In order to accurately predict short-term wind power

a new short-term wind power prediction method is proposed in this paper. Firstly

the Pearson correlation coefficient method was used to analyze the correlation between wind speed

wind direction and other influencing factors and wind power series. Secondly

the convolution neural network（CNN） was used to extract the features of the input sequence. Then

a forgetting gate and an input gate were added to the long short-term memory（LSTM） network to form an LSTM network with multi-level gating. Combined with CNN

an improved CNN-LSTM short-term wind power prediction model which can improve the feature extraction ability and prediction accuracy of input sequence was established. Finally

the simulation analysis was performed with the measured data of a wind farm in Gansu Province of China

and the prediction results were used as the basis for making the scheduling plan to analyze the influence of the test results on the operation cost of the system. The simulation results show that

compared with the LSTM model and the CNN-LSTM model

the root mean square error of the results predicted by this model is reduced by 63.9% and 47.9%

respectively

and the mean absolute error is reduced by 70.4% and 53.5%

respectively

which can improve the accuracy of wind power prediction to a certain extent. The wind power prediction results can effectively reduce the reserved spinning reserve capacity of the system and reduce the operation cost of the system

and provide a strong basis for the formulation of dispatching plan.