A Stability-Enhanced NSGA-TS Method for Online Three-Phase Unbalance Governance in Distribution Networks

To address the demand for online dynamic governance of three-phase unbalance in low-voltage distribution networks, this paper proposes a control strategy for Load Phase Switching (LPCS) that focuses on decision stability and optimization efficiency. The key innovation is a deeply integrated hybrid intelligent algorithm, the Non-dominated Sorting Genetic Algorithm with Tabu Search (NSGA-TS). The algorithm’s originality is twofold: 1) It introduces a “Gene Library” mechanism, which uses statistical information from elite solutions to probabilistically guide the crossover operator, accelerating the propagation and convergence of superior patterns. 2) It designs a dynamic switching mechanism between global exploration and local exploitation, activating Tabu Search in later iterations to deeply optimize the Pareto front’s elite set, effectively overcoming the premature convergence and result randomness of traditional genetic algorithms. A multi-objective optimization model, balancing governance effectiveness (unbalance degree) and economic cost (switching operations), was established. Simulation results show that the proposed NSGA-TS algorithm demonstrates superior optimization performance compared to several mainstream heuristic algorithms. More critically, in rigorous stability tests involving 100 independent runs, the algorithm’s decision results showed high convergence, proving it successfully resolves the core issue of uncertain results in single runs of heuristic methods and meets the stringent reliability requirements of online control scenarios.