Carbon Trading Price Prediction Method Based on the VMD-CNN-LSTM Model

Abstract

Accurately predicting carbon trading prices provides solid data support for the production and operation decisions of enterprises and the investment decisions of investors. This makes it easier to achieve the two goals of peaking carbon and reaching carbon neutrality. For predicting carbon trading price, this paper is established on the VMD-CNN-LSTM model; we introduce VMD method in combination with VMD-CNN-LSTM, thereby auto-separating frequencies through IMF and then predicting IMFs themselves separately. Datasets used for testing: the time-series data of 2701 carbon-trading prices in Shenzhen were selected. The verification tells us that the VMD-CNN-LSTM model is effective in forecasting Shenzhen carbon trading prices. VMD decomposition of sub-modes can be applied to deal with high-noise, non-stationary, and nonlinear carbon emission trading data sequences, thereby correcting the model prediction accuracy.