Accurate open-circuit voltage (OCV) modeling is critical for reliable state-of-charge (SOC) estimation in battery management systems (BMS). This study proposes a data-model hybrid-driven OCV framework to couple hysteresis and relaxation effects in lithium-ion batteries (LIBs). By integrating Long Short-Term Memory (LSTM)predicted factors into a physics-constrained formulation, the framework explicitly represents both path- and time-dependent voltage behaviors. Validation through dedicated hysteresis-loop and relaxation experiments confirms the model's accuracy and adaptability. The proposed hybrid OCV model is embedded into a dual polarization model (DPM) and evaluated under multiple dynamic driving profiles, achieving terminal-voltage prediction errors as low as an RMSE of 0.00834 V and an MAE of 0.00361 V, outperforming representative hysteresis-based OCV models. Building on this foundation, a multi-timescale dual central difference Kalman filter (MT_DCDKF) is designed for robust SOC estimation. Experimental validation shows that, when combined with the hybrid OCV model, the MT_DCDKF achieves the highest SOC estimation accuracy, with an RMSE of 1.019% and an MAE of 0.899% across diverse operating scenarios. These findings demonstrate that the joint framework enhances filter observability, improves estimation robustness, and provides an interpretable pathway toward sustainable and efficient battery management.

Lithium-ion batteries are essential energy storage devices that are extensively utilized in many different contexts. Battery performance, security and lifespan are all intimately correlated with the estimation accuracy of the state of charge (SOC). To improve estimation accuracy and reliability of SOC based on single sensor, the subject of multi-sensor distributed SOC estimation is examined in this work. Each sensor communicates measurement data with its neighboring nodes. Considering sensor random fading and communication overhead, random variables are introduced to characterize sensor fading and an incremental triggering mechanism is adopted to reduce communication overhead. A distributed SOC filter based on the measurements from node itself and its neighboring nodes is designed to minimize the upper bound of the variance (UBV) at every sensor node. It is demonstrated that the filtering error is exponential mean-square bounded. The method's efficacy is verified under different conditions by open-access datasets.

Accurate estimation of state of charge (SOC) for lithium iron phosphate (LiFePO4) batteries remains challenging due to cumulative errors from uncertain initial conditions and aging-induced capacity fade, which are often addressed in isolation by existing methods. To bridge this gap, this study proposes a novel switching-sequential hybrid framework (II-Ah) which synergistically integrates: (1) a threshold-triggered switching logic that seamlessly transitions from Unscented Kalman Filter (UKF)-based observer (for rapid initial optimization) to lowcomplexity ampere-hour integration, ensuring both high accuracy and computational efficiency; (2) a real-time capacity update mechanism, driven by an internal-resistance-based state of health estimator, that continuously compensates for capacity degradation without requiring offline intervention; and (3) the unified coordination of both mechanisms. Validated using offline measured data in MATLAB over 100 cycles, the proposed II-Ah estimator achieves a mean absolute error of 1.10% and a root mean squared error of 1.70% for a 2Ah battery. Its robustness and scalability are further demonstrated for 2Ah and 16Ah batteries via high-fidelity COMSOL-MATLAB co-simulation platform, maintaining estimation errors below 2% under dynamic operating conditions. This work presents a computationally efficient and adaptive solution, effectively mitigating the dual challenges of initial inaccuracy and capacity degradation for practical battery management systems.

For multi-sensor networked discrete-time stochastic linear systems with correlated noises, Kalman filters (KFs) in the linear minimum variance (LMV) criterion are deduced under the stochastic communication protocol (SCP) scheduling, bandwidth allocation, and uniform quantization effects. To resolve conflicts in multi-sensor data during communication, the SCP is utilized to ensure that only one node transmits its measurements to the filter at each step. To utilize the communication channel effectively, bandwidth is allocated according to the signal-tonoise ratio of measurement components of the sensor node selected to transmit data, and a uniform quantizer is utilized to quantize measurements before transmission. A random variable is introduced to indicate which sensor's quantized measurement is sent to the filter in each transmission. An optimal KF dependent on the value of the random variable is achieved in the LMV criterion, and the boundedness of the filtering error covariance matrix (ECM) is proved. It requires computing the filtering gain matrix online. To avoid the online computation of gain, a suboptimal KF dependent on the probability distribution of the random variable is developed, and its stability and steady-state properties are analyzed. The performance of the optimal and suboptimal KFs is compared. Ultimately, a simple target tracking example is used to illustrate the effectiveness of the achieved algorithms.

The distributed state estimation problem is studied for multi-sensor networked stochastic uncertain systems with correlated noises under a stochastic communication protocol (SCP). Random parameter matrices are utilized to describe the stochastic uncertainties within the system model. Given the limited channel bandwidth among sensor nodes, a general SCP is set to randomly select multiple components from the complete state prediction estimate for transmission. A set of random variables is introduced to indicate which combination of state prediction components is selected for transmission at each time step. In the case that the sensor node does not know which combination of state prediction components from each neighboring node is transmitted to it at each time step, a distributed Kalman-like recursive estimator structure that depends on the probability distributions of random variables is developed. Under this estimator structure, an optimal distributed estimation algorithm is presented based on the linear unbiased minimum variance criterion, which necessitates the computation of estimation error cross-covariance matrices between different nodes. To avert the computation of cross-covariance matrices, a suboptimal distributed estimation algorithm is also proposed, where optimal gains are achieved by minimizing the upper bound of estimation error covariance matrix at each node. In addition, the scalar parameters in the upper bound of the covariance matrix are optimized to obtain a minimum upper bound. Stability and steadystate properties of two distributed estimation algorithms are analyzed. Finally, the effectiveness of the presented algorithms is validated through a simulation example.

This article studies a distributed cooperative state estimation problem for nonlinear multiagent systems based on absolute and relative measurements. Each agent achieves cooperative state estimation by integrating its own absolute measurements and relative measurements with neighboring agents. Each agent receives local estimates from neighboring agents to linearize the nonlinear function in the relative measurement equation. An optimal distributed cooperative extended Kalman filter (EKF) algorithm is proposed by minimizing the filtering error covariance matrix (FECM), which requires calculation of cross-covariance matrices (CCMs) between agents. To avoid calculation of CCMs, a suboptimal distributed cooperative EKF algorithm is also proposed by minimizing an upper bound of the FECM. The exponential mean square boundedness (MSB) of the proposed distributed filters is proven. A cooperative localization tracking system with six mobile robots is used to evaluate the effectiveness of the proposed algorithm.

This paper concerns the modeling and distributed secure filtering for sensor networks with randomly varying nonlinearities, where sensor-to-filter channels and channels among filters are simultaneously corrupted by randomly hybrid attacks. A unified framework is proposed to model randomly varying nonlinearities and two-channel hybrid attacks. By means of an augmentation method, the global filtering error system with the dynamics of all nodes is obtained. Based on analysis of stability and H infinity attenuation effect for the global filtering error system, a novel distributed nonlinear H infinity filter which can suppress randomly varying nonlinearities and twochannel hybrid attacks of unknown prior information is designed. A sufficient condition is derived such that distributed filter gains are obtained by offline solving a linear matrix inequality. An electromechanical servosystem is provided to demonstrate the validity of the developed algorithm.

This paper is focused on the issue of distributed recursive linear fusion estimation in multi-sensor multi-rate linear discrete-time stochastic systems with non-Gaussian noises. The asynchronous sampling system is transformed into a synchronous sampling system through the pseudo-observation method. First, local filter in the minimum error entropy criterion is obtained at each sensor. Then, a distributed recursive linear fusion filter without feedback in the linear unbiased minimum variance criterion is presented based on local filters from all sensors. Estimation error cross-covariance matrices between local filters are derived. The proposed fusion filter is more accurate than the matrix-weighted fusion filter from local filters. Finally, to further improve the estimation accuracy, a distributed recursive linear fusion filter with feedback is presented, which avoids calculating cross-covariance matrices. The effectiveness of fusion algorithms is verified by simulations.

This paper proposes a novel reconfigurable non-isolated three-port converter for integrating photovoltaic (PV) power generation, energy storage, and loads. Based on a single shared inductor and a reconfigurable switching network, the topology achieves adaptive regulation across a wide input voltage range (24-120 V) at the PV port, substantially reducing conduction losses and magnetic core volume. By introducing a hybrid energy coupling mechanism-including series-boost coupling between the PV and battery ports, parallel current-sharing between the PV and backup source ports, and a composite power supply mode-the system enhances PV energy utilization under low irradiance conditions and improves overall output gain. A robust dual-loop PI controller, designed based on a full-range state-space model, effectively suppresses ripple oscillations caused by multisource coupling. A 150 W experimental prototype validates the feasibility of the proposed scheme, demonstrating efficiency above 90% across multiple operating modes, along with favorable dynamic performance and fault tolerance, making it suitable for PV-storage systems.

Addressing limitations in conventional nonisolated three-port converters (TPCs) for photovoltaic-energy storage (PV-ES) systems-such as restricted voltage range, high hard-switching losses, and soft-switching performance dependent on load power, this article proposes a reconfigurable, high power density, lossless nonisolated TPC. By integrating a dual-output architecture with the concept of power-path reconfiguration, single-stage wide-voltage-range energy conversion among the PV port, ES port, and load port is achieved. Combining an inductor series mechanism and switch parallel capacitor design, a hybrid reconfigurable quadrilateral modulation method for inductor current is proposed. This enables TPC to achieve full-load-range zero-voltage switching (ZVS) across a wide input voltage range of 29.43-80 V, increasing the maximum power transfer capability to 512.99 W. Furthermore, aiming to minimize the root-mean-square inductor current, the optimal duty cycle combinations for achieving full-load-range soft-switching in multiple operating modes are determined, significantly reducing switch conduction losses and inductor losses. Experimental results demonstrate that the proposed TPC achieves full-range ZVS in all three modes: single-input dual-output, dual-input single-output, and single-input single-output. It achieves a peak efficiency of 97.1% at rated power with voltage ripple below 0.71%. This work provides theoretical and technical support for efficient energy management in PV-ES systems.

This paper studies the distributed estimation problem for multi-sensor multi-rate networked systems with state equality constraints, where random transmission delays and packet losses occur during data transmission. The pseudo-measurement method is used to transform the multi-rate system into an equivalent synchronous system. To reduce computational cost, an unconstrained distributed filter that avoids computing cross-covariance matrices between sensor nodes is developed. Optimal gains are designed to minimize an upper bound of filtering error covariance matrix. The optimal scalar parameters coupled with gains are optimized to minimize the upper bound of filtering error covariance matrix. Subsequently, a constrained distributed filter is obtained by projecting the unconstrained distributed filter onto the constraint surface. A target tracking is used to verify the effectiveness of algorithms.

How to learn useful features hidden in process data for constructing an accuracy-acceptable soft sensor is still a challenge for the complicated process data with high complexity and strong nonlinearity. Deep learning (DL) has superior feature extraction ability and has been developed in many fields recently. In this work, evolved from the conventional stacked autoencoder (SAE), a novel DL framework called hierarchical double-parallel autoencoder (HDPAE) is designed for soft sensing. In the layer-wise feature extraction process of SAE, the reconstruction errors will be accumulated, which may reduce the modeling performance. In order to avoid this negative effect, the proposed HDPAE model takes the original input data and the extracted feature representations in previous DPAEs together as the input of high-level DPAE, which can not only avoid losing original input information in SAE, but also further enhance feature representation. The superior effectiveness of the proposed method can be validated on the debutanizer column industrial case.

This paper deals with the issue of state estimation for nonlinear non-Gaussian networks suffering from deception attacks. A novel distributed maximum correntropy filter with variance correction is proposed, which consists of local estimation and diffusion fusion. Firstly, the estimates from neighbor nodes are fused to obtain a common prediction estimate, as well as the upper bounds of the corresponding covariances are derived to avoid calculating correlated information. The chi-square test is employed to rectify the inaccurate noise variance caused by non-Gaussian noise and deception attacks. Based on minimum error and maximum correntropy criteria, distinct consensus gains and filtering gain for the local filter are derived. Then, local estimates of neighbor nodes are fused based on covariance intersection to enhance the accuracy of the node itself. The designed distributed structure combines the merits of consensus and diffusion filters, which can further enhance the estimation performance of the whole system. The statistical linearization and the cubature rule are applied to deal with nonlinear functions, avoiding Jacobi matrices and additional uncertain parameters. In addition, it is proved that the filtering covariance is bounded under certain conditions. Finally, simulation results verify the effectiveness and accuracy of the presented algorithm.

In the context of multisensor linear discrete networked descriptor systems, an equivalence transformation, achieved via singular value decomposition, leads to the derivation of two lower-dimensional non-descriptor subsystems. Each network node can perform state estimation based on data of its own and its neighboring nodes. Applying the Kalman consensus filter (KCF) framework, wherein one-step prediction estimates of a reducedorder subsystem are exchanged among network nodes, a distributed reduced-order KCF is designed for each sensor node, incorporating multiple consensus gains. This design facilitates collaborative state estimation by enabling nodes to leverage both their own measurements and the prediction estimates received from their neighbors. The optimal Kalman filtering gains and the optimal consensus filtering gains are determined by minimizing the trace of the filtering error covariance matrix. The investigation delves into the stability and steady-state characteristics of the tailored distributed reducer-order filtering systems. The performance of the algorithms is confirmed through illustrative simulation cases.

This paper addresses the issues of potential packet losses during communication and non-Gaussian noise interference in multi-sensor networked systems. Initially, the Cauchy kernel function-based minimum error entropy criterion is employed to measure the similarity of the information entropy of state estimation errors, thereby mitigating outliers caused by non-Gaussian noise. Subsequently, a set of independent and identically distributed Bernoulli random variables is utilized to characterize the phenomena of packet losses occurring during data communication between sensor nodes. What's more, the system model is reconstructed using an augmented coefficient matrix that incorporates random variables. Building upon this foundation, a distributed minimum error entropy fusion filtering algorithm with a fixed-point iterative form is derived. Finally, the effectiveness of the proposed filtering algorithm is validated through numerical simulations.

The robust fusion filtering problem of multi-sensor networked uncertain descriptor systems (NUDSs) with colored noise, uncertain noise variances and cyber-attacks is investigated. During data transmission in unreliable communication networks, the data can be maliciously attacked by attackers. In other words, the local filters (LFs) may receive false data or may not receive data because of the cyber-attacks. By adopting the singular value decomposition (SVD) method, the original NUDSs can be converted into two reduced-order subsystems with uncertain correlated fictitious white noises, and the cyber-attacks are transformed into the fictitious noises. Cross-covariance matrices between local filtering errors are derived. The robust LFs are obtained according to the minimax robust estimation principle. Under the linear unbiased minimum variance criterion, three weighted fusion algorithms are applied to fuse the LFs. For all allowable uncertainties of noise variances and cyber-attacks, the minimal upper bounds of covariance matrices of the local and distributed fusion filters are guaranteed. The proof of their robustness is established through the minimax estimation principle and Lyapunov equation method. Finally, the correctness and effectiveness of the proposed algorithms are verified by a circuit system example.

In this paper, an optimal predictor is designed for networked descriptor systems with multiple stochastic measurement delays under encoding-decoding strategy. The measurement delays phenomenon is described by a set of Bernoulli variables with known probabilities, where missing measurement is considered a special case of this phenomenon. By using the singular value decomposition (SVD) method, the original descriptor system is decomposed into a fast subsystem and a slow subsystem, and the measurement delay terms are eliminated by augmenting the system state. Based on the projection theorem, an optimal predictor is designed and a white noise estimator is used to solve the colored noise caused by the system model transformation. The effectiveness of the algorithm is validated through a simulation experiment.

The issue concerning distributed H, fusion filter (DHFF) design of joint fault and state for insecure multi-sensor networked systems is studied. In an insecure network environment, the channels from sensor to local joint filter and from local joint filter to fusion filter are simultaneously injected by probability-uncertain deception attacks. A novel attack-tolerant local joint filter (LJF) is proposed to estimate the fault and state of the system in a unified framework. A universal DHFF design algorithm with a smaller H, attenuation level than every LJF is also proposed. By means of the analyses of asymptotically mean-square stability and H, attenuation effect, the sufficient conditions for establishment of LJF and DHFF are given in the form of linear matrix inequalities (LMIs), respectively. Fusion estimates of the state and fault are established simultaneously via solving an LMI, which need not be designed separately. The validity of the achieved algorithm is demonstrated by a vehicle suspension system.

The distributed extended Kalman consensus filtering problem under stochastic communication protocol (SCP) is investigated for multi-sensor networked nonlinear systems. In the sensor network, each sensor node and its neighbor nodes occupy a limited number of communication channels when exchanging measurement data. Utilizing the SCP-equipped communication network ensures that the neighbor nodes of each sensor node randomly access these channels and send measurement data based on the number of channels at each step. A set of random variables is introduced to represent the neighbor nodes whose measurements are selected for transmission at each step. When each sensor node is aware of the measurement data received from its neighbor nodes at each step, a distributed extended Kalman consensus filter dependent on random variables is designed. To improve the estimation consensus among nodes, a consensus term is added to the performance metric, and a weighting factor is introduced to assign weight between estimation accuracy and consensus. The optimal filtering gain is derived by minimizing this performance metric. A sufficient condition for the exponential mean-square boundedness of the filtering error is given. Finally, the proposed algorithms effectiveness is validated through a simulation example.

This paper is concerned with the information fusion estimation problems for stochastic uncertain systems with quantized measurements and mixed network attacks including random deception attack and denial-of-service (DoS) attack. The cross-correlated random parameter matrices and addictive noises simultaneously exist in the studied system. By resorting to the optimal prediction compensation mechanism for DoS attack, an optimal centralized fusion filter in the linear minimum variance sense is proposed using an innovation analysis approach. In addition, the Kalman-like recursive distributed optimal linear fusion predictor and filter without feedback are presented based on local estimators from single-sensor subsystems. The estimation error cross-covariance matrices between two arbitrary local estimators, and those between local and prior fusion estimators are derived. They have good flexibility due to the parallel structure. However, they have lower accuracy than the centralized fusion estimators. To further improve the estimation accuracy, the distributed optimal linear fusion predictor and filter with feedback are also presented. They avoid the calculation of cross-covariance matrices. Moreover, it has been mathematically proved that they have the same estimation accuracy as the centralized fusion estimators. A simulation example demonstrates the effectiveness of the proposed algorithms.