Accurate battery State of Health (SOH) prediction is essential for battery management in electric vehicles and energy storage systems. This paper evaluates MS-DWHM, a multi-scale validation-guided weighted hybrid framework for lithium-ion battery SOH prediction. The framework integrates three complementary predictors: a physics-based Single Particle Model (SPM), a bidirectional Long Short-Term Memory (BiLSTM) network, and an IGWO-AdaBoost ensemble. A compact 17-dimensional feature representation is constructed from physical, temporal, and statistical descriptors, and validation-performance-guided inverse-MSE weighting is used to fuse model predictions without using test labels for weight estimation. On the NASA Prognostics Center of Excellence battery dataset under the adopted batterylevel split, the hybrid model obtains $\mathbf{R}^{2}$ of 0.872, RMSE of 0.037, and MAE of 0.031. Bidirectional-LSTM achieves the highest standalone accuracy in this experiment with $\mathbf{R}^{2}$ of 0.912, indicating that the proposed fusion framework currently emphasizes model complementarity and interpretability rather than universal accuracy dominance. The measured inference latency of 29.58 ms characterizes computational tractability on the test platform.

The lack of high-quality image datasets with quantifiable labels hinders the development of deep learning models for photovoltaic (PV) panel dust detection. To address this, this paper proposes a novel physical consistency modeling approach to generate a large-scale synthetic dataset with precise dust concentration labels. A particle cluster generation model is developed to simulate multiscale, heterogeneous dust aggregation governed by lognormal distributions, while an adaptive error feedback mechanism ensures accurate concentration estimation. Furthermore, an optical attenuation model based on the Beer-Lambert law and nonlinear hue, saturation, and value (HSV) color mapping is employed to ensure visual realism. Multi-dimensional evaluations demonstrate that the synthetic images achieve a histogram similarity exceeding 0.86, an entropy similarity above 0.94, and a comprehensive similarity score over 0.82 compared to real-world ground truth. These results significantly outperform conventional mask-based and generative adversarial networks (GANs) techniques, providing a reliable data source for training advanced dust detection algorithms.

As a core component of electric vehicles and energy storage systems, lithium-ion batteries require precise performance degradation diagnosis to ensure safe and reliable operation. Traditional state-of-health assessment methods primarily rely on macroscopic indicators such as capacity fade and resistance increase (RI), which fail to reveal complex electrode-level degradation mechanisms. This study proposes a novel deep learning and transfer learning (TL)-based framework for accurate quantification of battery degradation modes (DMs) at the electrode level. The method employs a parallel multi-branch convolutional neural network architecture that simultaneously extracts local and global features from low-rate discharge voltage curves using varying kernel sizes, eliminating the need for manual feature engineering. The model quantifies 4 key degradation mechanisms: lithium inventory loss, anode active material loss, positive electrode active material loss, and RI. Training on 26,521 low-rate discharge voltage curves from lithium nickel manganese cobalt batteries demonstrates high prediction accuracy, with mean absolute errors (MAE) below 2.4% for all DMs, including an MAE of 1.8% for lithium inventory loss. TL further enables model adaptation to lithium nickel cobalt aluminum and lithium iron phosphate battery chemistries, ensuring generalization capability. This work provides an accurate and universal degradation diagnostic tool for battery management systems, facilitating precise health monitoring and predictive maintenance throughout battery lifespan.

InAs nanowires (NWs) self-catalyzed grown on graphene surface frequently exhibit a large number of stacking-fault defects. However, the control of these defects in InAs NWs still remains a large challenge, which significantly limits the applications of InAs NWs in electronics and optoelectronics. In this work, the self-catalyzed growth of InAs NWs on graphene/Ge substrate by molecular beam epitaxy (MBE) is systematically investigated. Growth models for InAs NWs and parasitic islands on graphene/Ge are developed. Through rational design of growth parameters, the self-catalyzed growth of defect-free InAs NWs on graphene surfaces is ultimately achieved. Our experimental results indicate that lower growth temperature can effectively suppress the formation of stacking-fault defects in InAs NWs, no visible stacking-fault defects are observed in the samples grown below 510 degrees C, and the intrinsic mechanism for this is clarified with the density functional theory (DFT) calculations.

The self-catalyzed growth of InAs nanowires (NWs) on graphene is achievable but typically yields short, thick NWs with frequent stacking-fault defects. In this work, we demonstrate the successful synthesis of defect-free, ultralong InAs NWs by harnessing the synergistic effects of graphene and an Au catalyst. The distinct roles of graphene and Au in promoting the growth of high-quality NWs were systematically elucidated. Atomic resolution scanning transmission electron microscopy (STEM) imaging revealed that the Au-catalyzed InAs NWs grown on graphene predominantly exhibit defect-free wurtzite structures, whereas a small portion of nanosheets and the NWs grown atop them display defect-free zinc blende phases. Based on these observations, a growth model for Au-catalyzed InAs NWs on defective graphene/Ge surfaces was proposed.

Methane (CH4) emissions from paddies, driven by methanogens and methanotrophs, are major agricultural CH4 sources. While fertilization-induced soil nutrient dynamics and rice growth regulate CH4 fluxes, mechanistic responses to phosphorus (P) availability remain understudied relative to carbon and nitrogen. Based on a long-term paddy experiment without P input in Tai Lake region since 1980, we investigated P addition effects on CH4 emissions, microbial functional genes and communities across four treatments: chemical nitrogen and potassium fertilizer (CNK) vs. CNK+P, chemical NK combined with organic fertilizer (MNK) vs. MNK+P. Relative to CNK and MNK, P addition reduced CH4 emissions by 19.9 % and 35.4 %, respectively, with flux variation positively correlated with mcrA gene at tillering. Unlike genes, alpha diversity and community composition of methanogens/methanotrophs were unaffected by P addition, and treatment differences were mainly driven by historical N fertilizer types. Notably, P addition suppressed microbial P-mining capacity, evidenced by decreased phosphatase activities and phoC and phoD gene abundances, thereby reducing methanogenic substrate levels, reflected by lower dissolved organic carbon and root-secreted organic acids. Furthermore, P input did not disrupt the dominance of type I methanotrophs shaped by historical fertilization, which showed superior CH4 oxidation efficiency. The reduction of CH4 emissions under P addition was driven by the substrate limitation of methanogens, without altering the dominant position of type I methanotrophs. This study highlighted the critical role of P availability in regulating paddy CH4 emissions and advocated optimized P management to balance rice productivity and CH4 mitigation.

Context or problem: Short-term or single biochar (BC) applications have been observed to buffer soil water and temperature variations and improve soil fertility. However, the long-term effects of continuous biochar application on mitigating unfavorable weather conditions remain unclear. Objective or research question: This study aimed to assess whether continuous biochar application could enhance soil fertility, sustainably increase wheat yields, and improve wheat resistance to unfavorable weather conditions over eleven years. Methods: A long-term field experiment was conducted with five treatments: CK (no amendment), Straw amendment (6 Mgha(-1)yr(-1) straw), BC1 (2.4 Mgha(-1)yr(-1) biochar), BC3 (6 Mgha(-1)yr(-1) biochar), and BC5 (12 Mgha(-1)yr(-1) biochar). Wheat yields were monitored from 2011 to 2021, with weather data simultaneously collected from local weather stations. Soil physicochemical properties, including total carbon (TC), available nitrogen (AN), available phosphorus (Olsen-P), and available potassium (AK), and so on, were analyzed in 2021. Results: Wheat yields increased by 10.8 %, 18.7 %, 23.5 %, and 29.4 % in the Straw, BC1, BC3, and BC5 treatments relative to the CK, respectively. The TC increase rate in the 0-15 cm soil layer indicated that the carbon sequestration potential of biochar amendment (BC1) was about three times that of direct straw return (Straw) with an equivalent amount of feedstock. The BC1 treatment significantly increased soil AN, Olsen-P, and AK by 5.36 %, 64.78 %, and 85.61 % respectively, compared to the Straw treatment. A significant negative relationship between wheat yields and maximum temperatures during the growth period was observed in CK (R-adj(2) = 0.18, P < 0.01) and Straw (R-adj(2) = 0.14, P = 0.02) treatments. But this negative impact was not seen under biochar amendment. Notably, wheat yields increased with rising minimum winter temperatures and cumulative precipitation in the BC treatments. Conclusions: Continuous biochar addition can enhance soil fertility, sustainably increase wheat yields, and improve wheat resistance to unfavorable weather conditions.

With the widespread application of lithium-ion batteries in electric vehicles and energy storage systems, health monitoring and remaining useful life prediction have become critical components of battery management systems. To address the challenges posed by the high nonlinearity and long-term dependency in battery degradation modeling, this paper proposes a deep hybrid architecture that integrates Long Short-Term Memory networks with Transformer mechanisms, aiming to improve the accuracy and robustness of RUL prediction. Firstly, time-series samples are constructed from raw battery data, and physically consistent temperature-derived features—including average temperature, temperature range, and temperature fluctuation—are engineered. Data preprocessing is performed using standardization and Yeo-Johnson transformation. The model employs LSTM modules to capture local temporal patterns, while the Transformer modules extract global dependencies through multi-head self-attention mechanisms. These complementary features are fused to enable joint modeling of battery health states. The regression task is optimized using the Mean Squared Error loss function and trained with the Adam optimizer. Experimental results on the MIT battery dataset demonstrate the proposed model achieves excellent performance in a 7-step multi-point prediction task, with a Root Mean Square Error of 0.0085, Mean Absolute Percentage Error of 0.0200, and a coefficient of determination of 0.9902. Compared with alternative models such as MC-LSTM and XGBoost-LSTM, the proposed model exhibits superior accuracy and stability. Residual analysis and visualization further confirm the model’s unbiased and stable predictive capability. This study shows that the LSTM-Transformer hybrid architecture offers significant potential in modeling complex battery degradation processes and enhancing RUL prediction accuracy, providing effective technical support for the development of intelligent battery health management systems.

Accurately assessing the potential risk of cracks in photovoltaic (PV) panels is crucial for improving the system's energy conversion efficiency and safety. This paper develops a novel internal crack detection device for PV panels based on air-coupled ultrasonics and establishes a dedicated model for PV panel crack detection. Considering the anisotropy of silicon, the Rayleigh-Lamb equation is first used to obtain the dispersion curve of Lamb waves. The continuous wavelet transform (CWT) is then applied to analyze the A0 mode Lamb wave echo signals within a single cycle to distinguish between cracked and non-cracked regions. Further, the original signal is enhanced using singular value decomposition (SVD) to strengthen the signal features, and a direct relationship model between the signal echo and the crack length is proposed. Finally, the model's validity is verified by measuring actual crack data from PV panels. The results show that the optimal frequency of Lamb waves for crack detection in PV panels is 200 kHz, with an attenuation coefficient of 6.4 dB/mm and incidence and exit angles of 11 degrees. Compared to electroluminescence (EL) and inductive thermal imaging methods, the proposed approach achieves higher accuracy, reaching 84.26%.

The exceptional cycling stability of lithium-ion batteries in electric vehicles and large-scale grid energy storage applications necessitates the use of accelerated aging tests for rapid assessment. Overdischarge stress is an effective approach to accelerate battery aging, whereas its impact on solid electrolyte interphase (SEI) and battery aging performance remains elusive. Herein, the whole picture of SEI evolution under different overdischarge levels was quantitatively illustrated by combining the electrochemical analysis and spectrochemical techniques. Overdischarge leads to the decomposition of the organic components within SEI, such as ROCO2Li and CH3Li, while the damaged SEI is repaired during the subsequent charging process with its composition and structure reconstructed. Under overdischarge conditions, the SEI undergoes continuous cycles of destruction and repair, which suppresses its growth and evolution to inorganic components, resulting in a thinner and more uneven morphology with higher organic components and a lower Young's modulus. The unique SEI evolution mechanism of overdischarge effectively accelerates the loss of active lithium and exhibits similar thermodynamic degradation modes to normal aging, making overdischarge a potential accelerated aging method. This study provides a deeper understanding of the mechanisms behind accelerated aging in batteries and offers new insights into the evaluation and enhancement of battery performance.

Electrochemical energy storage systems are imperative to ensure the reliable operation of high-renewable energy grids. Battery reliability is a critical factor in the safety and cost-effectiveness of energy storage systems. The existing assessment methods are challenging to quantify the multidimensional degradation characteristics of batteries and ignore the influence of topology. In this paper, we propose a multi-state degradation reliability assessment model for energy storage batteries. This model is based on a multi-dimensional generic generating function, which is used to calculate the reliability level of the battery cell. This model helps to better manage and optimize the performance of the battery system through reliability level analysis. The model integrates various series-parallel structures of energy storage battery cells in diverse scenarios, thereby depicting the multiple states of battery clusters during the degradation process. It facilitates the acquisition of reliability grades for battery clusters that are fused with different series-parallel topologies. The experimental findings demonstrate the efficacy of the model in quantifying the reliability of battery cells. Additionally, the grading outcomes exhibit a substantial correlation with the actual degradation state of the batteries. This model offers a novel technical approach for predicting the lifespan and managing the health of energy storage batteries, with significant practical applications in enhancing system safety and economic efficiency.

Accurate estimation of lithium-ion battery capacity is essential for ensuring the reliability and safety of battery energy storage systems. This paper proposes an innovative online multi-time-scale capacity estimation method that combines improved Gaussian Process Regression with a dynamic calibration mechanism based on State of Charge turning points. The Whale Optimization Algorithm is employed to enhance the global search ability and predictive accuracy of the model. Additionally, a threshold calibration strategy based on the normal distribution of inflection points is introduced at the macroscopic time scale to correct estimation errors and quantify reliability. Experimental results indicate that the proposed method achieves an average relative error of less than 5 %, demonstrating superior stability and robustness compared to traditional approaches. The method not only improves the precision of battery capacity estimation but also provides a more reliable solution for battery health assessment, thereby extending battery life and enhancing overall performance.

Lithium-ion battery-powered mining trucks, as an emerging category in the field of heavy-duty electric commercial vehicles, have achieved cycle life exceeding thousands of cycles and a service lifespan of 5-8 years or more. However, the absence of systematic battery fault detection systems for electric mining trucks poses significant challenges, including safety risks, performance degradation, and resource waste caused by battery failures. To address these challenges, this study proposes an innovative method integrates convolutional neural networks (CNN), long short-term memory networks (LSTM), and dynamic autoencoder (DYAD)-based information entropy extraction. This method captures dynamic variations in battery states while addressing cumulative fault effects and nonlinear characteristics, achieving an area under the receiver operating characteristic curve (AUROC) of 88.7 %.The deep learning framework with an encoder-decoder architecture overcomes limitations in detecting subtle fault features. By analyzing reconstruction errors, the approach effectively mitigates false alarms caused by misjudged supercharging events. To enhance model interpretability, multimodal decoupling is introduced to monitor error variations in key features in real time, enabling early detection of latent battery faults. Global interpretability analysis decomposes multidimensional aging feature spaces, while latent space visualization reveals distinguishable characteristics of faulty batteries in energy dissipation dimensions, thereby improving model transparency. Experimental results demonstrate significant area under the receiver operating characteristic curve (AUROC) improvements of 26.2 % over Graph Dynamic Network GDN and 21.8 % over Autoencoder (AE), highlighting superior fault detection capabilities in complex environments. This work provides a robust solution for battery fault detection in electric mining trucks, aligning with industry standards and supporting sustainable development goals.

Accurate estimation of the State of Health (SOH) of lithium-ion batteries (LIBs) is of significant importance for the utilization of electrical devices powered by batteries, maintenance of battery energy storage equipment, and economic considerations for battery storage applications. Data-driven methods have gained increased attention due to their simple modeling and real-time learning capabilities. However, these methods have been constrained by cumbersome feature engineering and limited model transferability. Considering the emergence of time series foundation models, this study fine-tunes TimeGPT using cycling data from 143 LIBs with six different cathode materials to enhance the model's accuracy and generalization capability in SOH estimation. Experimental results demonstrate that the fine-tuned model can adapt to various LIBs and operating conditions, exhibiting strong adaptability and transferability. After 100 steps of fine-tuning, the model maintains an RMSE below 1.06% and MAE below 0.59% across different Test sets, achieving an average reduction of 21.55% in RMSE and 13.55% in MAE compared to zero-shot inference. This research treats LIBs SOH estimation as a time series predication downstream task, validates the feasibility of fine-tuning, and provides anew solution for developing next-generation battery management systems.

Growth on patterned graphene/Ge provides a route to improve the film quality of large mismatch heteroepitaxy and simultaneously facilitate the transfer of epitaxial films; however, the growth process and its associated technical challenges remain unclear. In this work, the molecular beam epitaxy (MBE) growth of CdTe films on micro-scale patterned strip-like graphene/Ge (100), containing selective area epitaxy (SAE) of CdTe seeds on exposed Ge and the merging process of CdTe seeds, were systematically investigated. The effects of growth temperature on the SAE of CdTe seeds were studied in detail, and a growth model of the CdTe seeds was proposed. Additionally, we examined the morphology and crystal quality of CdTe films at different growth stages, identifying the suppression of CdTe nucleation on graphene during the CdTe seed growth and merging as a key challenge to obtain high-quality films on patterned graphene/Ge.

Permanent magnet synchronous motors (PMSMs) are playing an increasing crucial role in industrial applications, where inaccurate system parameters can affect control performance and even threaten system stability. This paper aims to tackle the complex and cumbersome recognition algorithms for electrical and mechanical parameters of the PMSMs, which often result in long identification duration and low precision. We propose a full-parameter identification algorithm for PMSMs based on signal injection. This algorithm is simple and effective, easy to deploy, has a short identification duration, and high precision. It can achieve full parameter identification of resistance Rs, d-axis inductance Ld, q-axis inductance Lq, permanent magnet flux ψf, moment of inertia J, viscous damping coefficient Bm, and Coulomb friction coefficient Cm, contributing to the design and development of a high-performance PMSM servo control system. Firstly, this paper presents the identification algorithms for electrical and mechanical parameters. Then a simulation model is built using Maltab/Simulink. Finally, experiments are conducted on a 1.5kW built-in PMSM experimental platform. Both the simulation and experiments have verified the feasibility and effectiveness of the proposed method.

Perfect graphene substrates can improve the quality of films grown by large-mismatch heteroepitaxy and simultaneously enable the transfer of epitaxial layers. However, large-area chemical vapor deposition-grown graphene suitable for practical applications typically contains various defects, and the influence of these defects on heteroepitaxial growth remains unclear. In this work, we systematically investigated the molecular beam epitaxy growth of CdTe films on graphene/Ge substrates. The effects of graphene defects on the nucleation behavior and crystal structure of CdTe grains were elucidated, and epitaxial models for CdTe on three types of graphene/Ge surfaces were proposed. Furthermore, we evaluated the crystal quality of CdTe films grown on different graphene/Ge substrates and found that the quality of the epitaxial films is strongly correlated with the defects in graphene.

VES is a method of balancing the energy of a power system with other equipment or scheduling strategies, particularly with respect to controllable loads, owing to end-user electrification. This paper summarises the connotations, classifications, and typical modelling applications for VES users. Thereafter, the modelling methods, characteristics, and specific operation cases of five types of VESs are introduced, including electric vehicles, buildings, cold storage, industrial production and hydrogen storage. Furthermore, the energy storage capacity planning, energy scheduling strategy, and power control strategy of a VESS are realised through optimal control strategies. Finally, in conjunction with demand response, the development prospects of VES in modelling and control strategies are discussed to improve the economic and environmental benefits of microgrids.

Dust deposition on photovoltaic (PV) panels significantly reduces light transmittance and power conversion efficiency. Therefore, real-time dust detection systems are crucial for proactive cleaning and maintenance to improve light absorption and the operational efficiency of PV systems. This paper developed an end-to-end PV dust detection model, DVNET, based on light transmittance estimation. The model quantifies the dust density on PV panels using image processing to estimate light transmittance and determine optimal cleaning strategies. The DVNET architecture captures the mapping relationship between light transmittance and dust density. The model calculates the light transmittance of dusty images, enabling a quantitative assessment of dust deposition. Images are acquired and preprocessed to eliminate noise and irrelevant information and obtain optimal model training conditions. Experiments are conducted to evaluate model performance. The results indicate that the proposed model accurately estimates light transmittance using images of PV panels and generates transmittance maps to visualize the dust distribution on the panel surfaces. DVNET outperforms five benchmark models, achieving the lowest mean square error (MSE) of 0.00044. The performances of the squeeze-and-excitation block (SEBlock) module, the convolutional block attention module (CBAM), and the multiply-add-combine (MAC) attention mechanism are evaluated. The experimentally determined relative error (RE) is below 0.03 for uniform and non-uniform transmittance. The observed and predicted correlations between dust deposition and transmittance are highly similar for PV panels with different dust densities, demonstrating the model's applicability in real-world scenarios. The experiment demonstrated that the DVNET model exhibits low computational resource consumption, fast training time, as well as remarkable scalability and computational performance on large-scale datasets.

Two-dimensional transition metal dichalcogenides (TMDs) with piezoelectric effects are ideal materials for future wearable devices. While enhancing the piezoelectric performance by forming vertical heterojunctions, shortcomings such as contamination at the heterojunction interface and limited built-in electric field width have been noticed. In this work, a lateral heterojunction of monolayer WSe2-MoSe2 with type-II band alignment was employed to amplify the electromechanical optoelectronic efficiency. The considerable built-in field width (BFW) in the lateral heterojunction facilitates rapid separation of carriers. The lattice mismatch induced a flexoelectric effect during the lateral heterojunction growth. The flexoelectric and piezoelectric effects under external strain can regulate the photodetector performance of the device. Under the compressive strain of -0.93%, the photocurrent increased 9.1 times compared to the tensile strain of 0.47%. Flexoelectric effect can reduce the dark current under no external strain. This work reveals the roles of flexoelectric and piezoelectric effects in enhancing photoelectric conversion, suggesting lateral heterojunction devices may be applied in the field of flexible low-light detection.