摘要
当控排企业的配额交易以履约为驱动力时,碳市场会出现集中交易现象,并导致碳价格非线性、非平稳。针对这一问题,结合交易频率信息的相关特性,首先,采用迭代累积平方和算法分析碳市场的成交量,进而依碳配额交易频率对交易期进行划分;然后,借助小波变换(Wavelettransform,WT)提取碳价的市场发展趋势;最后,使用双向长短期记忆神经网络(Bi-directional long short-term memory,BiLSTM)对交易趋势进行预测。实验验证结果表明,若进行预测时能够考虑交易频率信息的影响,则能够提高模型预测精度;利用WT提取到的交易趋势信息进行预测,可使预测的效果优于直接对原序列进行预测;与长短期记忆模型相比,BiLSTM模型有更好的预测表现。
When the cap-and-trade of emission-controlled enterprises is driven by compliance,the carbon market will appear centralized trading phenomenon and lead to a nonlinear and non-stationary carbon price.In order to solve this problem,considering the relevant characteristics of trading frequency information,firstly,the iterative cumulative square sum(ICSS)algorithm is used to analyze the trading volume of carbon market,and then the trading period is divided according to the frequency of carbon allowances,wavelet transform(WT)is used to extract the market trend of carbon price;finally,BiLSTM is used to predict the trading trend.The experimental validation results show that the prediction accuracy of the model can be improved if the impact of trading frequency information can be taken into account when making predictions;the prediction can be better than the direct prediction of the original series by using the trading trend information extracted from WT;compared with the long and short-term memory model BiLSTM model has better prediction performance.
作者
李金颖
黄艺斌
LI Jinying;HUANG Yibin(Department of Economics and Management,North China Electric Power University,Baoding 071003,China)
出处
《电力科学与工程》
2023年第6期25-31,共7页
Electric Power Science and Engineering
基金
国家自然科学基金(72273042)。
