The security issues of cyber-physical systems (CPSs) have received widespread attention because they are vulnerable. Opacity is an important security property used to measure whether secrets are leaked to intruders. Existing works on opacity mostly focus on discrete event systems (DESs) and qualitative analysis. In order to quantitatively analyze the opacity in continuous state systems (i.e., the considered CPSs), this paper presents a new approximate opacity framework and proposes a multi-system T-delta approximate initial state opacity (MS-T-delta-AISO). Further, to improve the MS-T-delta-AISO, the fully cooperative game approximate initial state opacity enforcement (FCG-AISOE) is studied. Reinforcement Q-learning is used to optimize the MS-T-delta-AISO, which can achieve rapid iteration of control policy and converge to the optimal control policy, resulting in a rapid increase in opacity and ultimately tending towards the optimal opacity. Finally, simulation examples are given to verify the effectiveness of the proposed method.

Accurate and reliable estimation of lithium-ion battery state of health (SOH) is essential for battery management systems (BMSs), directly affecting safety, lifetime prediction, and energy efficiency. Although data-driven models have achieved promising accuracy, their practical deployment is often limited by insufficient physical consistency, weak cross-chemistry generalization, and poor interpretability. In this work, we propose a physicsinformed unified network (PIUN) for single-step SOH estimation, which integrates physically consistent constraints into a compact gated backbone termed the Update-Candidate Branch Network (UCBNet). By jointly enforcing data fidelity and degradation-aware physical regularization, PIUN captures both static health indicators and dynamic aging behaviors within a unified framework. The PIUN is evaluated on four public battery datasets (XJTU, TJU, MIT, and HUST). Across 11 evaluation subsets, PIUN reduces MAPE and RMSE by an average of 11.5% and 7.4%, respectively, compared with the strongest baseline. Under noise-augmented training, the average improvement further increases to approximately 16%, demonstrating enhanced robustness to measurement perturbations. Cross-dataset experiments additionally confirm superior zero-shot and fewshot generalization capability. To improve interpretability, SHAP analysis is conducted and reveals strong consistency between learned feature contributions and electrochemical degradation mechanisms. Guided by SHAP-derived importance within interpretable physical dimensions, a compact subset of 7 representative features is constructed. Experimental validation demonstrates performance comparable to the full 17-feature model, thereby experimentally confirming the reliability of the SHAP-based interpretations while enabling lightweight deployment. Overall, PIUN provides a physically consistent, interpretable, and generalizable SOH estimation framework with strong practical potential for real-world BMS applications.

This paper is concerned with the consensus of multi-agent systems (MASs) in the presence of fading channels and limited communication bandwidth. These issues will lead to inaccurate and energy constraints of communication data respectively, which will not be possible for MASs to achieve consensus. To this end, this paper proposes a variable structure encoding-decoding scheme (VS-EDS), which includes an encoder with a unitary signal generator and a corresponding decoder with an adaptive activation module. The advantages of the proposed VS-EDS are mainly in three aspects. First, it avoids the need for existing methods to perform a large amount of statistical work on the statistics of fading gains before the system runs. Second, it can eliminate the communication errors caused by fading. At the same time, it can adapt to the changing communication network environment. In addition, in order to achieve consensus control, a distributed protocol based on the VS-EDS is designed. Finally, the feasibility and effectiveness of the VS-EDS are demonstrated through comparative simulations.

In this paper, a resilient adaptive covariance Kalman filter is developed for state estimation under false data injection attack (FDIA) during the process of measurements transmission. The extreme measurement deviation caused by unknown injection vectors is clipped by an adaptive saturation function, and an adaptive noise covariance matrix triggered by prediction residual is constructed to enhance the estimation performance and stability of the filtering error system under FDIA. To analyze the asymptotic convergence of the algorithm, the error expression is constructed to analyze the upper limit of prediction error. Finally, a simulation experiment on an inverted pendulum car verifies the stability and effectiveness of the proposed method in reducing the impact of unknown attack vectors.

Aiming at the problem that a large number of electric vehicles randomly connected to the grid poses a huge challenge to the security of the power grid, this paper proposes a strategy to guide the orderly charging of electric vehicles by using the time-of-use electricity price policy. Firstly, an orderly charging scheduling model for electric vehicles taking into account the response level to the policy is constructed. Then, a hybrid algorithm combining Spider Wasp Optimization (SWO) and Particle Swarm Optimization (PSO) is used to optimize the peak-valley electricity price period. Finally, by using the Monte Carlo and probability statistics theory methods to simulate the daily charging load of electric vehicles, the experiment of different response level are carried out. And results of different optimization methods for solving the scheduling model are compared. Comparison results show that the proposed method achieves the smallest peak to valley difference with the lest iterations. The proposed method can provides an effective strategy for peak shaving and valley filling.

Taro production is predominantly manual, creating a need for automated defect detection in processing lines to improve efficiency and product quality. Traditional computer vision methods lack accuracy under practical industrial production conditions, while deep learning approaches, though robust, are often computationally expensive. This study presents a modified YOLOv8n model designed for automated defect detection of taro strips in industrial production environments where real-time processing and high accuracy are critical. The proposed model integrates several architectural innovations: a Bi-directional Feature Pyramid Network (BiFPN) to facilitate improved multi-scale feature fusion, the VoV-GSCSP module to replace the conventional C2f block and reduce computational complexity, and a shared parameter detection head to further lighten the model. In addition, an embedded Wise Intersection over Union (WIoU) loss function is used to accelerate convergence and improve prediction accuracy by optimizing bounding box alignment with ground truth data. Multiple data augmentation strategies, such as noise, motion, lighting, and background, are used in training to mitigate the risk of overfitting and enhance model scalability. Both offline and online experimental results demonstrate that the optimized model achieves an average mean detection accuracy (mAP50) of over 99%, with a precision and recall of over 0.99, and 3.23 G FLOPs. These results are significantly better than those of the original YOLOv8n, which achieved an mAP50 of 94.63%, a precision of 0.955, and a recall of 0.934. When deployed on a Raspberry Pi 5, the modified model demonstrates good robustness and generalization by accurately detecting defects in previously unseen taro-strip data. These results clearly show the advantages of the proposed method in terms of accuracy and computational efficiency, making it ideal for real-time taro-strip defect detection.

The operational stability of small hydropower stations is crucial to ensure grid security requirements when renewable energy is connected to the grid. However, due to the complex nonlinearity of the hydropower station system, and the strong random external hydrological and meteorological disturbances such as water inflow and reservoir evaporation, the current single-loop control strategy based on proportional-integral-differential control has poor control effect, resulting in continuous and large fluctuations in reservoir water level, and even exceeding the safe water level. To address this problem, this paper proposes a dual-loop predictive control strategy based on mechanism-data driven. This strategy decouples the control task structure into two collaborative loops. While, the outer loop is a forward-looking optimization decision-making layer which is used to actively absorb the impact of disturbances and maintain water level. The outer loop is deployed based on the BI-GRU network to accurately predict the dynamic evolution trend of the water level and solve the optimal unit target power online. The inner loop is a fast execution layer, which is based on the mechanism model to track the target power with high fidelity and response speed by optimizing the parameters of the original underlying proportional-integral controller. The experimental results on a real hydropower station show that under the continuous injection of composite random disturbance flow, this strategy successfully reduced the water level fluctuation within an extremely narrow range of ±0.015 meters above and below the target value (4.8m), effectively solving the problem of large water level fluctuations.

This paper proposes a leader-follower control method for multiple snake robot formation. Based on the simplified snake robot model, this work improves the traditional Serpenoid gait mode to a time-varying frequency form. Combined with the line-of-sight (LOS) method, a snake robot trajectory tracking controller is designed to enable the leader to track the desired trajectory at the ideal velocity. Then, the leader-follower following error system of a snake robot formation is established. In this framework, the follower can maintain a preset geometric position relationship with the leader to ensure the fast convergence of the formation location. Lyapunov’s theory proves the stability of a snake robot formation error. Simulation and experimental results show that this strategy has the advantages of faster convergence speed and higher tracking accuracy than other current methods.

Solving the problem of water level fluctuations in small hydropower station systems is challenging under traditional industrial control methods. This difficulty arises from the system’s high nonlinearity and the complexities involved in mechanism modeling. To address this, an improved neuro-fuzzy approach is proposed. In which, the multi-head attention mechanism based long short-term memory network is used to describe complex water level change patterns, and the fuzzy controller is introduced to dynamically adjust the control parameters to reduce water level fluctuation. Simulation-based on real hydropower station system data is carried out, and the superiority of the improved model under complex dynamic conditions is verified by comparing the prediction accuracy of different neural network methods and the effects of fuzzy controller and traditional PID control.

To address the issues of optimizing the operation of island electric-hydrogen coupled systems and high curtailment rates, a dispatch optimization strategy considering refined hydrogen energy modeling is proposed. The article analyzes the operational characteristics and electrolysis efficiency of alkaline electrolyzers and develops a refined model, balancing the economic and reliability aspects of the integrated energy system. An island integrated energy dispatch operation model is established. Case study results show that, compared to traditional electrolyzer models, the proposed model effectively reduces the number of start-stop cycles, extends the lifespan of the electrolyzer array, and enhances system economics.

Wind power forecasting is an effective way to reasonably schedule wind power generation and ensure the stable operation of power systems. However, the impact of physical attribute data related to wind power on forecasting varies, and long-term sequences of original features often contain redundant information and noise, making wind power forecasting a challenging task. To address this, a wind power prediction model based on CNN-BiLSTM (KG-RCBM) is proposed. This model uses Convolutional Neural Network (CNN) for short-term feature extraction to obtain local high-dimensional features, which are then analyzed by Bidirectional Long Short-Term Memory Network (BiLSTM) to capture the long-term trends of these local high-dimensional features and more comprehensive sequence information. This approach effectively reduces inaccuracies caused by the mixing of original data. The attention mechanism is utilized to dynamically allocate weights to the output data, addressing the issue of the model's inability to distinguish the importance of different data. The RIME algorithm is used to optimize the hyperparameters corresponding to each type of prediction model after the Gaussian Mixture Mode (GMM) classification, enabling adaptive wind power forecasting. Finally, Wind power data from Galicia, Spain, 2019, as well as hourly updated climate data from NASA, were utilized to validate our proposed wind power forecasting model. Compared with seven other models, the proposed model's prediction accuracy improves by 2.32% to 4.27%, indicating a higher forecasting precision.

The study is dedicated to the H∞ filtering of switched networks in the conditions of hybrid cyber attacks, the main types of which are deceptive attacks and denial-of-service (DoS) attacks. Two key assumptions are first made for the analysis the attack frequency follows a Bernoulli random distribution, and the switched signal as well as the measured signal are transmitted through a single data packet over the network. On this basis, a mathematical model is constructed for switched systems in the presence of hybrid cyber attacks. In this model, both the switched signal and measured output signal may encounter loss or tampering during data transmission. Subsequently, a switched filter is designed. A novel switching Lyapunov function is constructed by incorporating the random characteristics inherent in the switched signal. Based on this, the linear matrix inequality (LMI) method is embraced to come up with the adequate requirements that ensure the mean-square exponential stability of the filter error system as well as the desired H 8 performance index of the system. The outputs of the numerical simulations confirm that the designed H -filter can significantly reduce the packet loss issue resulting because of cyber attacks.

The cascaded H-bridge inverter exhibits the characteristics of high voltage, large capacity and low harmonic distortion, and has a vital impact on application fields such as battery energy storage and photovoltaic power generation. Fault diagnosis of inverter switches is essential to enhancing equipment dependability. However, the limited number of fault samples and severe overlap of fault signals in real-word applications present difficulties for inverter fault diagnosis. In view of this, this paper introduces a hierarchical classification fault diagnosis strategy founded on an improved siamese network to achieve high-precision fault diagnosis. Firstly, for the purpose of addressing the issues of multiple fault categories and limited samples, an improved Siamese network based on long short-term memory and attention mechanism is proposed to extract more subtle fault difference features, thereby improving the recognition accuracy of overlapping fault classes. Then, to solve the problem of serious overlap of fault samples of different types in the preliminary grouping, a hierarchical fault diagnosis model is proposed to realize high precision fault diagnosis. Finally, the fault data of the cascaded H-bridge inverter was obtained through the semi-physical simulation platform to complete the diagnosis experiment. The experimental results demonstrate the recommended model offers clear benefits in terms of diagnostic accuracy when compared to the conventional model.

In the field of precision manufacturing, error compensation of parts is the key to improve product quality and manufacturing efficiency. This paper presents a Long Short-Term Memory Network (LSTM) model based on the Gray Wolf optimization algorithm designed to optimize part error compensation. First, we introduce the sources of part errors and their impact on the manufacturing process. Then, we elaborate the application of LSTM network in predicting and compensating part errors by selecting appropriate features through correlation analysis. Through experiments, we verify the effectiveness of the Gray Wolf optimization-based LSTM model in part error prediction and compensation. The experimental results show that compared with the traditional method, the model in this paper has a significant improvement in both error prediction accuracy and compensation efficiency.

Anoectochilus roxburghii (Wall.) Lindl. (Orchidaceae) is one of the most precious Chinese medicine with extraordinary effects in medical treatment and health protection. Planting and tissue-cultured are two main cultivated methods of A. roxburghii. There are slight characteristic differences between Planting and tissue-cultured A. roxburghii, but they show significant differences in medicinal and market value. Therefore, the identification of cultivated methods plays an important role in effectively securing the medicinal efficacy of A. roxburghii and maintaining a good market order. However, due to the influence of composite differences such as different cultivars, different geographical origins and different times of cultivation, the difficulty and complexity of identification in cultivated methods increase heavily. This paper proposes an effective model to discriminative different cultivated methods of A. roxburghii based on improved 1D-inception-CNN. The experiments were conducted on two kinds of A. roxburghii, and their NIRS data were collected by a Fourier transform near-infrared spectrometer. Considering the unbalanced proportion of planting and tissue-cultured samples,the NIRS data was over sampled by using SMOTE first. Secondly, a one-dimensional convolutional neural network based on improved Inception was constructed to identify planting and tissue-cultured A. roxburghii though both include different varieties, different geographical origins and different cultivating times. Finally, Bayesian optimization was used to optimize the hyperparameters of the model. The final average identification accuracy, precision, recall, and F1-score of five-fold crossvalidation reached 97.95%, 96.16%, 100%, and 98.02%. The identification model proposed in this experiment provides a useful method to identify planting and tissue-cultured A. roxburghii effectively and rapidly and provides an idea for the identification of cultivation methods of other Chinese herbal medicines.

The inverter is one of the most important components in photovoltaic and wind power generation systems, and its stability is crucial to the smooth operation of the system. Power devices are the most fragile components in the inverter. Fault diagnosis and timely processing can greatly improve the reliability of the power generation system. Existing data-driven fault diagnosis methods are designed based on fixed working conditions. Once the system parameters change, the diagnosis accuracy will significantly decrease. To solve these problems, this study proposes a three-phase inverter open circuit fault diagnosis method based on domain adversarial neural network. This method selects the three-phase inverter phase voltage as the input signal, improves the convolutional neural network through the Inception structure, and then uses the domain adversarial neural network to learn domain invariant features. Finally, the diagnosis results are obtained based on the output of the fault classifier. Experimental results show that in transfer diagnosis tasks across different systems, the method achieves an average diagnosis accuracy of 95.01% and exhibits robustness in various noisy environments.

Anoectochilus roxburghii from different origins has different nutritional content and different price. Achieving origin identification is of great significance to the development and standardization of the Anoectochilus roxburghii industry However, it is influenced by complex factors, including the specific strain and growth environment. Achieving a high accuracy in origin identification presents significant challenges and may not always meet the stringent requirements. To account for the unique characteristics of Anoectochilus roxburghii dataset from different origins, such as limited sample size, imbalanced samples, and numerous sample interference factors, a method based on improved SMOTE and CatBoost is designed to address the need for precise origin identification. First, a Fourier transform near infrared spectrometer was used to collect sample information of Anoectochilus roxburghii from three different origins, and then the improved SMOTE algorithm was used to balance the dataset. Finally, CatBoost classifier was used to identify the different origins. Comparative experimental results show that the method proposed in this article has the highest identification accuracy, reaching more than 97%, which is 6.9% and 2.8% higher than using original data and the original SMOTE algorithm respectively. The model constructed can efficiently identify Anoectochilus roxburghii of different origins and can be served as a useful reference for quality supervision of Anoectochilus roxburghii.

This paper studies the problem of state estimation in wireless sensor networks affected by false data injection attack. An event-triggered mechanism-based diffusion Kalman filter algorithm is proposed to solve this problem. Firstly, an event-triggered mechanism based on saturation function is designed to reduce the sampling rate of nodes and clipping abnormal measurements. To maintain the stability of node estimation and prevent the diffusion of contaminated data between nodes, a data transmission mechanism based on local outlier factor is designed to control the data exchange between nodes. Simulation results show that the designed algorithm improves the estimation performance in the case of false data injection attack, the communication rate between nodes and the transmission of contaminated data are reduced.

Potential malicious attacks have been a significant security concern for network system applications. However, there are few studies on filtering for hybrid network attacks in switching systems. This paper considers a Kalman filtering problem for the switched systems that suffer from deception attacks and denial-of-service attacks. A new network transmission model for switching systems is established. Then, based on the minimum mean square error criterion, a Kalman filter with low conservatism is designed for the discrete-time switched system. The newly switched Kalman gain matrix is deduced including the random variation of the switching signal after being attacked by the network. Finally, the effectiveness of the proposed filter is verified by the numerical simulation. Note to Practitioners —A switching system is considered to be a typical hybrid system. As it often works in a network environment, the switching signal is vulnerable to the network and thus to various cyber-attacks (e.g. spoofing attacks and denial-of-service attacks). Few studies have been conducted on the impact of switching signals in network transmission. To address this problem, this paper proposes a Kalman filter design method with low conservativeness, which describes network attacks and data loss by building a Kalman filter model associated with the switching signal in terms of satisfying Bernoulli random variables. For practitioners, the ability to recover data to a certain extent when the data transmission is affected by different network attacks will save a lot of costs and is important for improving the stability and performance of switching systems. Our future work will aim to improve the performance of different filtering algorithms under different cyber-attacks.

Advanced wind power prediction technique plays an essential role in the stable operation of the grid with largescale grid integration of wind power. Most research focuses on distance-based static classification where the subjective nature of initial center selection increases the uncertainty of the prediction. And the data classification on a daily basis neglects the potentially significant climate changes at smaller time scales. To address these issues, the improved snake optimization-long short-term memory (ISO-LSTM) model with Gaussian mixture model (GMM) clustering is proposed to forecast wind power from an adaptive perspective. By exploiting the merits of the probabilistic classification, the K-means optimized GMM clustering enables an appropriate feature modelling for substantial climate changes at smaller time scales. Then the ISO algorithm exhibits higher search accuracy and is better suited for finding hyperparameter combinations for LSTM neural networks. The data from the National Aeronautics and Space Administration (NASA) of the US is used to validate the effectiveness of the proposed method. Compared to the traditional K-means clustering, the K-means optimized GMM clustering has increased accuracy by 2.63 %. Simultaneously, with the adoption of the enhanced ISO algorithm, the accuracy further increases by 7.27 %. Different existing models have also been tested; it shows that the proposed model demonstrates higher prediction accuracy.