As the global climate crisis deepens, countries are accelerating the transition to low-carbon and sustainable energy systems. In this context, generating high-quality renewable energy scenarios is crucial for robust power system planning and operation. This study proposes a novel temporal scenario generation method based on the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP). By introducing a gradient penalty to enforce the 1-Lipschitz constraint, the model improves training stability and mitigates issues such as mode collapse, which are common in conventional GANs. To assess the fidelity of generated time-series data, we introduce a new metric called Spectral Entropy Deviation (SED), which quantifies frequency-domain complexity and evaluates dynamic consistency. Experimental results on photovoltaic and wind power datasets demonstrate that the proposed WGAN-GP model outperforms Variational Autoencoders (VAE) in capturing statistical distributions, temporal dependencies, and dynamic behaviors of renewable energy outputs. These findings highlight the model’s potential to serve as a reliable tool for uncertainty modeling in renewable integration and energy transition planning.