摘要
为更好地挖掘大量采集数据中蕴含的有效信息,提高碳交易量预测精度,文中提出一种基于小波变换和卷积神经网络(CNN)、门控循环单元(GRU)模型的碳交易量组合预测方法。首先利用小波变换算法对上海市历史碳交易量数据进行去噪处理,接着将去噪后数据序列输入到CNN-GRU模型中进行预测,最终将预测结果与使用原始数据的GRU模型和CNN-GRU模型的预测结果进行对比,结果表明使用小波变换和CNN-GRU的组合预测方法可有效提高碳交易量预测的精度。
In order to better exploit the effective information contained in the large amount of collected data and improve the accuracy of carbon trading volume prediction, a combined carbon trading volume prediction method based on wavelet transform and convolutional neural network (CNN), gated re-current unit (GRU) model is proposed in this paper. Firstly, the wavelet transform algorithm was used to denoise the historical carbon trading volume data in Shanghai, and then the denoised data series were input into the CNN-GRU model for measurement. Finally, the prediction results were compared with those of the GRU model and CNN-GRU model using the original data, and the results showed that the combined prediction method using wavelet transform and CNN-GRU could effec-tively improve the accuracy of carbon trading volume prediction.
出处
《建模与仿真》
2023年第3期2040-2048,共9页
Modeling and Simulation