A new method incorporating the time sequence modeling and intelligent algorithm modeling is presented to forecast the day-ahead electricity price. With genetic algorithm（GA） adopted to optimize the model's parameters， support vector machines（SVM） model is applied to forecast the price sequence. The generalized autoregressive conditional heteroscedasticity（GARCH） models are applied to adjust the error series of price forecasted by SVM-GA models， eliminating their autocorrelations and heteroscedasticity effects. A case of forecasting the day-ahead price of PJM market in August， 2005 demonstrates the proposed method has a desirable performance with an overall mean absolute percentage error（MAPE） of 8.19 percent， which is nearly 4 percent less than data forecasted by common methods.