Lithium-ion batteries often exhibit significant cell-to-cell variability during capacity grading and pack integration, which can be further amplified at the pack level and affect system consistency and service life. Therefore, evaluating cell lifetime potential at the capacity grading stage or during very early cycling is critical for battery screening and consistency-based grouping. To address the limitations of traditional methods that rely on long-term aging tests, this study proposes an early lifetime assessment approach using capacity grading-stage single-cycle charge-discharge data. First, full-lifetime cycling data were divided into source-training and target-batch assessment sets, and multidimensional degradation-related features were extracted at several key state-of-health (SOH) nodes. Recursive Feature Elimination (RFE) was applied to select the most informative features. Subsequently, a Hierarchical Degradation Learning Network (HDL-Net) was developed, in which the Macro Degradation Structure Layer (MDSL) captures the global degradation trend, while the Adaptive Residual Refinement Layer (ARRL) models cell variability and local nonlinear characteristics. A Simulated Annealing Weight Optimizer (SAWO) was further incorporated as a lightweight residual distribution calibration module to refine the ARRL output. Experimental results show that the proposed method can effectively evaluate lifetime differences using capacity grading-stage single-cycle charge-discharge features, achieving an MAE of 11.015, an RMSE of 17.969, and an Acc±5% of 93.75%. Validation on two additional datasets from Tongji University further demonstrates the robustness and applicability of the proposed approach for battery screening and consistency-based grouping.
In this paper, a novel hybrid pole permanent magnet (HPPM) machine with combined Halbach-array and V-shape magnet arrangements is presented for special electric vehicles (EVs). The simplified magnetic equivalent circuit (MEC) is established to illustrate the flux-concentrating mechanism of the proposed hybrid pole design, and the design parameters are optimized so as to improve performance. Then, the contributions of harmonic components for different permanent magnets (PMs) are analyzed and researched, which reveals the mechanism of torque improvement and torque ripple suppression. Moreover, the electromagnetic performances of the proposed HPPM machine are obtained and compared with traditional PM machines, including back electromotive force (back-EMF), torque characteristics, and efficiency. The results show that the HPPM machine shows higher magnetic flux density and stronger torque output capability, and the torque density is achieved at 8.51N·m/kg to accommodate the specific needs of the special EVs. Finally, a 40kW 60-slot/8-pole prototype is manufactured and tested to verify the effectiveness of the proposed machine.
High-frequency operation, high resonant voltage stress, and harsh operating environments make wireless power transfer (WPT) systems highly susceptible to various faults. However, most existing fault diagnosis methods in electrical engineering are targeted at motors and power systems, and are not directly applicable to WPT systems with different operating conditions and circuit topologies. This paper proposes a multi-fault diagnosis method for dynamic wireless power transfer (DWPT) systems based on resonant capacitor voltage. By monitoring and calculating the DC and AC components of the resonant capacitor voltage to determine the system fault characteristics, the proposed method realizes the diagnosis and localization of inverter power transistor open-circuit, transmitting rail inter-turn short-circuit, and resonant capacitor breakdown faults, providing a basis for subsequent fault-tolerant control and targeted maintenance. After analyzing the impacts of different faults on DWPT system characteristics, this paper identifies the corresponding fault features. On this basis, a fault diagnosis method based on resonant capacitor voltage detection is proposed, and the diagnosis procedures for each fault are presented. The method is simple to implement, as it avoids the effects of dead time, leakage inductance, power transistor switching delays, parameter deviations, and sampling errors, front-end grid fluctuations, and vehicle lateral misalignment. It offers high real-time performance, reliability, and robustness, providing a brand-new solution for the online real-time fault diagnosis of DWPT systems. An experimental prototype is built for verification, and the results confirm the effectiveness and accuracy of the method in diagnosing all targeted faults.
In this paper, by using a nonlinear control strategy, we prove that the elimination of the fuel cell power converter in an electric vehicle powertrain is feasible. This strategy indirectly regulates the power delivered by the fuel cell through a nonlinear control implemented by a single DC-DC converter connected to a supercapacitor. This configuration allows for efficient adjustment of the operating point, in combination with an energy management strategy, to adapt to power demands. As a result, there is a reduction in the size of the powertrain and an increase in efficiency by eliminating the need for multiple DC-DC converters to directly control the fuel cell. Experimental test has been carried out, on a real electric vehicle powertrain, to validate the development.
At high speed, permanent magnet synchronous motors (PMSMs) exhibit increased stator iron loss due to the high fundamental electrical frequency. The resulting eddy-current reaction changes the magnitude and phase of the observed flux linkage, which may introduce angle estimation errors in position-sensorless control. Existing methods usually introduce an equivalent iron-loss resistance into the voltage model, but the eddy-current-induced flux attenuation and phase shift are not explicitly embedded into the active flux used for rotor-position estimation. To address this issue, a vector magnetic circuit (VMC) active-flux observer is proposed for high-speed PMSM sensorless control. The VMC flux-linkage relationship is reformulated into an active-flux-compatible form through linearization, allowing the eddy-current-induced flux attenuation and phase shift to be represented by the VMC active flux and the corresponding equivalent inductance. Considering the speed-dependent characteristic of the VMC model, a stationary-axis speed observer based on the mechanical model is also developed to improve transient performance and reduce steady-state speed error. Simulation and experimental results show that the proposed method reduces the angle estimation error by 48.9% compared with the conventional active-flux observer, confirming its effectiveness for high-speed PMSM sensorless control.
To effectively attenuate current ripple of the permanent magnet synchronous motor (PMSM), this paper proposes an enhanced active disturbance rejection control (EADRC) based on a dual error constraint observer with novel resonant controller (NRDEO). Within the proposed architecture, the dual error constraint observer (DEO) is designed to improve disturbance estimation performance by introducing an auxiliary error correction term on top of the conventional single error extended state observer. Meanwhile, a novel resonant controller (NRC), which is embedded in the disturbance estimation loop of the DEO, is developed to eliminate the resonant peak near the target frequency. With the direct current and alternating current disturbances accurately estimated, the feedback control law of EADRC is presented to effectively reject these disturbances. Furthermore, theoretical analysis comprehensively investigates the disturbance estimation and rejection capabilities, the convergence of the NRDEO, and the stability of the EADRC, which reveals that the proposed scheme can effectively enhance current dynamic performance and steady-state accuracy. Finally, the feasibility of the proposed scheme is evaluated on the PMSM experimental platform under various operational conditions.
This paper proposes a novel zero-sequence filter (ZSF) for open-end winding (OEW) permanent magnet (PM) motors and verifies its effectiveness in suppressing harmonics originating from non-ideal back electromotive force (BEMF). Although OEW configurations offer higher voltage utilization and advantages in high-speed operation, they are vulnerable to zero-sequence current (ZSC) harmonics due to the absence of a neutral point. These harmonics increase torque ripple, vibration, and power losses. To mitigate these effects, a distributed-winding-based ZSF structure that can be retrofitted and seamlessly integrated with the motor stator is proposed. The ZSF increases zero-sequence inductance without compromising motor output performance or requiring additional compensation control. The size and parameters of the filter can be determined based on mechanical constraints, and the resulting ZSC reduction can be analytically derived. A structurally compatible ZSF design with a conventional OEW PM motor is developed based on the slot and tooth configuration of a 4-pole, 12-slot combination. The proposed structure is magnetically equivalent to a ZSF with a toroidal core while maintaining mechanical compatibility with the motor winding layout. Finite element analysis (FEA) using Ansys Maxwell is conducted to evaluate the zero-sequence inductance and magnetic saturation characteristics across operating conditions. The results confirm that the designed filter achieves the target inductance and provides over 80% harmonic suppression at rated speed. Experimental validation is performed and the measured results demonstrate that the proposed integrated filter effectively suppresses ZSCs caused by BEMF harmonics.
The transition to electric mobility is essential to meet global climate targets, but it poses critical challenges to the stability of the power grid due to the high, concentrated demand from electric vehicles. This paper proposes a discrete-event optimisation model for the allocation and scheduling of electric vehicle charging, focusing on electric buses, which are characterised by large battery capacities and predictable schedules. The developed mixed-integer linear model introduces a detailed battery representation that captures nonlinear charging behaviour and incorporates key constraints such as power limits, time windows, charging station availability, time of use electricity tariff, and provides multiobjective optimisation for makespan and energy cost minimisation. Compared to traditional discrete-time approaches, the proposed formulation significantly reduces problem size while maintaining modeling accuracy. The proposed framework is validated using a real case study provided by Iveco S.p.A. company, involving a fleet of maximum fifty electric buses. The robustness of the model with respect to key parameters, together with the uncertainty analysis performed on simulated scenarios and the scalability analysis with an increasing number of electric buses, demonstrates the good performance of the proposed approach under different operational conditions.
Oil cooling is widely adopted to increase the power density of electric vehicle traction motors. However, part of the coolant often enters the air-gap through winding dripping, rotor splashing, and bottom accumulation. The air-gap oil produces strong viscous shear, increasing friction loss and causing noticeable torque fluctuation. This work introduces a rotor surface design that achieves active oil evacuation through droplet-scale transport mechanisms. The design is based on the observation that droplets experience force imbalances among centrifugal force, viscous stress, and surface tension when interacting with properly shaped microstructures. Two types of surface features are applied to the rotor. At the end region, small conical elements create a preferred path for outward motion by converting the combined effect of centrifugal force and surface tension into radial displacement. In the central region, shallow slots aligned with the circumferential flow generate an axial force component that gradually transports droplets toward the end. Together, these features form a continuous evacuation route and allow droplets to leave the air-gap more rapidly. Experiments confirm the proposed design reduces the amount of oil entering the air-gap, accelerates the oil evacuation, lowers residual oil volume, and decreases the associated viscous torque.
To address the problems of insufficient transient disturbance rejection capability and degraded steady-state current performance in surface permanent magnet synchronous motor (SPMSM) speed control, this paper proposes a finite-time switching active disturbance rejection control (FSADRC) method. The proposed method consists of a finite-time cascaded extended state observer (FCESO) and a switching control law. The FCESO is composed of a first-layer finite-time ESO and a second-layer linear ESO. The first-layer finite-time ESO and the second-layer linear ESO are designed to provide speed and disturbance estimations for the transient and steady-state stages, respectively, achieving both high disturbance rejection capability, rapid response, and high steady-state current performance. To suppress oscillations during the transition between transient and steady-state modes, a linear smooth switching mechanism is developed to ensure the continuity of the control signal. Experimental results demonstrate that the proposed FSADRC significantly improves transient disturbance rejection capability, convergence speed, and steady-state speed and current performance, thereby validating the effectiveness and robustness of the proposed FSADRC.
Fault detection and quantitative assessment are critical for ensuring the reliability of proton exchange membrane fuel cell. However, existing methods often suffer from limited generalization capability of data-driven models and insufficient robustness in fault assessment. To address these challenges, this study proposes a fault diagnosis framework that systematically fuses structural and mechanistic prior knowledge. First, a topology-aware graph network is constructed to simulate the information interaction of real system, and prior voltage equations are embedded into graph nodes to establish the reference model for residual-based fault detection. For online severity quantification, a multi-perspective metric based on improved dynamic time warping distance is developed, which comprehensively measures discrepancies in value deviation and fluctuation levels between reference and fault signals. Experiments on a typical commercial stack show that, compared with the latest methods, the proposed method increases the fault identification accuracy by 14.6% and 7.2% respectively. Furthermore, the proposed method exhibits superior capability in tracking fault evolution and robustness against false fluctuations.
To achieve a favorable trade-off between output capability and thrust cost for long-stroke rail applications, this article proposes a hybrid Halbach magnetic lead screw (HH MLS). The outer translator employs a Halbach array, while the inner rotor adopts a flux-focusing unit composed of axially magnetized permanent magnets and soft magnetic materials. An aluminum rotor core is used to maintain a favorable magnetic flux path while reducing rotor mass. A finite-element (FE) model is established to verify the operating principle and support the design optimization. Using thrust density and thrust cost as the optimization objectives, a sensitivity-based optimization is adopted to determine a favorable parameter combination. Comparative results with representative MLS topologies indicate that the proposed HH MLS achieves a balanced compromise between output capability and material-related cost, and its thrust-cost advantage becomes more pronounced as the stroke increases. An M1.6 screw-fastened axial-offset arc-segment PM assembly is adopted for prototype realization. Experimental results agree well with the FE predictions, validating the proposed HH MLS and the adopted manufacturing scheme.
Maintaining a stable anode pressure and hydrogen excess ratio (HER) in the hydrogen supply system (HSS) is critical towards the operational efficiency and reliability of a PEM fuel cell. In this paper, we present a novel multiple-input multiple-output (MIMO) double-loop controller based on an active inference control (AIC) approach for coordinated regulation of pressure and HER in the HSS. An HSS model is first established, incorporating a segmented anode model to approximate the stack’s distributed dynamics. Inspired by the neuroscientific free energy principle, the AIC is formulated in a variational Bayesian framework to address the pronounced nonlinearities inherent in the stack pressure and pump flow dynamics. The AIC serves as the outer-loop controller, in which a goal-directed conscious perception mechanism is designed based on a probabilistic generative model to infer system states and proactively predict their higher-order dynamics. By minimizing the free energy function, the AIC achieves robust and offset-free tracking of anode pressure and HER setpoints without requiring an accurate physical plant model, while dynamically generating reference values for the inner loop. The MIMO inner-loop controller employs a nonlinear state-feedback strategy to manage the well-established dynamics of the manifold pressure and pump motor response. Compared with existing controllers, our method mitigates transient overshoot in both anode pressure and HER, while concurrently reducing the respective root mean square errors by 88.1% and 38.2%. Hardware-in-the-loop experiments are conducted to validate the effectiveness of the proposed method.
The Brushless Direct Current Motor (BLDCM) is widely adopted for electric vehicles due to its high efficiency and reliability. However, speed fluctuations and load changes under cross-operating conditions lead to significant feature distribution drift. This phenomenon severely limits the generalization performance of traditional models. Moreover, existing data-driven solutions face a critical trade-off between diagnostic accuracy and computational efficiency. This often results in models that are either too fragile for dynamic environments or too resource-intensive for edge deployment. To address these challenges, this paper proposes a lightweight Temporal Differential Fusion Attention Neural Network (TDFANN). This network is combined with a physics-informed hierarchical transfer learning strategy to enable efficient inter-turn short-circuit diagnosis under cross-operating conditions. First, the TDFANN integrates a feature decoupling module, a backbone for long-range temporal feature extraction, and a lightweight attention mechanism. Second, a grouped transfer strategy based on parameter sensitivity analysis is introduced. This strategy enables low-data and low-latency adaptation from a single source domain to complex target domains. Finally, experiments on the platform validate the generalization and effectiveness of the model. Experimental results demonstrate that the proposed model achieves a fault diagnosis accuracy of 98.84% under complex operating conditions. With a model size of 2.18 MB and an inference time of 16.8 ms, it satisfies the real-time constraints of industrial edge devices while ensuring robust diagnosis.
Dual three-phase permanent magnet synchronous motor (DTP-PMSM) drives commonly use four current sensors for high-performance current control. This paper proposes a topology-adaptive two-sensor control method for DTP-PMSM drives based on current reconstruction. Considering different installation combinations of two current sensors, a reduced-order model and a unified two-sensor mapping are built to analyze the full state observability and sensor placement effects on reconstruction robustness. Then, a topology-adaptive multidimensional extended state observer (TA-MESO) is designed to reconstruct four-dimensional d-q-x-y observation errors from two phase-current residuals. A multi-order quasi-proportional resonant (MO-QPR) controller and a speed-adaptive notch filter (SANF) are used to suppress the coupling between the fundamental d-q and harmonic x-y residual channels. Furthermore, sliding window integration (SWI) is introduced to suppress the dc offset of current sensors. Experimental results under different sensor placements are compared with conventional four-sensor control and a traditional two-current-sensor method to verify the accuracy, robustness, and topology adaptability of the proposed method. The dc-offset suppression capability is further verified by injecting artificial dc offset.
Conventional design of permanent magnet motor rotors follows a topology-first paradigm that requires repeated finite-element analysis (FEA) iterations separately for each candidate topology, resulting in low efficiency and strong topology dependence. To address these limitations, this paper proposes a requirement-driven motor topology generation (RMTG) framework that directly generates rotor topology designs under user-specified performance targets. The framework integrates a conditional generative adversarial network with a pre-trained, frozen parameter predictor to enforce consistency between the generated topology and the desired electromagnetic characteristics, and a coordinated stabilization strategy resolves the training instability introduced by performance conditioning. Trained on 40,000 rotor topology images across four IPMSM families, RMTG achieves R2 > 0.97 for all three performance parameters (average torque, torque ripple, and core loss) and 100% topology-type classification accuracy. Two fabricated prototypes further confirm close agreement (within 2% discrepancy) among generated designs, FEA, and experimental measurements, demonstrating that RMTG provides an efficient end-to-end path for requirement-driven cross-topology design of permanent magnet motors.
Interior permanent magnet synchronous motors (IPMSMs) for electric vehicles predominantly operate in torque control mode, requiring accurate flux linkage and torque information for high-performance control and health monitoring. However, due to the nonlinear variation of permanent magnet (PM) flux linkage and cross-coupling effects, the inductances and PM flux linkage both exhibit nonlinear variations, which cause difficulties in flux linkage and torque estimation. Hence, this paper proposes a multi-effect equivalent inductance (MEI) model-based flux linkage and torque simultaneous estimation method. MEIs are defined to reflect variations for the PM and dq-axis flux linkages, respectively. To improve the identification speed and accuracy of MEIs, an asynchronous axis high-frequency injection (ASAHI) based method is proposed. By observing the current amplitudes with the reduced injection angles, MEIs can be identified rapidly. Furthermore, for the PM flux linkage estimation, the PM flux linkage equivalent inductance is decoupled by the magnetic curve variation characteristics, and the PM flux linkage can be predicted by this inductance. To obtain high-precision dq-axis flux linkage by MEIs with limited sampling counts, an optimized path integration (OPI) method is presented, which selects the optimal integration path by the integration error evaluation function. Finally, the proposed method is validated on a 200 kW IPMSM platform for vehicles.
Despite numerous benefits, the growing penetration of electric vehicles (EVs) brings new challenges to distribution network operation and reliability. Transactive energy (TE) has emerged as an effective approach for managing these resources within a market environment. This study proposes a network-constrained, TE-based scheduling framework for vehicle-to-grid (V2G)-capable EVs to optimize real-time operations of local distribution networks (LDNs) with multiple charging stations. The proposed framework constitutes a single-sided flexibility procurement market in which EV owners participate by submitting their preferences and operational parameters, based on which the local market operator (LMO) constructs individual response curves at each time interval–serving as offer functions. These response curves encode each owner’s willingness-to-accept for different levels of flexibility provision, derived analytically from the submitted inputs and V2G cost components. Considering the constructed response curves and incentive signals from the distribution system operator (DSO)–including dynamic pricing and a reward-penalty load-flattening scheme–the LMO runs a model predictive control optimization to clear the TE market, accommodating operational uncertainties. The effectiveness of the proposed model is validated through case studies on the IEEE 33-node and IEEE 69-node distribution test systems, demonstrating reduced LDN operational costs, improved load profile flatness, effective utilization of V2G flexibility, and computational feasibility across the tested configurations.
The progression of wound rotor synchronous machines (WRSMs) toward megawatt‑class power ratings in hybrid electric aircraft (HEA) propulsion systems necessitates dramatically higher field magnetomotive force (MMF), resulting in excessive stator armature winding losses and rotor field winding losses that critically constrain efficiency and reliability. Furthermore, high‑power‑density designs intensify machine cooling difficulties, making loss reduction imperative. This paper applies a profiled magnetic slot wedge (PMSW) to a WRSM. The PMSW features a variable cross-section and is composited of magnetic and non-magnetic materials. Utilizing magnetic equivalent circuit (MEC) and finite element analysis (FEA), this paper investigates the influence of the PMSW on the machine's electromagnetic performance. By modulating the air-gap magnetic field, the PMSW effectively reduces armature AC losses, suppresses rotor eddy-current losses, and decreases field current demand. A multi‑objective optimization of the PMSW’s structural parameters is conducted with respect to machine efficiency, output power, and rotor core loss. A 600kW prototype employing the optimized design is fabricated and experimentally validated.
The dual-inverter fed open-end winding motor (OEWM) systems have advantages such as multi-level output characteristic, inherent fault tolerance, and high DC-link voltage utilization. Thus, OEWM systems have gained significant attention for electric traction applications, such as electric vehicles and aircraft. This paper provides a comprehensive review of recent research on OEWM systems, focusing on innovations in inverter topologies, control strategies, modulation methods, and fault-tolerant operation. First, the common DC source topology and the isolated DC source topology are compared in terms of their characteristics and control methods. Then, various modulation strategies for the two topologies are qualitatively compared, with the advantages of different strategies analyzed in detail. Furthermore, the paper reviews recent progress in fault diagnosis and fault-tolerant techniques for OEWM systems. Finally, an outlook on future trends and challenges of OEWM systems is presented.
