基于改进条件生成对抗网络的可控场景生成方法

可再生能源发电具有较强的随机性和波动性，为实现高效可靠的场景建模，提出一种基于改进条件生成对抗网络的可控场景生成方法。提出基于条件生成对抗网络的场景生成框架，结合Transformer的全局注意力机制以及常规卷积和深度可分离卷积的局部泛化机制，设计适用于提取可再生能源发电不同维度特征的网络结构；利用条件生成对抗网络模型建立低维气象特征隐空间和高维可再生能源发电数据之间的映射关系，提出一种可控场景生成方法，并建立随机场景生成、场景约减、极端场景生成和连续日场景生成4种生成策略。基于实际光伏、风电数据和气象数据的仿真结果表明，所提模型与方法能够有效学习可再生能源发电的随机性、时序性、波动性及空间相关性，实现对不同策略下场景的可控生成。

Controllable scenario generation method based on improved conditional generative adversarial network

The renewable energy generation has strong randomness and volatility, in order to achieve efficient and reliable scenario modelling, a controllable scenario generation method based on an improved conditional generative adversarial network is proposed. A scenario generation framework based on the conditional generative adversarial network is proposed. Combining the global attention mechanism of Transformer and the local generalization mechanism of conventional convolution and depth-separable convolution, a network structure suitable for exacting different dimensional features of the renewable energy generation is designed. The conditional generative adversarial network model is used to establish the mapping relationship between the low-dimensional meteorological feature latent space and the high-dimensional renewable energy generation data, a generation method of the controllable scenario is proposed, and four generation strategies of random scenario generation, scenario reduction, extreme scenario generation and continuous daily scenario generation are established. The simulative results based on the actual photovoltaic, wind power data and meteorological data show that the proposed model and method can effectively learn the randomness, timing, fluctuation and spatial correlation of renewable energy generation, and realize the controllable generation of scenarios under different strategies.