Zheng Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zheng Liu has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
Chunshan Yang
School of Artificial Intelligence Guilin University of Aerospace Technology Guilin China
5 shared publicationsBenqin Jing
4 shared publications- Xuanju Dang
School of Electronic Engineering and Automation, Guilin University of Electronic Technology, Guilin, China
4 shared publications Yuan Qiu
4 shared publicationsJianbo Ji
School of Electronic Information and Automation, Guilin University of Aerospace Technology, Guilin, China
3 shared publicationsZhenhua Zhao
3 shared publications
Latest Publications
- 26 May 2026
- Published by Wiley in Energy Technology
- Vol. 14 (5)
A Hybrid Framework for Lithium‐Ion Battery State‐of‐Health Estimation via Dynamic Local–Global Feature Fusion
Accurate estimation of the state of health (SOH) is essential for ensuring the safety and reliability of lithium‐ion batteries. Conventional data‐driven approaches often struggle to simultaneously capture long‐term temporal dependencies and dynamically adapt to evolving degradation behaviors. To address this limitation, a hybrid deep learning framework, termed TCN–SE–TE, is developed. The framework integrates a temporal convolutional network (TCN), a squeeze‐and‐excitation (SE) attention module, and a transformer encoder (TE). Based on experimental cycling data, four health indicators (HIs) highly correlated with SOH are extracted through data preprocessing combined with Pearson correlation analysis. The TCN captures local degradation dynamics between adjacent cycles, while the SE module adaptively adjusts feature channel weights to emphasize dominant HIs at different aging stages. The TE further models global temporal dependencies to characterize overall degradation trends. Evaluated using leave‐one‐out cross‐validation on the University of Maryland CS2 battery dataset, the proposed framework effectively combines local–global feature fusion and dynamic feature attention to enhance the adaptability and stability of SOH estimation across different degradation stages.
- 26 May 2026
- Published by Wiley in Energy Technology
- Vol. 14 (5)
A Hybrid Framework for Lithium‐Ion Battery State‐of‐Health Estimation via Dynamic Local–Global Feature Fusion
Accurate estimation of the state of health (SOH) is essential for ensuring the safety and reliability of lithium‐ion batteries. Conventional data‐driven approaches often struggle to simultaneously capture long‐term temporal dependencies and dynamically adapt to evolving degradation behaviors. To address this limitation, a hybrid deep learning framework, termed TCN–SE–TE, is developed. The framework integrates a temporal convolutional network (TCN), a squeeze‐and‐excitation (SE) attention module, and a transformer encoder (TE). Based on experimental cycling data, four health indicators (HIs) highly correlated with SOH are extracted through data preprocessing combined with Pearson correlation analysis. The TCN captures local degradation dynamics between adjacent cycles, while the SE module adaptively adjusts feature channel weights to emphasize dominant HIs at different aging stages. The TE further models global temporal dependencies to characterize overall degradation trends. Evaluated using leave‐one‐out cross‐validation on the University of Maryland CS2 battery dataset, the proposed framework effectively combines local–global feature fusion and dynamic feature attention to enhance the adaptability and stability of SOH estimation across different degradation stages.
- 17 April 2026
- Published by Wiley in Energy Technology
- Vol. 14 (4) , e202502662
Enhanced State of Charge Estimation for Battery via Adaptive Wavelet‐Based Hybrid Deep Learning
Accurate state of charge (SOC) estimation is essential for lithium‐ion battery management systems. However, the accuracy of SOC estimation can be significantly degraded by unknown noise under complex operating conditions. According to this issue, an SOC estimation method that integrates an adaptive wavelet and a convolutional neural network (CNN)–long short‐term memory (LSTM) hybrid neural network is proposed. First, we have designed a locally adaptive signal‐to‐noise ratio‐based dynamic wavelet threshold algorithm. By optimizing the decomposition level and adopting an improved semi‐soft threshold function, this algorithm can effectively suppress noise while preserving the signal features. Furthermore, a CNN–LSTM hybrid model is constructed. The CNN is used to extract the local features of the battery data, and the LSTM is employed to model the long‐term dependencies of the battery data, thereby enhancing the model's robustness against noise. The experimental results show that this method can suppress the influence of various types of noise in multiple working conditions. The errors in this method are lower than those of traditional methods, and the R2 values are all above 0.95, indicating a high degree of fitting. This study provides a reliable solution for estimating the battery's SOC in complex noise environments.
- 17 April 2026
- Published by Wiley in Energy Technology
- Vol. 14 (4) , e202502662
Enhanced State of Charge Estimation for Battery via Adaptive Wavelet‐Based Hybrid Deep Learning
Accurate state of charge (SOC) estimation is essential for lithium‐ion battery management systems. However, the accuracy of SOC estimation can be significantly degraded by unknown noise under complex operating conditions. According to this issue, an SOC estimation method that integrates an adaptive wavelet and a convolutional neural network (CNN)–long short‐term memory (LSTM) hybrid neural network is proposed. First, we have designed a locally adaptive signal‐to‐noise ratio‐based dynamic wavelet threshold algorithm. By optimizing the decomposition level and adopting an improved semi‐soft threshold function, this algorithm can effectively suppress noise while preserving the signal features. Furthermore, a CNN–LSTM hybrid model is constructed. The CNN is used to extract the local features of the battery data, and the LSTM is employed to model the long‐term dependencies of the battery data, thereby enhancing the model's robustness against noise. The experimental results show that this method can suppress the influence of various types of noise in multiple working conditions. The errors in this method are lower than those of traditional methods, and the R2 values are all above 0.95, indicating a high degree of fitting. This study provides a reliable solution for estimating the battery's SOC in complex noise environments.
- Open Access
- 19 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 42068-42082
Novel Adversarial Unsupervised Subdomain Adaption Multi-Channel Deep Convolutional Network for Cross-Operating Fault Diagnosis of Rolling Bearings
Rolling bearings in production practice usually serve in a healthy state. Some fault state labels are scarce or even no labels, resulting in unbalanced data categories. Meanwhile, frequent working condition switching results in significant differences in data distribution among working conditions, and labeled data in some working states cannot be fully utilized. To deal with the challenge of low fault identification accuracy caused by these practical factors, this paper proposed a novel adversarial unsupervised subdomain adaption multi-channel deep convolutional network (ASMDCN). Firstly, a parallel three-channel depth feature extraction module is built, and a multi-scale convolution kernel is used to fully extract the rich features of vibration signals under various working conditions. Secondly, a novel loss function is designed to adequately consider the classification difficulty of samples and the degree of class imbalance. Finally, the adversarial training strategy is used to force the feature extractor to extract the domain invariant features, and the Local Maximum Mean discrepancy (LMMD) is used to align the global and related subdomains of the source and target domains. The experimental results show that the designed feature extraction can fully extract the domain-invariant features of the rolling bearings under different working conditions. Under the proposed objective function optimization, the network model can fully align the features of multi-source and single-target domain under unbalanced data and has strong generalization performance.
- 19 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 42068-42082
Open AccessNovel Adversarial Unsupervised Subdomain Adaption Multi-Channel Deep Convolutional Network for Cross-Operating Fault Diagnosis of Rolling Bearings
Rolling bearings in production practice usually serve in a healthy state. Some fault state labels are scarce or even no labels, resulting in unbalanced data categories. Meanwhile, frequent working condition switching results in significant differences in data distribution among working conditions, and labeled data in some working states cannot be fully utilized. To deal with the challenge of low fault identification accuracy caused by these practical factors, this paper proposed a novel adversarial unsupervised subdomain adaption multi-channel deep convolutional network (ASMDCN). Firstly, a parallel three-channel depth feature extraction module is built, and a multi-scale convolution kernel is used to fully extract the rich features of vibration signals under various working conditions. Secondly, a novel loss function is designed to adequately consider the classification difficulty of samples and the degree of class imbalance. Finally, the adversarial training strategy is used to force the feature extractor to extract the domain invariant features, and the Local Maximum Mean discrepancy (LMMD) is used to align the global and related subdomains of the source and target domains. The experimental results show that the designed feature extraction can fully extract the domain-invariant features of the rolling bearings under different working conditions. Under the proposed objective function optimization, the network model can fully align the features of multi-source and single-target domain under unbalanced data and has strong generalization performance.
- 19 October 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (1) , 015905
Novel imbalanced subdomain adaption multiscale convolutional network for cross-domain unsupervised fault diagnosis of rolling bearings
Data on the vibration signals collected from rolling bearings mostly belongs to health conditions, leading to an imbalanced data distribution. In addition, frequent switching of operating conditions results in unlabeled data collected under a specific working condition. This paper proposes a novel network for cross-domain unsupervised fault diagnosis of rolling bearings considering the imbalanced data to address these challenges. First, a multiscale parallel features extraction is developed, which can fully mine the rich high-level feature representation of various fault types from the original data and has a high value for fault identification. Second, a squeeze-and-excitation attention mechanism is constructed to enhance features conducive to model classification and suppress redundant features. Finally, a new loss function is proposed to optimize the model, which can accurately classify imbalanced source domain and easily align related subdomains of two domains. The proposed method was validated on multiple unsupervised cross-domain diagnostic tasks on two bearing datasets. Experimental results manifest that the proposed method has stable generalization performance and excellent robustness.
- 19 October 2023
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (1) , 015905
Novel imbalanced subdomain adaption multiscale convolutional network for cross-domain unsupervised fault diagnosis of rolling bearings
Data on the vibration signals collected from rolling bearings mostly belongs to health conditions, leading to an imbalanced data distribution. In addition, frequent switching of operating conditions results in unlabeled data collected under a specific working condition. This paper proposes a novel network for cross-domain unsupervised fault diagnosis of rolling bearings considering the imbalanced data to address these challenges. First, a multiscale parallel features extraction is developed, which can fully mine the rich high-level feature representation of various fault types from the original data and has a high value for fault identification. Second, a squeeze-and-excitation attention mechanism is constructed to enhance features conducive to model classification and suppress redundant features. Finally, a new loss function is proposed to optimize the model, which can accurately classify imbalanced source domain and easily align related subdomains of two domains. The proposed method was validated on multiple unsupervised cross-domain diagnostic tasks on two bearing datasets. Experimental results manifest that the proposed method has stable generalization performance and excellent robustness.
- Open Access
Torque Ripple Suppression of Switched Reluctance Motor with Reference Torque Online Correction
High torque ripple dramatically affects the switched reluctance motor (SRM) application. To reduce the torque ripple, a reference torque neural network (RTNN) is proposed to adjust the reference torque online. Firstly, the RTNN is built on the torque sharing function (TSF) method. Furthermore, the RTNN is designed as a single-input and -output network. As the periodic relationship between the torque ripple and the rotor angle, the rotor angle constitutes the central node parameter of the implicit function in RTNN. Therefore, one-step adjustment of the RTNN can perform well at restraining reference torque. Lastly, the torque error is used to adjust the parameters of RTNN to reduce the torque ripple. In the MATLAB environment, through the simulation comparison with fuzzy torque and PD current compensation method, the effectiveness of RTNN at torque ripple suppression is proven with different loads and speeds.
Open AccessTorque Ripple Suppression of Switched Reluctance Motor with Reference Torque Online Correction
High torque ripple dramatically affects the switched reluctance motor (SRM) application. To reduce the torque ripple, a reference torque neural network (RTNN) is proposed to adjust the reference torque online. Firstly, the RTNN is built on the torque sharing function (TSF) method. Furthermore, the RTNN is designed as a single-input and -output network. As the periodic relationship between the torque ripple and the rotor angle, the rotor angle constitutes the central node parameter of the implicit function in RTNN. Therefore, one-step adjustment of the RTNN can perform well at restraining reference torque. Lastly, the torque error is used to adjust the parameters of RTNN to reduce the torque ripple. In the MATLAB environment, through the simulation comparison with fuzzy torque and PD current compensation method, the effectiveness of RTNN at torque ripple suppression is proven with different loads and speeds.
- 13 December 2021
- Published by Wiley in Advanced Theory and Simulations
- Vol. 5 (3) , 2100331
A Combined State of Charge Estimation Method for Lithium‐Ion Batteries Using Cubature Kalman Filter and Least Square with Gradient Correction
Reliable state of charge (SOC) is the core of the energy storage system for electric vehicles. An efficient method of parameter identification and SOC estimation for lithium‐ion batteries (LIBs) using cubature Kalman filter (CKF) and least square with gradient correction is proposed. Firstly, the recursive gradient correction (RGC) strategy is introduced based on the recursive least square with forgetting factor (FFRLS) method, and a novel way is proposed for a simplified equivalent circuit model (ECM) based LIBs parameter identification using the combination of RGC and FFRLS method. Then, on the basis of CKF, the singular value decomposition (SVD) is used instead of the Cholesky decomposition to solve the non‐positive definiteness matrix for cubature transformation in the priori estimated covariance, thereby improving the effect of SOC estimation. Finally, the dynamic stress test (DST) and the federal urban driving schedule (FUDS) are used to verify the effect of the proposed method. The test results reflect that the combination of RGCFFRLS‐SVDCKF method and simplified ECM have advantages in terms of LIBs SOC estimation convergence ability and tracking accuracy, and the SOC estimation error can converge to 1% and 1.5% error boundary when the initial SOC are 0.5 and 0.8, respectively. In addition, the robustness of the proposed method is verified. Compared with the FFRLS‐EKF method, the proposed RGCFFRLS‐SVDCKF method has better precision and robustness to noise disturbance.
- 13 December 2021
- Published by Wiley in Advanced Theory and Simulations
- Vol. 5 (3) , 2100331
A Combined State of Charge Estimation Method for Lithium‐Ion Batteries Using Cubature Kalman Filter and Least Square with Gradient Correction
Reliable state of charge (SOC) is the core of the energy storage system for electric vehicles. An efficient method of parameter identification and SOC estimation for lithium‐ion batteries (LIBs) using cubature Kalman filter (CKF) and least square with gradient correction is proposed. Firstly, the recursive gradient correction (RGC) strategy is introduced based on the recursive least square with forgetting factor (FFRLS) method, and a novel way is proposed for a simplified equivalent circuit model (ECM) based LIBs parameter identification using the combination of RGC and FFRLS method. Then, on the basis of CKF, the singular value decomposition (SVD) is used instead of the Cholesky decomposition to solve the non‐positive definiteness matrix for cubature transformation in the priori estimated covariance, thereby improving the effect of SOC estimation. Finally, the dynamic stress test (DST) and the federal urban driving schedule (FUDS) are used to verify the effect of the proposed method. The test results reflect that the combination of RGCFFRLS‐SVDCKF method and simplified ECM have advantages in terms of LIBs SOC estimation convergence ability and tracking accuracy, and the SOC estimation error can converge to 1% and 1.5% error boundary when the initial SOC are 0.5 and 0.8, respectively. In addition, the robustness of the proposed method is verified. Compared with the FFRLS‐EKF method, the proposed RGCFFRLS‐SVDCKF method has better precision and robustness to noise disturbance.
- 30 August 2021
- Published by Springer Nature in Journal of Electrical Engineering & Technology
- Vol. 17 (1) , 197-208
A Parameter Identification Method for Lithium-Ion Batteries Using Simplified Impedance Model and Fractional Order Kalman Filter
- 30 August 2021
- Published by Springer Nature in Journal of Electrical Engineering & Technology
- Vol. 17 (1) , 197-208
A Parameter Identification Method for Lithium-Ion Batteries Using Simplified Impedance Model and Fractional Order Kalman Filter