This paper proposes a method to predict net power demand of a few hours ahead using fixed information at the time of prediction, result value of net power demand, air temperature, and solar radiation. Because the result value is obtained easily, the proposed method updates the prediction value at short intervals using latest result value. We evaluate the performance of the proposed method using the demand data of an actual power company. The prediction error of the proposed method is 0.97% when we predict net power demand of 90 minutes ahead.