Short-Term Energy Consumption Prediction of High-Speed Train Production Based on Small Sample Data Repair and Augmentation

High-speed trains are an important support for national development, and their production process requires a significant amount of resources. Effectively achieving low carbon emissions and reducing energy waste in the production process is an important problem to be addressed. However, at present, the energy consumption assessment of the workshop is based on the readings from electric meters only, and the effective data of the workshop is small with abnormal jumps in the data. Based on this, we propose a method for short-term energy consumption prediction in high-speed train production based on small sample data repair and augmentation. In the first step, the preprocessed data undergoes repair for abnormal values using the Long Short-Term Memory Network (LSTM) technique, while the Fourier transform and polynomial fitting techniques are applied to extract the periodicity and trend characteristics of the data. Subsequently, the Deep Convolutional Wasserstein Gener-ative Adversarial Network (DCWGAN) method is employed to augment the data, effectively addressing the challenge of a small sample size. Finally, energy consumption prediction is performed using a convolutional neural network incorporating an attention mechanism and a bidirectional gated recurrent neural network. To verify the effectiveness of the method, experiments are conducted using energy consumption data and production data from the workshop of a high-speed train manufacturing company. The experimental results demonstrate that the proposed method achieves higher accuracy compared to commonly used deep learning methods.