As the existing deterministic short-term load forecasting methods hardly meet the demands of uncertain risk analysis and decision-making in electricity market， a probabilistic load forecasting method based on the statistics of load forecasting errors’ characteristic is presented at length. First， a statistic analysis model for the distribution regularities of forecasting error is established in two dimensions. The principle and method to verify the validity of statistical regularity is then proposed. At last， by combining the verified distribution regularity and the deterministic load forecasting result， the probability distribution of load forecasting can be gained. According to the result， envelopes of load forecasting curve under certain confidence level can also be obtained. The validity and practicability of the proposed method are tested with the actual data. It is expected that the proposed approach can provide a new feasible solution for the probabilistic short-term load forecasting.