Wind turbines are characterised by complex structures, difficult operation and maintenance, and long-term exposure to harsh operating environments. The demand prediction of wind turbine spare parts helps to equip wind farms with the most suitable number of spare parts to ensure the smooth and efficient operation of wind farms. Principal component analysis-back propagation (PCA-BP) model was constructed for the complex spare parts affected by multiple factors, and the elements affecting the wind power spare parts were first screened by PCA, and then the BP neural network algorithm was used to obtain the most accurate prediction results. The results of autoregressive integrated moving average (ARIMA) model, BP neural network prediction and PCA-BP neural network prediction were compared. It shows that PCA can significantly reduce the neural network prediction error, and the accuracy of the prediction is 93.94%, which is higher than the 88.39% of the BP neural network prediction and the 85.31% of the ARIMA model, so the PCA-BP neural network model has accurate prediction accuracy and reliable results, and it can be applied to the wind turbine spare parts demand prediction.