Abstract: The price changes in the carbon trading market, as the core indicator for measuring the market operation, provide decision-making basis and reference for the regulators and investors of the carbon emission rights trading market. Affected by the nonlinear, high noise and strong fluctuation characteristics of carbon prices, the prediction research of carbon emission rights prices faces considerable challenges. Based on the daily average transaction prices of the Shanghai carbon market from 2013 to 2023, this article constructs the ARIMA-LSTM combined model. Firstly, the ARIMA model is used to predict the data to capture the linear and cyclical trends of carbon prices. Secondly, the residuals generated by the linear prediction of carbon prices by the ARIMA model are taken as the input of the LSTM model to learn the nonlinear parts in the sequence. Finally, the results of the two models are weighted and combined to obtain the final prediction data. The experimental results show that the ARIMA-LSTM combined model is superior to the single ARIMA or LSTM models in both the prediction accuracy of carbon prices and the interpretability of the results, and can better capture the fluctuation trend of carbon market prices. This study would provide assistance for market participants to understand the internal mechanism of carbon emission rights pricing, and be conducive to promoting the sustainable development of China's carbon market.