Research on building flexible load quantification method based on users’ behavior prediction and multi-timescale optimized operation method for integrated energy system

Coordinated dispatch of flexible supply and demand resources mitigates flexibility limitations in integrated energy systems (IES). However, existing scheduling strategies often overlook users' inherent variability and behavioral uncertainty when quantifying flexible loads. Furthermore, they predominantly rely on single-timescale optimization, resulting in sub-optimal overall energy utilization efficiency. To address these limitations, this study proposes a building flexible load quantification method based on user behavior prediction, integrated into a multi-timescale (day-ahead, intra-day, and real-time) optimal supply-demand operation framework. First, a probabilistic prediction model for user energy consumption behaviors is developed utilizing a non-homogeneous Markov Chain combined with the CART decision tree method. Subsequently, a flexible load quantification model is constructed to accurately calculate users' shiftable, transferable, and reducible load capacities based on the predicted consumption patterns. This comprehensive framework aims to minimize overall IES operating costs by actively incorporating these flexible load characteristics. Validated against the UK Time Use Survey database, the behavior prediction method achieves accuracies exceeding 90% for most behaviors and 72% for the remainder. Case studies simulating a typical winter day demonstrate that the proposed scheduling reduces the electric load peak-valley difference by 9.7%. Moreover, the multi-timescale operation increases the system's PEE and PREU by 3.94% and 8.79%, respectively, while decreasing PEC by 10.5% and operating costs by 10.29%. Ultimately, this cooperative optimization approach effectively reduces costs, achieves peak-shaving and valley-filling, and enhances renewable energy consumption and the comprehensive benefits of the IES.