In recent years, photocatalytic reduction of CO2 using renewable solar energy has emerged as a promising technology for addressing the energy crisis. To achieve efficient CO2 photoreduction, enhancing the sunlight absorption capacity and charge separation efficiency of g-C3N4-based photocatalysts was crucial. Hydrothermal preparation of dual Z-scheme Cu2-xS/g-C3N4/NiAl-LDH composite catalysts was used in this study. The optimized Cu2-xS/g-C3N4/NiAl-LDH photocatalyst demonstrated CH4 and CO yields of 28.13 and 4.47 & micro;mol g-1h-1, respectively. Notably, the CH4 yield was 12.5 times higher than that of the LDH monomer, and the selectivity of CH4 was as high as 96.2%. The promotion mechanism could be attributed to the photothermal effect of composite catalyst and the augmented light absorption, as well as the efficient internal charge transfer originated from the bi-heterostructure. Consequently, combining Cu2-xS/g-C3N4/NiAl-LDH and photothermal effects, this photocatalyst design provides insightful information for creating full-spectrum responsive photocatalysts for CO2 reduction.
The widespread adoption of Power Electronic Converters (PECs) generates substantial high-frequency (HF) switching oscillations caused by switching transients. These oscillations not only compromise system reliability but also result in energy losses. This paper proposes a novel non-contact energy harvesting technique to capture and utilize the energy associated with HF switching oscillations in PECs. The approach employs a magnetically coupled HF resonant energy harvesting device, which simultaneously enables efficient energy extraction and suppresses switching oscillations. The study begins with an analysis of the generation mechanism of HF switching oscillations in PECs, followed by an examination of the converter’s equivalent circuit, the design of the external energy harvesting device, and its impact on the original system. Experimental validation is conducted using a 300 V/390 W SiC MOSFET Buck converter operating at its resonant frequency. Results demonstrate that the proposed method achieves power output in the hundred-milliwatt range, reduces switching oscillations by 32.4%, and enhances overall system efficiency. The proposed method provides a promising solution for simultaneous switching oscillation suppression and energy recovery in power electronic converters.
Layered oxide cathodes for potassium-ion batteries (PIBs) undergo substantial structural strain during cycling because of the large K+ radius. Here, Co modification is introduced into O3-type Na0.8Cr0.8Ti0.2O2 to regulate structural evolution and electrochemical reversibility. The resulting Na0.8Cr0.8Ti0.15Co0.05O2 (NCTCO) contains substitutional Co in the layered host together with localized CoCr2O4 spinel domains. NCTCO delivers 70.3 mA h g−1 at 1000 mA g−1 and retains 96.4% capacity after 500 cycles at 500 mA g−1 in K-metal half-cells. Electrochemical analysis shows a lower charge-transfer resistance after Co modification. In situ XRD reveals reduced lattice variation during the first charge, with a three-layer-equivalent c-axis expansion of 1.10% compared with 2.50% for NCTO, together with suppression of a high-voltage P1-type component. DFT calculations using idealized layered models indicate that substitutional Co lowers a representative K-ion migration barrier from ∼0.45 to 0.39 eV. These results show that Co modification improves the structural reversibility and electrochemical stability of the O3-type layered cathode while clarifying the distinct experimental and modeled roles of the two Co environments.
Driven by carbon neutrality goals and increasing wind penetration, accurate turbine-level wind power forecasting is essential for grid operation, wind farm digital twins, and intelligent dispatch. However, wind power evolution involves coupled wake-induced spatial interactions, non-stationary temporal dynamics, and long-range spectral periodicity, which are difficult to capture with single-domain forecasting models. This paper proposes TriModal-STGNN, a physics-guided tri-modal spatio-temporal graph neural network that formulates turbine-level prediction as a coupled spatial-temporal-spectral representation learning problem. The framework integrates physics-guided spatial encoding, adaptive temporal feature extraction, multi-scale frequency-domain representation, and gated tri-modal fusion to jointly model wake-aware turbine interactions, transient dynamics, and periodic evolution. Experiments on the 134-turbine SDWPF dataset from China and the 14-turbine Penmanshiel dataset from the UK evaluate its forecasting performance and cross-dataset transferability. On SDWPF, TriModal-STGNN achieves aggregate MAE reductions of 17.22% and 22.03% relative to the horizon-wise strongest baseline and the mean of the top-five strongest non-proposed baselines, respectively. On Penmanshiel, the corresponding reductions are 3.97% and 12.18%. Ablation, interpretability, and efficiency analyses further show that the performance gains arise from coordinated spatial-temporal-spectral modeling rather than model scaling.
Multi-sensor fault diagnosis in rotating machinery is vulnerable to corrupted or missing sensor signals, which can distort shared representations before reliability is assessed. We propose a Physics-Informed Mahalanobis Informatics Framework (P-MIF) that evaluates source reliability before feature fusion. Target frequencies selected from physical priors or spectral analysis are encoded by a Morlet continuous wavelet transform filter bank, while grouped convolutions isolate sensor–subband streams. Channel-wise Mahalanobis distances measure departures from clean class-conditional manifolds; split conformal calibration converts them into comparable reliability scores for down-weighting unreliable streams during fusion. P-MIF was evaluated against six baselines on the Paderborn, GearEcc, and MCC5-THU Motor datasets under localized Gaussian noise, sensor loss, and realistic degradation. Under localized Gaussian noise at −10 dB, P-MIF achieved mean accuracies of 97.28% with Paderborn Sensors 1 and 2 simultaneously corrupted and 95.29% with four of the six MCC5 sensors corrupted, outperforming the corresponding strongest baselines by 87.81 and 77.92 percentage points, respectively. These results show that physics-guided isolation and geometry-calibrated weighting enable robust, interpretable multi-sensor diagnosis under partial source degradation.
