基于条件生成对抗网络与迁移学习的暂态电压稳定超前判别

Lead discrimination of transient voltage stability based on conditional generation adversarial network and transfer learning

为解决样本不平衡导致的暂态电压稳定判别准确性不足的问题以及实现暂态电压稳定超前判别,提出一种基于条件生成对抗网络(CGAN)与迁移学习的暂态电压稳定超前判别方法。考虑暂态电压稳定样本类型,利用CGAN定向扩增暂态电压样本集,解决样本不平衡问题,从而提升暂态电压稳定判别准确性;考虑到CGAN生成器与暂态电压时序预测模型具有相似的学习任务,将CGAN生成器模型迁移至暂态电压时序预测模型,结合工程判据实现暂态电压稳定超前判别,并进一步提升暂态电压稳定判别准确性。在CEPRI-VC暂态电压稳定分析系统中验证了所提方法的有效性。

In order to solve the problem of insufficient accuracy in transient voltage stability discrimination caused by sample imbalance and achieve the lead discrimination of transient voltage stability,a lead dis⁃crimination method of transient voltage stability based on conditional generative adversarial network(CGAN)and transfer learning is proposed.Considering the types of transient voltage stability samples,CGAN is used to directionally expand the transient voltage sample set for solving the problem of sample imbalance,thereby improving the accuracy of transient voltage stability discrimination.Considering that the CGAN generator and transient voltage time series prediction model have similar learning tasks,the CGAN generator model is transferred to the transient voltage time series prediction model,the lead discrimination of transient voltage stability is realized combining with the engineering criteria,and the accuracy of transient voltage stability discrimination is further improved.The effectiveness of the proposed method is verified in CEPRI-VC transient voltage stability analysis system.