Electricity price forecasting error presents a dual random process. One is the forecasting ability state of the model， showing the model's forecasting accuracy on the ideal data without stochastic influence; the other is the forecasting error of real price series including the stochastic validities. This paper adapts hidden Markov model to analyze the forecasting error and to find the regularity of forecasting ability state of the model and the error distribution of the model in different states. According to the above dual random process analysis， the forecasting error in next step is analyzed and forecasted to adjust the model's forecasting results. A case study of a USA electricity market shows that the proposed method improves the forecasting results significantly.