Electricity quantity prediction is an important basis for power supply units to purchase electricity, and it is the fundamen‐ tal foundation for the stable operation of the electricity economy. This article proposes using the SARIMA-ARCH fusion model to predict electricity consumption. Firstly, the seasonal time series model (SARIMA) is used to model the monthly electricity consump‐ tion time series. The optimal model order is determined through ACF and PACF graph screening, and the basic prediction model of the seasonal time series model (SARIMA) is obtained. Subsequently, the regression residuals of the basic model are subjected to ARCH effect testing, and an autoregressive conditional heteroscedasticity (ARCH) model is established. Finally, this article com‐ pares and analyzes the predicted values of the SARIMA-ARCH model with those of conventional SRIAM and ARIMA. The results show that the prediction accuracy of the hybrid model of SARIMA-ARCH is high.