Abstract With the continuous development of intelligent power system, high-precision short-term load forecasting becomes one of the important issues on the research of power system operation. Based on the introduction of the principle of Kalman filter, the Kalman filter step recursive equations are given. Combined power system load conditions, the Kalman filter short-time forecasting model of is established, follows its predictions. By introducing the holiday factor, the advanced algorithm is put forward. The prediction accuracy of the Kalman filter is improved, confirming the validity and effectiveness of the advanced algorithm.