This paper takes the thermal coal at Qinhuangdao Port as the research object, collects the daily price time series data of the thermal coal, performs stationarity test and first-order difference on the series, obtains the stable daily thermal coal price time series after the difference. Autocorrelation graph and partial autocorrelation graph are distinguished. By introducing the arithmetic average of AIC and BIC as the main basis for determining the model parameters, the ARMA(7,4) model is finally determined to fit the sequence. Secondly, in view of the "volatility aggregation" in the time series diagram, the LM test is used to determine that the residual square is not a white noise sequence, and it is concluded that the sequence has a GRACH effect. Finally, we revised the ARMA model based on the GRACH model, determined that GARCH(2,2)-ARMA(7,4) was the model with the highest degree of fit, and used this model to extrapolate the fitting results, and found that the price of thermal coal was After that, the volatility will gradually become smaller, prices will stabilize, and market supply and demand will gradually return to balance.