Qian Ning

The development of clean and sustainable energy has become a global focus, with nuclear energy playing an important role in low-carbon energy systems. To address the complex cable routing problem in nuclear island environments, this study proposes a three-dimensional cable routing optimization framework based on an improved A* algorithm. The proposed method incorporates multiple engineering-related constraints, including cable type, protection level, turning angle, and construction standards, while optimizing routing paths in constrained spatial environments. A simplified spatial expansion strategy is adopted to approximate routing safety margins while maintaining computational efficiency. Experimental results demonstrate that the proposed method effectively reduces the number of bends and shortens routing paths under predefined engineering constraints, thereby improving routing efficiency and reducing installation costs. The proposed framework provides a preliminary engineering-oriented routing optimization approach and preserves extensibility for future integration of thermal dissipation and electromagnetic compatibility constraints.

Source-Free Domain Adaptation (SFDA) adapts a pre-trained source model to an unlabeled target domain without accessing source data, alleviating the need for direct access to source data during adaptation, which reduces the risk of data transmission. Existing methods leverage pre-trained Vision-Language (ViL) models to provide supervisory signals, but their predictions are often noisy, limiting adaptation performance. Moreover, current approaches refine ViL predictions only within individual source domains, overlooking the complementary discriminative evidence available across multiple sources. To better select the most decisive refinement signal from each source model, we propose a novel criterion, termed M aximum R efinement D ecision score ( MRD-score ), which replaces softmax with median centering to normalize source models’ predictions along both positive and negative axes, thereby harnessing both affirmative and complementary guidance. Building upon MRD-score, we present M ulti-source D ual- R efining A daptation N etwork ( MDRAN ), which, to the best of our knowledge, is among the first attempts to exploit ViL-based guidance for multi-source-free domain adaptation (MSFDA). MDRAN first aggregates the MRD-score-based predictions from multiple source models to construct S elected Source- L ogit S et ( SLS ), which is then used to customize a collective ViL model via prompt learning. To mitigate noise in the collective ViL models’ outputs, we design a D ual- V iew R efinement ( DVR ) module that refines ViL model’s predictions using source model predictions. Extensive experiments on four benchmarks demonstrate that our method achieves state-of-the-art performance on the more challenging Office-Home, VisDA-C, and DomainNet-126, while remaining highly competitive on the comparatively easy Office-31. The code can be found in https://anonymous.4open.science/r/MDRAN-098E/ .

Based on a fully reproducible integrated nonlinear dynamic model of a pressurized water reactor (PWR), this paper presents a decentralized control structure comprising four control loops—coolant average temperature, pressurizer pressure, pressurizer level, and turbine speed—and five PI controllers. To improve controller parameter tuning efficiency and enhance the overall control performance of the PWR system, the NSGA-II algorithm is adopted to perform multi-objective collaborative optimization of the multi-loop PI controller parameters, with the integral of time-weighted absolute error (ITAE) of coolant average temperature, ITAE of turbine speed, and maximum negative overshoot of reactor power as the objectives. Through the obtained Pareto front, the non-dominated trade-off relationships among nuclear island thermal stability, conventional island regulation performance, and reactor power safety recovery characteristics are analyzed. An optimal solution with balanced comprehensive performance is selected from the Pareto set, and closed-loop comparative simulations against engineering-experience-tuned parameters are conducted under two load reduction scenarios. The results demonstrate that the optimized controller parameters achieve superior control performance in both step load rejection and linear load reduction transients, with improved load-following capability.

Cable routing is a pivotal design component for electrical systems and safety-critical engineering fields, such as nuclear propulsion systems, nuclear power plants and aircraft. Scientific and optimized routing schemes are essential for efficient and safe power and signal transmission and for mitigating system failure risks. Previous studies have adopted heuristic search and swarm intelligence optimization algorithms for cable path planning; however, these methods tend to converge to local optima under complex constraints and cannot theoretically guarantee global optimality, failing to address multi-constraint, high-dimensional optimization challenges of confined-space cable routing. This paper proposes a mathematical programming-based systematic optimization model: it first discretizes continuous three-dimensional space into a grid coordinate system and constructs a composite cost field integrating geometric distance and thermal interference, then formulates a multi-objective optimization model considering path length, thermal impact and routing feasibility, which is converted into a single-objective problem via normalized weighting coefficients and solved by exact mathematical programming techniques, yielding a best feasible solution together with a provable lower bound and an optimality gap. When the solver converges within the time limit, global optimality for the discretized model can be certified. Simulation results show the proposed method reduces overall path cost by an average of 31.8% compared with classical algorithms like the A* algorithm, Dijkstra’s algorithm, Rapidly-exploring Random Tree (RRT), Particle Swarm Optimization (PSO), and Genetic Algorithm (GA). Furthermore, it cuts decision variables by an average of 70% (up to 82% in complex scenarios) against the 0–1 Integer Linear Programming (ILP) model and the graph-theoretic Multi-Commodity Flow (MCF) model with multi-cost considerations. These results preliminarily validate the favorable solution quality, computational efficiency and engineering applicability of the proposed model for confined-space cable routing optimization.

Flexible load-following operation of pressurized water reactors is becoming increasingly important in low-carbon power systems with high shares of variable renewable generation. However, the stochastic and fast-varying nature of wind power is poorly matched to the large thermal inertia of nuclear units. Direct tracking of high-frequency disturbances may lead to excessive actuation of turbine governor valves, thereby increasing mechanical burden and disturbing primary-loop thermal behavior. To address this issue, this paper proposes a decentralized constrained control architecture for a nuclear-wind-storage hybrid system aimed at supporting flexible load-following operation of a pressurized water reactor. In the proposed framework, a constant-coefficient linear model predictive controller regulates turbine-side power tracking under hard actuator rate constraints, while an independent PID loop maintains primary-loop temperature regulation. Fast net-load variations that exceed the dynamic response capability of the nuclear unit are compensated through battery energy storage system support based on the measured frequency-deviation signal within the adopted simulation framework. Simulations driven by empirical wind-power data show that the proposed architecture maintains acceptable load-following performance and primary-loop thermal stability while substantially reducing the high-frequency energy content of turbine-valve commands. Compared with an unconstrained PID benchmark, the high-frequency energy ratio of the turbine-valve signal is reduced by nearly 90\%. These results indicate that actuator constraints can provide a practical coordination mechanism for improving the flexible operating capability of nuclear units in hybrid energy systems while alleviating unnecessary mechanical burden associated with fast disturbances.

Semantic communications (SemCom) and uplink rate-splitting multiple access (RSMA) are promising enablers for efficient 6G systems, yet their integration poses new resource allocation challenges due to the intrinsic coupling between semantic compression and multi-user interference management. This letter investigates semantic-aware resource allocation for an uplink RSMA system. Joint optimization of semantic compression, bandwidth allocation, and transmit power is formulated to maximize long-term task transmission efficiency. The problem is modeled as a Markov decision process and solved using a Soft Actor-Critic framework. The simulation results demonstrate robust performance gains over baseline schemes under various resource constraints.