In the smart grid era, Short-Term Load Forecasting (STLF) is the building block of a secure, reliable, and economical power system. Therefore, researchers have spent much time trying different methods to improve load forecasting accuracy. Despite the advances in the STLF area, load forecasting is still difficult. This difficulty comes from two facts: 1- The behavior of the electric load is complex and shows different levels of seasonality; 2- The electric load is strongly influenced by other external factors such as meteorological variables and calendar variables. To overcome these issues, in this paper, a two-stage Kalman filter-based method is used to enhance the accuracy of STLF. In the first stage of the proposed method, the Kalman filter and Rauch-Tung-Striebel smoother are applied to the short windows of the past electric load series to obtain an initial prediction of the load series. To produce the final forecast, in the second stage, the initial prediction of the load series along with other calendar and meteorological variables are used to form a load forecasting model whose parameters are obtained based on another Kalman filter. The effectiveness of the proposed method is evaluated by performing a case study on the real dataset from a power utility in Iran, which shows the excellent performance of the proposed method with 1.98% mean absolute error.