A data-driven hierarchical dispatch framework for village renewable energy systems: Integration of user portrait cluster and cross-tier coordination based on ADMM

Village distributed energy systems are often constrained by low communication infrastructures and strong user heterogeneity, which hinder efficient coordination and reduce renewable self-consumption. Conventional centralized or synchronous optimization approaches either impose heavy communication burdens or neglect socioeconomic differences among users, which limits their applicability in village contexts. In order to address these challenges, this study develops a bi-level clustering and asynchronous ADMM framework. First, a Bayesian-based user portrait module classifies households according to socioeconomic attributes and demand flexibility. Then, a lower-level mixed-integer linear programming model optimizes intra-tier dispatch, while an upper-level asynchronous ADMM layer coordinates cross-tier energy exchange by exchanging only boundary power and multipliers. A linear cross-tier subsidy mechanism is further embedded to balance heterogeneous benefits. Case studies on a village system demonstrate that the proposed framework increases renewable self-consumption to 93.31 %, reduces PV and WT curtailment by 23.7 %, and lowers total costs from 2527.39 to 2011.60, which represents a 20.41 % improvement compared with baseline methods. The ADMM converges within 15 iterations under low communication, while the subsidy scheme ensures that low-consumption households (T5) achieve a 9.4 % cost reduction without undermining system efficiency. The results verify that the proposed asynchronous ADMM with cross-tier subsidies effectively balances economic efficiency, renewable utilization, and equity among heterogeneous users under low-communication conditions.