Short-term prediction plays an important role in the construction of smart grid

and profoundly affects the intelligent transformation of all aspects of power grid generation

transmission and distribution. Short-term predictions are generally based on system measured data

while sensor failures

data transmission errors and other reasons will lead to a decline in data quality

which will seriously affect the accuracy of short-term predictions. In order to establish an accurate short-term prediction model under the condition of data quality impairment

this paper proposes a short-term prediction framework bi-directional long short-term memory data preprocessing (Bi-LSTM-DP) that combines data preprocessing and bi-directional long short-term memory (Bi-LSTM). In Bi-LSTM-DP

the acquired data first fills in the missing values by mean

then denoises the data based on the Savitzky-Golay filter

and finally uses Bi-LSTM to extract time series information to achieve short-term predictions. In order to evaluate the performance of the proposed method

the measured public data sets are used to predict the wind power generation capacity and load demand respectively

and the comparison with other reference methods shows the effectiveness and robustness of the proposed method.