【摘要】 阐述为提高天然气负荷数据异常检测的准确性，基于Transformer构建生成对抗网络（GAN）的生成器和判别器，并引入改进的Wasserstein损失函数来解决GAN在训练过程中的问题，构建基于Transformer-GAN的天然气负荷数据异常检测模型。实验结果和对比分析说明Transformer的使用能够提高异常检测的效率，提出的异常检测模型具有更高的异常检测精度和良好的可解释性。更多还原

【Abstract】 This paper describes the construction of a Generative Adversarial Network(GAN)generator and discriminator based on Transformer to improve the accuracy of anomaly detection in natural gas load data. An improved Wasserstein loss function is introduced to solve the problem of GAN in the training process, and a natural gas load data anomaly detection model based on Transformer GAN is constructed. The experimental results and comparative analysis indicate that the use of Transformer can improve the efficiency of anomaly detection, and the proposed anomaly detection model has higher anomaly detection accuracy and good interpretability.更多还原