Exponential smoothing model is one of the main load forecasting models of power systems, the accuracy of the model depends on smoothing coefficient. In this paper, a study on how to seek the best smoothing coefficient is given. The results show that using the principle of weighting more on near data and weighting less on far data to seek the optimal smoothing coefficient can get a better result. Based on the result, a method of how to weight more on near data and weight less on far data is proposed. A practical example of load forecasting with the method is also shown here.