission control enterprises. Then, based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM), this paper proposes a carbon price prediction model based on the combined CNN-LSTM method, which can fully consider the timing characteristics of carbon prices and effectively improve the problem that traditional models cannot extract valid features from time series data. Finally, the carbon price examples of the European Energy Exchange and the carbon market in Guangzhou, China are carried out, and the proposed method compared with other common prediction model, and the results show that the carbon price prediction method proposed in this paper has higher prediction accuracy for carbon price prediction at home and abroad. It provides a certain reference for the future research model selection of carbon price prediction.