Bi-level Collaborative Access Optimization of Demand-Side Resources Based on External-Interactive Characteristics of Users

With the extensive integration of user-side resources such as distributed new energy, electric vehicle charging piles, and energy storage, the volatility and randomness of regional distribution networks have intensified. This makes it challenging to accurately characterize the external-interactive characteristics of users and grid, posing significant challenges for the optimization of source-load-storage access under various planning scenarios. To address this issue, a bi-level collaborative access optimization method of demand-side resources based on external-interactive characteristics of users is proposed. Firstly, the variational Bayesian Gaussian Mixture Model (VBGMM) clustering algorithm is employed to construct a set of typical daily scenarios. Secondly, external-interactive characteristics indicators are defined, and a multidimensional stereoscopic profile of incremental load is constructed based on the principles of correlation-based feature selection and best-first search strategy. Finally, based on the analysis of the load profile of users, the load adjustment potential and response capabilities are determined. These serve as the basis for model optimization. A bi-level collaborative programming model for multivariate demand-side source-load-storage access is established and linearized and convex-relaxed using techniques such as second-order cone relaxation, the Big-M method, and McCormick linearization. The model is then transformed into a multi-objective mixed-integer programming model and solved using the Improved Normalized Normal Constraint (INNC) method. Simulation results demonstrate that the proposed algorithm exhibits superior performance in enhancing the bearing capacity and economic efficiency of regional distribution networks, promoting peak shaving and valley filling, and maintaining voltage stability.