Short-term load forecasting methods for the bus charging station are studied prior to proposing a combined forecasting model based on data freshness and cross entropy. First

the load characteristics are analyzed to show the daily charging load has the features of large fluctuation

periodicity

and being closely related to meteorological conditions(including temperature and rainfall). Secondly

in the accumulation process of historical prediction errors

the combined forecasting model is improved in the following aspects: (1)Considering the time characteristics and fluctuating characteristics of the charging load sample data

the selecting method of similar days based on grey relational degree is proposed; (2)Considering the precision and stability of a single method

a combined forecasting model based on cross entropy and normal distribution probability density function is developed to dynamically adjust the weight coefficients; (3)Considering the time effectiveness of the data source

the concept of freshness function is put forward

which improves the probability density distribution function of the single forecasting method to further optimize the weight coefficients of the combined forecasting model

improving the accuracy of the model. Finally

the training samples and test samples based on the historical data of a Beijing bus charging station are developed. Compared with single models and other combined forecasting methods

the validity of the combined forecasting model proposed is proved. This work is supported by National Key Technologies R&D Program(No. 2013BAA02B02).