Developing a Fine-Grained Modeling System for Virtual Power Plant User-Side Loads via PSO-AdaBoost Ensemble Strategy

Accurate user-side load modeling is fundamental for virtual power plants (VPPs) to achieve coordinated scheduling and intelligent energy management, where reliable communication infrastructures and distributed information exchange play a critical role in supporting large-scale resource aggregation. To address the limitations of conventional load modeling methods in handling multi-factor coupling, heterogeneous load characteristics, and dynamic operating conditions, this study proposes a fine-grained modeling framework based on a Particle Swarm Optimization (PSO)- AdaBoost ensemble strategy. A multidimensional feature engineering scheme is established by integrating historical load patterns, meteorological variables, user behavior, and production process information, while PSO is employed to optimize weak classifier weights and adaptive iteration parameters within the AdaBoost framework. The proposed method enables differentiated modeling for industrial, commercial, and residential loads and improves robustness under dynamic operating scenarios. Furthermore, the framework is compatible with communication-enabled VPP architectures that rely on real-time sensing and distributed data transmission, providing efficient support for coordinated energy dispatch. Experimental validation demonstrates enhanced modeling accuracy, adaptability, and computational efficiency compared with conventional approaches, offering a practical solution for intelligent load characterization in advanced energy systems and communication-assisted smart grid applications.