Motivated by the development of photovoltaic (PV)-enhanced power engineering, this study presents a structured review of year-round PV scenario generation and power forecasting methods across multiple timescales, including day-ahead and intraday horizons, from the perspective of statistical relational artificial intelligence (SR-AI). A unified analysis framework is established from the viewpoint of representation, inference, and constraint to elucidate the relationship between data-driven modeling and system-level decision-making. At the annual timescale, the evolution of PV scenario generation is reviewed from probabilistic models to deep generative models. At the day-ahead timescale, high-dimensional feature representation, uncertainty modeling, and robust learning methods are summarized. At the intraday timescale, approaches for fast fluctuation modeling, multimodal data fusion, and spatial correlation modeling are discussed. Finally, key challenges, including extreme-event characterization, forecasting robustness under changing weather conditions, and the practical deployment of scenario-generation and forecasting methods, are highlighted. This article provides a comprehensive reference for AI-enhanced systems engineering in PV energy systems.