To further explore the influence factors of load time sequence forecasting errors and improve the accuracy of load forecasting, an approximate entropy algorithm is proposed for characterizing the regularity of load time series quantitatively and recognizing the causes for load forecasting errors fully. Approximate entropy algorithm is applied to analyze load time series and determine the regularity of it. Based on the above research, this paper carries out the related research according to the relationship between the regularity and forecasting errors of load time series. The results show that, approximate entropy algorithm can describe the regularity of load time series effectively, and there is a strong correlation between regularity of load time series and load forecasting errors, proving the correctness and validity of the method. This work is supported by National Key Basic Research Program (973 Program) (No. 2013CB228201) and National Natural Science Foundation of China (No. 51307017).