Abstract: Recirculating aquaculture systems (RAS) have emerged as a key driver for the green upgrading of the aquaculture industry amid the global shift toward intensification and ecological sustainability. However, the widespread application of RAS still faces challenges such as the limited removal of characteristic pollutants (e.g., high-concentration inorganic nitrogen and trace organic compounds) and high operational energy consumption. From a systems engineering perspective, this review comprehensively discusses the coupling mechanisms and targeted regulation strategies of the core water treatment units in RAS. For biological nitrogen removal, we evaluate autotrophic nitrogen removal pathways (e.g., Feammox) that potentially break through the efficiency bottlenecks of conventional nitrification-denitrification. While these novel pathways demonstrate promising energy savings (up to 44.7% reduction compared to complete nitrification-denitrification), their practical application remains largely at the laboratory scale, with unresolved challenges in reactor stability and process control under real aquaculture conditions. Furthermore, to address the current difficulties in controlling biofilm thickness within existing processes, we propose that future optimizations should focus on achieving the targeted enrichment of microbial communities and steady-state maintenance through filler modification and fluid shear stress regulation. Regarding pathogenic risk control, conventional chemical disinfection and antibiotics lack selectivity, disturbing the system's ecology and exacerbating antimicrobial resistance. Cost-effective and environmentally friendly options like performic acid are therefore gaining attention. In contrast, precision technologies such as quorum quenching and gene silencing are theoretically attractive but face significant engineering challenges including activity maintenance and scaling costs in the near term. For multiphase separation, conventional mechanical micro-screens fail to intercept fine particles (< 30 μm), leading to excessive organic accumulation. Microbubble-driven dissolved air flotation offers a highly efficient alternative to overcome this physical interception limit, significantly reducing the biochemical oxygen demand on downstream biofilters. Furthermore, while advanced oxidation processes (AOPs), such as ozone and photoelectrocatalysis, effectively degrade refractory dissolved organics and trace antibiotics, they introduce an operational trade-off where excessive oxidation substantially increases energy demands and risks inducing the horizontal transfer of antibiotic resistance genes (ARGs). Thus, balancing oxidation dosages for both ecological safety and energy efficiency remains paramount. Looking forward, we outline the development trajectories of RAS toward system integration and digital operation and maintenance. We propose that under the Water-Energy-Food nexus framework, concrete pathways for technology implementation include: (i) computational fluid dynamics-assisted hydrodynamic optimization to reduce hydraulic energy consumption; (ii) artificial intelligence-enabled predictive water quality control to achieve feedforward regulation rather than lagged responses; (iii) the integration of renewable energy sources (e.g., solar and wind) to power electrolytic oxygen supply and temperature control, thereby lowering operational costs; and (iv) digital twin technologies integrating real-time sensor data with mechanistic models for flow field reconstruction and early warning. Collectively, these cross-disciplinary innovations are driving the transformation of aquaculture models toward the deep fusion of "data-driven and mechanistic models". This review aims to provide theoretical and technological support for achieving a low-carbon transition of aquaculture and the full-process steady-state operation of RAS.