Embedding metal foam into phase change materials is an effective approach to enhance the thermal performance of latent heat thermal energy storage systems. However, heat transfer mismatches caused by coupling between conduction and natural convection remain a challenge. Unlike most existing studies focusing on uniform or one-dimensional porosity-gradient structures, this work investigates the effect of the inclination angle of the porosity-gradient interface on melting behavior and heat transfer performance. A three-dimensional pore-scale numerical simulation is employed to analyze the melting process of gradient-porosity metal foam composite PCMs. The effects of the inclination angle (theta), porosity gradient magnitude (Delta epsilon), and thermal energy storage temperature difference (Delta T) are evaluated. The results show that melting performance first improves and then deteriorates with increasing inclination angle, with optimal performance at theta = 30 degrees. Compared with a uniform-porosity structure, this configuration reduces the total melting time by 12.16% and increases the thermal energy storage rate by 11.66% by suppressing melting dead zones and improving the balance between heat conduction and natural convection. The effect of porosity gradient depends strongly on interface orientation: a moderate gradient (Delta epsilon = 0.12) is optimal for theta = 0 degrees and 30 degrees, whereas smaller gradients are preferable at theta = 90 degrees. Increasing Delta T consistently accelerates melting, for the optimal structure (theta = 30 degrees, Delta epsilon = 0.12), increasing Delta T from 25 K to 40 K reduces melting time by 29.06% and enhances the energy storage rate by 55.14%. Overall, this study provides new insight into pore-scale heat transfer regulation in gradient-porosity LHTES structures and offers practical guidance for optimal design.

To improve the energy efficiency of 5G telecommunication base station (TBS) cooling systems, a cabinet-level variable-speed vapor compression system (VVCS) is developed. However, the system's multiple operating parameters are complex and strongly coupled, which makes real collaborative regulation and optimization difficult. To overcome this challenge, a 5G TBS simulated experimental platform is established based on the VVCS for system testing and performance analysis. Initially, the operating parameters, which include compressor speed (Vcom), condenser fan speed (Vc), evaporator fan speed (Ve), and electronic expansion valve (EEV) opening degree, are regulated and optimized under the conditions of an ambient temperature of 30 degrees C and a thermal load of 6 kW. This optimization takes into account factors such as superheating and discharge temperature control, ultimately aiming to enhance the system's energy efficiency ratio (EER). Furthermore, the operational characteristics and energy efficiency of the VVCS are investigated under three different operating modes: variable speed parameter optimization, variable speed on-off, and constant speed on-off. The results indicate that the optimized parameters are Vcom (60 %), Vc (80 %), Ve (80 %), and EEV opening (40 %). Under these conditions, the evaporator superheating degree is 6.3 degrees C, and the compressor discharge temperature is 51.5 degrees C. Additionally, the optimized EER is 4.2, which is 8.81-33.33 % higher than the EERon-off of the variable speed on-off operation mode and 38.61 % higher than the EERon-off of the constant speed on-off operation mode. This study provides valuable references for optimizing control strategies and enhancing energy efficiency in 5G TBS cooling systems.

This study numerically investigates the transient thermal response of a raised-floor data center under dynamic AI server loads. Power profiles of DGX A100 servers running GEMM, VGG-19, and ResNet-152 workloads were used as time-varying boundary conditions in CFD simulations. Six load scenarios were analyzed to evaluate the dynamic responses of cooling capacity, rack inlet and outlet temperatures, and net airflow rate. The results show that the response lag of air-conditioning units causes a temporary mismatch between cooling supply and server heat dissipation, leading to transient inlet-temperature fluctuations. Fluctuating workloads produce larger outlet-temperature amplitudes and higher temperature change rates than stable workloads, while mixed workloads change the overall fluctuation intensity and affect neighboring-rack airflow through cold-aisle pressure interactions. These findings can support workload-aware rack arrangement and adaptive cooling control for future AI-oriented data centers.

Waste heat recovery from data center cooling systems offers a viable pathway toward energy reduction. Cooling parameters and waste heat recovery strategies mutually influence the operational performance of both systems, further complicating collaborative optimization under multiple parameters and conflicting objectives. To address the above challenges, this paper introduces a double-layer multi-objective optimization framework for data center combined cooling and waste heat recovery systems. The global layer minimizes power usage effectiveness (PUE) by optimizing cooling parameters, including the computer room air handler and chilled water supply temperatures, and the local layer simultaneously minimizes total system's energy consumption and operational cost under peak-valley electricity pricing by adjusting operation strategies. Compared to a rule-based control strategy, the double-layer multi-objective strategy under optimal supply temperatures (18 degrees C for chilled water and 25 degrees C for computer room air handler) reduced PUE by about 0.88%, energy consumption by 9.37-11.16%, and operational cost by 16.26-18.26%. When evaluated against single-objective strategies, the double-layer multi-objective framework showed balanced performance: against the energy minimization strategy, it increased energy consumption by only 0.16-0.68%, but lowered cost significantly by 3.23-5.87%. Against the cost minimization strategy, it reduced 0.18-0.78% energy consumption with a slight cost increase of 0.03-0.54% in winter and transition seasons. In summer, energy consumption decreased by 0.07% despite a 2.57% increase in operational cost due to low heating demand. The proposed framework enables collaborative optimization of cooling and waste heat recovery, achieves a trade-off among conflicting objectives, delivers more comprehensive benefits, and provides a replicable control paradigm for data center optimization.

The development of low-carbon energy systems constitutes a critical pathway for industrial parks. However, neglecting the correlation between energy sources and loads will elevate decision-making risks during system optimization. This study proposed a multi-objective optimization framework for the load coordination of electric and heat in low-carbon energy parks. Firstly, the heating system and the battery energy storage system were constructed as adjustable resources in the low-carbon energy system of the industrial park. Secondly, it indicated that a fundamental conflict was between carbon emissions and the total costs in the planning process. Increasing the capacity of photovoltaic and wind power had significantly reduced carbon emissions, but had also led to an increase in the total cost. The concurrent management of electric and heat loads allowed surplus green power to be diverted for heating purposes, providing a pathway to phase out gas boilers. Finally, based on the optimized trade-off between carbon emissions and the total costs, three representative configuration plans were proposed to support the implementation of the low-carbon industrial parks, including the economical practical (Ⅰ), balanced coordinated (ⅠI) and low-carbon (ⅠII). Besides, The typical days were selected by k-means from the 8760 h of electric and heat loads in the actual operating park, thereby making the optimized configuration of electricity and heat coordination more in line with the constraints of the real world. The findings provided direct and reliable references for the design of low-carbon energy systems.

Traditional thermosyphon systems (TS) for 5G telecommunication base stations (TBS) typically control fan on-off operating based on the return air temperature (Treturn), resulting in inefficient utilization of natural cold sources and reduced energy efficiency. Furthermore, temperature oscillation during TS startup compromises precise temperature control. To address these issues, this study developed a cabinet-level variable-speed loop thermosyphon system (VLTS) and established the 5G TBS simulated experimental platform. Through a series of experiments, the impact of evaporator fan speed (Ve) and condenser fan speed (Vc) on the system's operating performance was investigated. The influence of control parameter selection on the system's energy efficiency improvement is analyzed. In addition, based on experimental analysis, two energy efficiency optimization strategies are proposed: the variable-speed parameter optimization strategy (VPOS) and the high-low speed switching strategy (H-LSS). These strategies aim to enhance the system's EER while ensuring the thermal safety of IT equipment and preventing temperature oscillation. A comparison of the operational effects between the two strategies is also conducted. Results indicate that the Ve has a more significant impact on system performance than the Vc. Maintaining the Ve above 50% can effectively prevent system temperature oscillations. Compared to using Treturn as the control parameter, using the outlet air temperature of IT equipment (TL,o) as the control parameter can further enhance the system's energy efficiency. The optimized parameter combination (Ve 50%, Vc 40%) increases the system's EER to 38.8 through the VPOS. Both VPOS and H-LSS strategies can effectively avoid system temperature oscillation. Compared to the H-LSS, the VPOS improves the system's EER by 20.8% to 112.7%. For the H-LSS, raising the lower limit of the Treturn setpoint from 37 degrees C to 38 degrees C increases the system's EER by 12.8% to 29.9%.

The application of cold plate technology enhances the cooling efficiency and waste heat recovery potential of data centers. This paper proposes a hybrid cold plate system with waste heat recovery (HCPHR) to simultaneously reduce the energy consumption of data center cooling and residential district heating. At the case data center, energy efficiency, environmental and economic potential of the proposed system are simulated and analyzed in the five typical climate cities, by comparing the previous system (PS) and the row-level thermosyphon heat recovery system (RLTHR). In the selected cities, the annual energy consumption of cooling system of RLTHR and PS is similar, and the cooling system's annual energy consumption and Power Usage Effectiveness (PUE) of HCPHR are 40.87-53.56 % and 0.0393-0.1125 lower than those of PS, respectively. In addition, district heating systems' annual energy consumption of RLTHR and HCPHR is 26.94-50.66 % and 90.96-94.35 % lower than that of PS, respectively. Thus, the annual energy consumption and CO2 emissions of HCPHR are 41.98-70.77 % and 2701.51-6261.00 tons lower than those of other two systems, respectively. Moreover, the dynamic payback periods of RLTHR are only 1.88-3.76 years and 2.59-4.59 years compared to PS and HCPHR, respectively. Therefore, HCPHR has good energy efficiency, environmental and economic potential.

To improve the thermal energy storage performance of latent heat thermal energy storage (LHTES) units under high-power conditions, this study develops a bio-inspired spiderweb-fin structure with dual-gradient thickness distributions in both the radial and circumferential directions. Existing spiderweb-fin studies have mainly focused on uniform-thickness designs or single-directional thickness variation, while the coupled effects of multidirectional thickness gradients on phase change heat transfer remain insufficiently understood. To address this gap, four key gradient parameters are introduced to characterize the thickness variation of radial and circumferential fins, and their effects on liquid fraction evolution, temperature distribution, natural convection, and thermal energy storage rate (TESR) are systematically investigated. The results show that the circumferential thickness gradient of radial fins has the strongest influence on TESR, and increasing it from 1.0 to 1.3 enhances TESR by 34.12%. Increasing the radial thickness gradient of radial fins significantly strengthens natural convection, resulting in a maximum TESR improvement of 14.38%. In contrast, the two thickness gradients of circumferential fins mainly improve heat conduction in regions far from the heated wall, thereby promoting a more favorable melting pattern. Response surface optimization identifies an optimal parameter combination with a predicted TESR of 287.39 J/s. This study establishes a dual-gradient design framework for multiparameter fin optimization in LHTES units.

With the rapid advancement of digital transformation and the growing demand for artificial intelligence applications, the performance improvements of server chips have driven an exponential increase in thermal flux density, posing significant challenges for thermal management and energy consumption in data centers. In the traditional method, the cooling terminal and information technology (IT) system are independently controlled, resulting in inefficient cooling and potential thermal risks. Consequently, this study proposes a co-optimization methodology integrating thermal-aware workload scheduling with deep reinforcement learning (DRL)-based cooling terminal operation control, designed to minimize energy consumption while satisfying thermal safety constraints. Specifically, first, a model for predicting server outlet temperature was established based on a deep neural network (DNN). Second, a thermal-aware workload scheduling algorithm is proposed. Finally, the Soft Actor-Critic (SAC) algorithm is employed to achieve coordinated control between the cooling terminal and IT equipment. Comparison results with traditional methods demonstrate that the proposed thermal-aware workload scheduling algorithm reduces the maximum server outlet temperature by 0.12-0.43 degrees C and achieves optimal temperature uniformity. Compared with the conventional control method, the proposed methodology reduces energy consumption by up to 8.6 % while significantly improving temperature control effectiveness, reducing the number of temperature threshold violations from 184 to 1.

Latent heat thermal energy storage offers high energy density and nearly isothermal operation, yet its practical deployment is restricted by the inherently low thermal conductivity of phase change material. In this study, a numerical investigation is conducted to enhance the melting performance by incorporating a rotational mechanism with a bionic spider-web fin structure. The enthalpy-porosity model is employed to simulate the melting behavior of RT55, and the effects of rotational speed (m), circumferential fin spacing reduction parameter (alpha), and radial fin thickness increment ratio (beta) are systematically examined. Performance is evaluated in terms of melting time, total stored energy, and energy storage rate, and response surface methodology is applied for multi-objective optimization. Results show that rotation at 0.3 rpm shortens the melting time by 24.94% and increases energy storage rate by 25.78%; alpha = 4 reduces melting time by 14.2% and raises energy storage rate by 16.54%; beta = 1.4 decreases melting time by 30.67% and enhances energy storage rate by 40.32%. Response surface methodology identifies beta as the dominant factor, and the optimal parameters (alpha = 4, beta = 1.4, m = 0.035 rpm) achieve a predicted melting time of 3607.02 s and a total stored energy reaches 1.40 x 107 J, with a verification error below 0.55%. The study demonstrates that the coupling of rotation and non-uniform spiderweb fins markedly improves heat transfer uniformity and energy storage efficiency, offering practical guidance for solar energy collection and waste heat recovery applications.

With the rapid development of artificial intelligence (AI) computing and edge computing infrastructure, high-power-density data centers are placing increasingly stringent demands on the heat dissipation capability, rapid deployment, and flexible scalability of cooling systems. In addition, the mixed deployment of AI and general computing, along with the differentiated heat dissipation requirements of various electronic components, makes traditional air cooling methods increasingly inadequate. To address these challenges, this study proposes a novel air-liquid hybrid cooling system for containerized AI data centers, integrating a water-cooled pump-driven heat pipe in-row air conditioner, a staged two-phase cooling distribution unit (CDU), and dry-wet hybrid cooling towers. Heat transfer and power consumption models were developed for the system, and the corresponding models for the aforementioned three subsystems were validated against measured data. Compared with a baseline air-liquid hybrid cooling system, the proposed system achieved higher energy efficiency ratios (EER) across different outdoor temperature conditions. Specifically, the EER ranged from 9.94 to 31.49 at an air-to-liquid cooling load ratio of 3:7 and from 12.87 to 33.79 at 2:8. A year-round assessment was conducted for 21 representative cities across different climate zones in China. The results showed that the annual EER (AEER) of the proposed system ranged from 10.97 to 16.88 at an air-to-liquid cooling load ratio of 3:7 and from 13.61 to 19.79 at 2:8, representing improvements of 13.15%–32.98% and 16.46%–30.38% over the baseline system. The corresponding power usage effectiveness (PUE) ranged from 1.109 to 1.141 and from 1.101 to 1.123, while the water usage effectiveness (WUE) ranged from 0.565 to 0.988 L/kWh and from 0.375 to 0.789 L/kWh. The results can serve as a reference for the design and deployment of hybrid cooling systems for containerized AI data centers in different climate zones.

This study investigates the charging performance of a cascaded latent heat storage (CLHS) system under periodic linear fluctuating heat sources, which are representative of solar thermal and chip waste heat applications. A novel CLHS structure is proposed based on the characteristic evolution patterns of the solid-liquid phase change interface. By establishing a two-dimensional transient model using the enthalpy porosity method, the effects of the fluctuation period, amplitude, and dimensionless parameter related to the initial inlet temperature and heat capacity are systematically investigated. The results indicate that the heat source fluctuation period is the dominant factor governing thermal storage performance. The optimal period of 6000 s was determined, which maintains stable natural convection while achieving the shortest melting time (4446.8 s). Compared with a constant-temperature heat source, the melting time was reduced by 41.8 %, and the average heat transfer rate increased by 76.7 % (reaching 143.34 W). This study revealed that increasing amplitude enhances transient temperature gradients and natural convection, shortening the melting time by 17.3 % (Amplitude 30 degrees C), but this advantage diminishes at shorter periods. Conversely, initial inlet temperature has a negligible influence, primarily affecting the sensible heat phase change process, while having a limited impact on latent heat absorption and natural convection mechanisms. This study provides new insights and theoretical foundations for designing high-performance thermal energy storage systems capable of addressing practical non-steady-state heat sources.

Model predictive control (MPC) has been widely used to optimize the energy consumption of data center cooling systems. However, performance degradation caused by equipment aging progressively degrades MPC model accuracy, causing the model predictions to deviate from actual data and consequently reducing energy-saving effectiveness. To overcome this limitation, a novel aging-aware adaptive MPC (AMPC) strategy is proposed. The key innovation is the embedding of real-time aging equipment status as time-varying constraints within the MPC optimization, combined with a recursive correction mechanism that continuously updates the model's aging factor to eliminate prediction mismatch. By simulations, the detrimental impact of aging on a traditional MPC (TMPC) is quantified and the performance of AMPC against both TMPC and an unoptimized strategy is compared. Results show that TMPC's energy-saving benefit declines over the equipment life; by year 12 its energy consumption is merely higher than the unoptimized strategy by 1.14%. In contrast, AMPC achieves a total lifecycle energy consumption of 55,395.63 MWh, representing 12.40% savings relative to the unoptimized strategy and 6.26% savings relative to TMPC. This demonstrates that the proposed adaptive strategy effectively counteracts the adverse effects of equipment aging on data center energy efficiency.

Parameter coupling of a combined data center cooling and waste heat recovery system increases control complexity.Model,measurement,and execution errors significantly reduce control accuracy and limit improvements in energy efficiency.To address multi-objective conflicts affecting system benefits and quantify performance fluctuations from uncertainty parameters,this study proposes a multi-objective optimization strategy to collaboratively optimize the energy consumption and operation cost of the combined cooling and waste heat recovery system in the Dongjiang Lake water source data center and uses Monte Carlo simulation to quantify the robustness of the control strategy under different uncertainty parameters.Compared with those of rule-based control,the multi-objective optimization strategy reduces the total energy consumption by 11.07%,operational costs by 16.25%,and PUE by 0.01.Relative to those of single-objective energy optimization,energy consumption increases marginally(0.28%),whereas costs decrease significantly(3.20%).Compared with those of single-objective cost optimization,energy consumption decreases by 0.77%,with only a 0.54%cost increase.Although multi-objective optimization exhibits slightly higher variation coefficients for individual performance metrics than those of single-objective optimization strategies,its energy consumption variation is 2.8%lower than that of single-objective cost optimization,while cost variation is 2.2%lower than that of single-objective energy optimization.This strategy maintains relatively low heat storage/release mode misjudgment rates,confirming the global robustness advantages under multi-parameter uncertainty.

Model predictive control (MPC) strategy is one of advanced control methods that can enhance the energy efficiency of data center cooling systems. However, traditional MPC (TMPC) strategies fail to account for the impact of performance degradation on accuracy of the model prediction under equipment aging, which leads to a reduction in energy-saving effect. In this study, an aging-aware improved MPC (AMPC) strategy is proposed and investigated to overcome the limitations of TMPC, which incorporates the actual performance of aging equipment as constraints, dynamically updates the aging factor of the predictive model in a rolling manner. For comparison, the unoptimized strategy and TMPC strategy are adopted as benchmark schemes. The results show that the total energy consumption of the cooling system throughout the entire life cycle of the data center under the unoptimized strategy, TMPC strategy and AMPC strategy are 63237.41MWh, 59091.89MWh and 55395.63MWh respectively. TMPC strategy has a good energy-saving effect in the early stage of equipment operation. However, in the 12th year, the total energy consumption of the cooling system under the TMPC strategy was 1.14

A novel micro channel separated heat pipe system coupled with radiative sky cooling is proposed to reduce the cooling energy consumption in data centers. The system can operate in two modes, including the conventional heat pipe mode and the coupled heat pipe mode with radiative sky cooling. Two prototype systems with different structures of radiative cooling heat exchanger (RCHE) were designed, built and experimentally tested. The surface of RCHE is pained with a spectrally selective absorbing material, which has a low absorptivity in the solar irradiation band and a high emissivity in the atmospheric window band. The temperatures of air and refrigerant along with the refrigerant pressure were measured. The subcooling degree, cooling capacity and energy efficiency ratio (EER) were calculated and analyzed. Experimental results indicate that compared to the conventional system, refrigerant can be cooled further in the coupled system due to the assistant cooling by the RCHE through radiating heat to the cold outer space, with a higher subcooling degree at the inlet of evaporator by 0.4 degrees C on average. Moreover, the coupled system had 18.91% more cooling capacity and higher EER. Then, the performances of the coupled system with different structures of RCHE were compared. It was observed that 10.7% more cooling capacity and higher EER were obtained by the coil tube structure coupled system over the parallel tube structure coupled system.

With the increasingly serious shortage of resources and environmental pollution caused by fossil fuels, green renewable energy has become a key focus of global development. As a crucial field, wind power is developing towards large-scale and high efficiency. The dynamic response of wind turbine blade, as a core load-bearing component, directly affects the safety and stability of the whole machine. Aiming at the trade-off between the efficiency and accuracy of traditional multibody dynamics (MBD) and finite element analysis (FEA), this study proposes two blade response prediction models based on the long-short-term memory network (LSTM) and Bayesian optimization (BO), which can efficiently predict the response of MBD and FEA simulation data, respectively. The results show that the optimized LSTM model achieves mean square error (MSE) 0.0011, mean absolute error (MAE) 0.0196, and coefficient of determination (R2) 0.9964 in MBD prediction, and MSE 0.0025, MAE0.0322, and R2 0.9922 in FEA prediction, which demonstrate high fitting accuracy and generalization ability. For the first time, based on the high-precision simulation data of WeMoLab platform, the modeling and prediction of two types of dynamic responses are realized with high efficiency, and the computational cost is greatly reduced. The proposed model provides reliable data-driven support for wind turbine blade load prediction, structural optimization, and operational state assessment and has wide engineering application value.

Literature study indicates that drivers spend about 79 min on average each day traveling to their destination. However, continuous driving in an enclosed vehicle can lead to a significant increase in the concentration of contaminants inside the vehicle, especially CO2. Despite this, automotive indoor air quality (IAQ) receives less attention. In this study, a gasoline-powered passenger vehicle was selected to study car cabin IAQ. Sensors were used to measure the concentration of O2, CO, CO2, total volatile organic compounds (TVOCs) and PM10 as well as temperature and humidity in the vehicle. The results revealed that the O2 content remained within adequate levels and TVOCs concentration remained low and safe, with values less than 1 ppm. However, CO2 concentration could exceed 2000 ppm within just 30 min when the ventilation flap was closed, both for idle and driving mode. Opening the flap could lead to increases in carbon monoxide (CO) and PM10 concentrations to more than 10 ppm and 0.15 mg/m3, respectively, particularly during traffic congestion. Even when the ventilation mode is switched to closed circuit mode, exfiltration and infiltration of contaminants may still occur.