A C-DCGAN-based method for generating extreme risk scenarios of high percentage new energy systems

Extreme events have occurred frequently in recent years, and the extreme uncertainty of source -loads of high percentage new energy systems brings risks to the safe and stable operation of power systems. Therefore, this paper proposes a C-DCGAN-based extreme risk scenario generation method for high percentage new energy systems. Firstly, an improved deep convolutional conditional adversarial generation network C-DCGAN suitable for new energy output and load scenario generation is designed by combining the advantages and disadvantages of CGAN and DCGAN. secondly, considering the source-load risk features of the extreme scenarios of the system, the respective extreme measures are established from the three risk perspectives reflecting the threats of wind and power abandonment, power supply tension, security and stability, which are used as the extra conditions for generating samples in line with the extreme scenario risk characteristics. Then, in order to improve the model training stability and reduce the metric error, the Wasserstein distance is adopted and the extreme loss term is added as the loss function of the adversarial generation network. Finally, training the historical data of a region's scenery in Python, and the source-load risk scenario close to the real scenario is generated using the power supply tension as the input of the extreme condition, thus verifying the stability of the model training and the validity of the method proposed in this paper.