The increasing penetration of renewable energy in global clean energy transitions introduces significant challenges to planning, operation, and optimal dispatch due to intermittency, uncertainty in generation, and dynamic load fluctuations. To strategically mitigate these risks, scenario generation techniques have emerged, constructing plausible future scenarios to support data-informed decision-making in power systems. This paper comprehensively reviews mainstream methods for wind, solar, and load scenario generation, which are widely applied in Monte Carlo simulation (MCS), Latin hypercube sampling (LHS), Copula functions (CF), generative adversarial networks (GANs), variational autoencoders (VAEs), physics-informed data-driven fusion approaches (PDF), Markov chain Monte Carlo (MCMC) methods, and time-series analysis techniques (TAT). Through analyzing the principles, characteristics, technical merits, and application cases of these methodologies, we summarize recent research advances and propose future directions. Findings demonstrate that scenario generation significantly enhances power systems’ resilience against uncertainties, addressing the growing demand for robust operational frameworks, while confronting persistent challenges in multi-source data fusion, high-dimensional uncertainty modeling, computational tractability, and real-time dynamic adaptability in practical implementations.