Accurate prediction of electric vehicles charging loads is essential to enable fine-grained energy scheduling and ensure grid stability under large-scale vehicle-grid interaction. This study addresses this task by proposing a fine-tuning method based on TimesFM, a large-scale pure decoder-based model for time series prediction. The study first extracts charging behavior segments of typical vehicles from seven major cities in China to construct a minute-level load dataset, and introduces meteorological information and holiday factors as covariates. In the fine-tuning stage, three key strategies are adopted: (1) periodic window segmentation based on Fourier spectral analysis; (2) introduction of covariates to enhance the model’s ability to model non-stationary behaviors; and (3) multi-step rolling prediction mechanism based on a sliding window for overlaying long prediction periods. The results show that the fine-tuned model achieves an accuracy of about 85

The rapid adoption of electric vehicles (EVs) with increasingly advanced battery technologies is reshaping electricity demand patterns. Yet existing studies often assume static behaviors, overlooking that real-world charging patterns are transient and evolve in response to technological change. This study applies a scenario-aware generative modeling framework to project weekly EV charging demand in Beijing for 2030, capturing behavioral shifts driven by evolving battery technologies, usage patterns, and infrastructure conditions. Results indicate that total charging load could increase by 457-509% compared to a 2021 baseline under two different scenarios of battery size growth. Though medium-power (4-20 kW) charging remains dominant in event frequency, high-power ( > 20 kW) charging contributes substantially to loads, implying the growing risk of stress on the power grid in the absence of coordinated scheduling. The proposed framework provides a scalable, data-driven method to simulate EV usage and load patterns and offers valuable insights for transportation and energy planners confronting the rapid electrification of private mobility.

Degradation and safety are the fundamental imperatives of batteries, tightly intertwined and inseparable. Only by understanding the coupled mechanisms of degradation and thermal runaway (TR) can we gain control over battery safety across the entire lifecycle. In this work, we investigate commercial 20 Ah NCM/graphite pouch cells subjected to low-temperature and fast-charging coupled protocols, and track how their degradation from the beginning of life (BOL) to the end of life (EOL) reshapes TR behavior. By combining electrochemical diagnostics with multi-scale post-mortem characterization, adiabatic TR tests on full cells, electrode-level differential scanning calorimetry and post-TR gas chromatography, we build a stage-resolved framework that links interfacial composition, lithium plating and structural damage to changes in TR onset, heat release and gas composition. The results reveal a clear degradation critical point at around 90% remaining capacity, at which the dominant degradation mode shifts from mild loss of lithium inventory to lithium-plating-driven accelerated aging with pronounced interface reconstruction, leading to reduced T1/T2, stronger low-temperature anode exotherms and a transition from CO2-dominated to H2-and hydrocarbon-rich vented gases. Notably, loss of active material already initiates in the BOL -> CP stage, although capacity fade remains predominantly governed by loss of lithium inventory. With further degradation towards EOL, superposed interfacial thickening, cathode cracking and loss of active material further narrow the safety margin and increase TR severity relative to the remaining energy. This work provides a concise mechanistic link between specific degradation signatures and TR metrics, offering practical guidance for defining aging-aware safety envelopes and designing low-temperature fast-charging/discharging strategies for energy storage systems. (c) 2026 The Authors. Published by Published by Elsevier B.V. and Science Press on behalf of Science Press and Dalian Institute of Chemical Physics, Chinese Academy of Sciences. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

The realization of zero voltage switching (ZVS) plays a crucial role in improving the system electromagnetic compatibility (EMC), so wide-range ZVS operation is necessary for electric vehicles (EVs) wireless power transfer (WPT) system. However, to realize wide-range ZVS, the system impedance needs to be adjusted by the design of topology parameters and control strategies, which introduces large reactive current at light and medium loads, thus resulting in system efficiency degradation. To fill this gap, this paper presents a novel WPT converter integrated with dual-side LC branches that helps realize ZVS while significantly reducing the RMS current on switches, thereby minimizing the system conduction loss. Moreover, based on the mathematical model for the proposed system, a universal design method for LC branches and a combined modulation strategy are proposed to ensure the full operating range of ZVS and maintain high efficiency. Finally, a 3.3 kW experimental prototype is built to verify the proposed system and make a comparison with traditional phase-shift WPT systems. The measured efficiency is over 91.7% from 12% to 100% loads with a maximum efficiency improvement of up to 6.47% at the light load.

The rapid growth of electric vehicle ownership and advancements in vehicle-to-grid (V2G) technologies have created an urgent demand for bidirectional charging-discharging interfaces. Wireless power transfer (WPT) technology, known for its convenience, safety, and flexibility, is a promising solution for energy transfer between vehicles and the grid. This paper presents the design and demonstration of a highly interoperable and highefficiency bidirectional WPT system, addressing key challenges such as wide voltage output adaptation, multipower level compatibility, and efficient operation over a broad power range. The front-end converter uses a power module combining a three-phase fully controlled rectifier and a cascaded buck converter to provide a wide DC voltage range. Modular activation technology ensures the grid interface operates efficiently under varying power demands. For the bidirectional inductive power transfer (BIPT) link, an integrated scheme for the resonant networks in the ground assembly (GA) with cross-frequency compatibility is proposed, and its performance is validated through calculations and simulations. A bidirectional power flow control strategy is implemented, with voltage regulation and operation mode switching as the main method. Experimental results demonstrate interoperability between the same grid-side equipment and different vehicle-side equipment rated at 6, 11, and 30 kW. Under specified operating conditions at the aligned position, the system achieves a grid-to-battery efficiency from 91.7% to 94.3%, and a battery-to-grid efficiency ranging from 89.5% to 93.5%.

Accurate estimation of battery state of health (SOH) of power batteries is crucial for reliable assessment of driving range, safety, and service life in electric vehicles (EVs). Since SOH is commonly defined as the ratio of the current capacity to the rated capacity, accurate capacity estimation provides a practical approach for SOH assessment. However, many existing methods rely on a single type of feature, such as microscopic time-series signals or macroscopic statistical indicators, which limits their ability to capture complementary degradation information. To address this issue, this study proposes a time-frequency-statistical multi-domain feature fusion framework for battery capacity estimation using real-world EV data. Time-domain features are extracted from voltage and current sequences using a convolutional neural network-long short-term memory (CNN-LSTM) module, while time-frequency features are obtained through multi-level wavelet decomposition (MLWD). In addition, statistical features describing battery usage patterns are processed using a CNN. These features are fused through an attention mechanism and used to predict battery capacity. Experiments on data from 184 retired EV batteries show that the proposed time-frequency-statistical feature fusion method outperforms baseline approaches. The method effectively overcomes the limitations of single-feature-based capacity estimation and provides a promising solution for battery aging assessment.

In dynamic wireless power transfer (DWPT) systems employing transmitting (Tx) coil arrays, output power fluctuations aggravate battery charging current ripples and compromise system stability. To address this issue, this work proposes a CLC-S compensation topology based on a dual-input single-output (DISO) architecture along with a cooperative coil design methodology. Theoretical analysis of the DISO CLC-S compensation network reveals that equivalent mutual inductance fluctuation is the main cause of output power instability, and that cross-coupling between Tx coils can disrupt the zero-voltage-switching (ZVS) condition. Parameter sensitivity analysis identifies the Tx coil series compensation capacitance as the key tunable parameter decoupled from the system output power. Based on the ZVS boundary condition, a tuning strategy is proposed to suppress the cross-coupling effect. To minimize equivalent mutual inductance fluctuations, an analytical model for the mutual inductance between the receiving (Rx) coil and adjacent dual transmit (Tx) coils under lateral movement is proposed. The model enables the determination of the optimal center-to-center distance between Tx coils, and supports the design of an overlapping Tx coil layout. A 1.3 kW prototype is developed for validation. Test results demonstrate that, compared to a single-input single-output (SISO) system, the proposed solution reduces output power fluctuation by 85.87%, while full ZVS operation during dynamic charging is achieved. In this case, the average output power has been increased by 6.58%.

The design of magnetic couplers is crucial to the efficiency of inductive wireless power transfer (WPT) systems in EV charging application, where coil geometries and material layouts must be engineered under tight electromagnetic and packaging constraints. Conventional finite-element modeling (FEM) provides high fidelity but is computationally expensive for iterative design, while purely data-driven surrogate models are fast but often neglect electromagnetic constraints and material interactions. This paper presents a physics-informed neural network (PINN) framework that embeds the frequency-domain quasi-static A-form of Maxwell’s equations with boundary and interface conditions, incorporates material-aware eddy-current effects, and leverages lightweight FEM-based calibration for absolute scaling. The method parameterizes DD-shaped and rectangular coils together with ferrite and aluminum shielding, trains a physics-constrained network on these inputs, and performs forward inference to directly predict inductive characteristics from geometry parameters. Validation on representative prototypes at 85 kHz shows agreement within a few percent relative to FEM and bench measurements, while reducing evaluation time from tens of minutes to a few minutes per design, thereby enabling rapid and accurate modeling of practical WPT couplers.

To enhance State-of-Charge (SOC) variation prediction accuracy for Electric Vehicles (BEVs) under real-world driving conditions, this paper proposes a novel framework integrating multi-dimensional feature engineering with a multi-model stacking ensemble. Leveraging real-world operating data from 20 BEVs over two months, a comprehensive multidimensional feature set is constructed, incorporating available battery capacity, driving behavior characteristics, and operating condition distributions. Subsequently, feature selection is performed using a Support Vector Regression (SVR)-based importance evaluation. A stacking ensemble model is then developed, aggregating six heterogeneous machine learning base learners with a Linear Regression meta-learner. Experimental results demonstrate the superior performance of the proposed method, achieving an R2 of 0.97. Relative to the best-performing single model, the stacking ensemble reduces Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) by 3.3% and 5.1%, respectively, validating its robustness under complex real-world driving conditions.

To achieve wide-range power regulation without using additional dc-dc, phase-shift full-bridge (PSFB) for wireless power transfer (WPT) system is a promising way. However, for the PSFB WPT system, switching loss is a key factor affecting the system efficiency. Single-side PSFB can regulate the power, but hard switching results in low efficiency and EMI problems. Dual-side PSFB can realize ZVS throughout a wide power range, but at the cost of high conduction loss and turn-off loss at light load. In this article, a WPT converter combining tandem half bridge (THB) and active rectifier (AR) is proposed to reduce the switching loss of the system at light and medium loads. Meanwhile, an optimized hybrid triple-phase-shift (TPS) and extended-phase-shift (EPS) control strategy is proposed to achieve ZVS and power regulation with adaptive modes, thus further reducing conduction loss and switching loss and optimizing light-load efficiency. Finally, A 1.2 kW experimental prototype was built, and the experimental results indicate strong agreement with theoretical analysis. All switches realize ZVS within the entire power range and maintain high efficiency >= 94.1%) in the power range of 16.7%-100%.

As global electric vehicle (EV) adoption accelerates, granular analysis of empirical usage and charging patterns remains scarce. This study presents a unique large-scale empirical examination of 1.6 million EVs, including a broad array of vehicle types—private, taxi, rental, official, bus, and special purpose vehicle—across seven major Chinese cities with over 854 million observations of driving and charging events. Our findings illuminate significant heterogeneity in EV usage, battery energy, and charging behavior across vehicle types with notable city differences. Day-time high-power charging presents high loads on the electricity grid across all vehicle types, particularly from service-oriented vehicles, including taxis, rental cars, and buses. The maximum loads also are the highest in the center of the cities. Our study of large-scale EV usage offers critical insights for developing charging infrastructure, managing energy grids, and providing flexibility services, which are pivotal to the evolution of future transport ecosystems.

Precise fault identification and evaluation of battery systems are indispensably required to facilitate safe and durable operation for electric vehicles. With the core objective of addressing the challenges of inaccurate evaluation and misdiagnoses of multi-fault in existing methods, this paper proposes a deep-learning-powered diagnosis and evaluation scheme for series-connected battery systems. First, we conduct series-connected cycling experiments to simulate the two most common faults including capacity anomaly fault and short circuit fault happening concurrently to observe the failure phenomena of different faulty batteries and fault-free batteries. Then, the evolutional processes of various faults are analyzed and compared for a deeper understanding of the battery fault mechanism. In addition, we establish an elaborate deep-learning-based model, achieving satisfactory realizations on predicting the reference voltage (with the mean square error of 7.84 x 10(-5) V) while categorizing the current fault state (with an accuracy of 98.2 %). At last, a comprehensive fault identification and quantification strategy is constructed to minimize the misdiagnosis. All proposed methodologies demonstrate the advancement compared to other state-of-the-art algorithms. And the results are thoroughly validated with two different experimental datasets and real-world cloud vehicle datasets, affirming the efficiency and practical applicability, contributing to enhancing the active safety capabilities of battery systems.

In wireless charging applications for electric vehicles (EVs), multireceiver inductive power transfer (IPT) systems hold enormous potential. However, the coupling and charging rates between vehicles exhibit significant disparities and unpredictable variations, which causes the systems' efficiency to deviate from their optimal state. This article presents an analytical model for solving optimal control variables, which considers arbitrary loads and couplings. Based on the proposed model, with further consideration of coupling and load restriction, the feasible regulation trajectories of the primary inverter and secondary active rectifier phases are derived. Besides, a maximum efficiency point tracking (MEPT) control strategy is designed for achieving selective power distribution and constant current output characteristics simultaneously. Finally, an IPT experiment with dual loads is designed and carried out. The experimental results show that the desired power distribution can be maintained for each receiver (RX) even in the face of load voltage, demand current, and coupling variations. Furthermore, the efficiency of the proposed system can be improved by up to 3%, reaching a maximum of 89.13% compared to the system without MEPT.

This paper proposes a non-isolated converter topology capable of achieving a wide output voltage range (200 V to 1000 V). This design addresses the limitations of traditional two-stage isolated and single-stage chargers, where the use of high-frequency resonant converters often restricts the power density and narrows the output voltage range. The proposed topology integrates a buck-stage rectifier circuit with a three-level DC/DC boost stage. To ensure high-efficiency operation and prevent redundant high-frequency switching, a coordinated modulation strategy is developed. When the system operates in buck mode, the high-frequency switches in the boost stage are clamped; when in boost mode, the buck-stage switches are clamped; and during transition mode, both stages coordinate their switching behavior based on the required output voltage. This modulation strategy ensures that the minimum number of high-frequency switches are active at any given time, thereby reducing switching losses caused by hard switching and enabling a smooth transition across the entire output voltage range. Simulation results validate the feasibility and effectiveness of the proposed topology and control method.

As the growing deployment towards transportation electrification, a critical focus has emerged on quantifying the reduction contribution of greenhouse gas emissions from electric vehicles towards achieving carbon neutrality under diverse policy scenarios in the future. This necessitates a dynamic model that captures the evolving composition of the vehicle fleet and accurately forecasts the penetration and developmental trajectory of the electric vehicles in the car market. However, previous studies have largely overlooked the heterogeneity in user usage attributes, rendering them less effective in evaluating the impact of usage-based incentives on electric vehicle market penetration. To bridge this research gap, this study introduces an innovative, data-driven framework that integrates system dynamics and agent-based model. The proposed model can predict levels of electric vehicle penetration and corresponding greenhouse gas emission reductions within the private passenger vehicle sector, under a variety of policy scenarios. Our findings indicate that usage-based incentives, when implemented with optimal intensity, yield more significant emission reduction impacts and long-term economic benefits compared to conventional purchase-based subsidy. These insights not only furnish actionable policy suggestions to expedite the electric vehicle industry's growth in China but also offer valuable implications for other countries seeking to implement effective strategies for combating climate change and fostering sustainable transportation initiatives.

The energy efficiency of battery electric vehicles has reached a high level, but there remains significant potential for optimizing the efficiency of battery thermal management systems (BTMS). Existing control strategies for BTMS often fail to adequately consider preview information of the trip, leading to suboptimal energy efficiency. In this paper, an online Markov Chain (MC)-based model predictive control (MPC) strategy is used to reduce the energy consumption of the battery thermal management system. The MC-based speed predictor gathers speed data during driving to continuously update the transition probability matrix(TPM), thereby enhancing prediction accuracy. Firstly, the predicted vehicle speed sequence is used as a disturbance of the model predictive control. Then the battery degradation cost, electric power consumption cost, and deviation of battery temperature from target are selected as the objective functions, and finally the dynamic programming (DP) algorithm is employed to solve the local optimization problem within the prediction time domain. For simulation verification, the CLTC-P cycle was repeated 17 times as a test condition. Compared with the rule-based controller, the MPC algorithm reduced power consumption by 5.4% and battery degradation by 3% under test conditions. In addition, the BTMS with MPC tracks the battery reference temperature better.

China, the world leader in automobile production and sales, confronts the challenge of transportation emissions, which account for roughly 10% of its total carbon emissions. This study, utilizing real-world vehicle data from three major Chinese cities, assesses the impact of Battery Electric Vehicles (BEVs) on air quality. Our analysis reveals that BEVs, when replacing gasoline vehicles in their operational phase, significantly reduce emissions, with reductions ranging from 8.72 to 85.71 kg of CO2 per vehicle monthly. The average monthly reduction rate is 9.47%, though this effect is less pronounced during winter. Advanced BEVs, characterized by higher efficiency and newer technology, exhibit greater emission reduction benefits. While private BEVs generally contribute positively to environmental outcomes, taxi BEVs, due to their intensive usage patterns, show less environmental advantage and may sometimes worsen air quality. Looking ahead, we project substantial emission reductions from the replacement of gasoline vehicles with electric alternatives over the next decade. Policymakers are urged to adopt proactive measures, focusing on promoting medium to large electric vehicles and fostering the use of private and ride-hailing electric vehicles.

The efficiency of the wireless power transfer (WPT) system is highly dependent on the load resistance. To meet the demand of wide load variation, traditional methods employ dual-side control with multivariable to improve the efficiency for light loads. However, these methods often necessitate high-frequency signal synchronization and involve complex control. Therefore, a crucial challenge in WPT systems is to maintain high efficiency over a wide load range while using a simple control method. To address this research gap, a tandem-half-bridge (THB) WPT converter is proposed to improve the system efficiency and reduce the switching drain-source voltage. Based on the THB WPT converter, a hybrid modulation strategy is proposed, and only pulse frequency modulation or phase shift modulation is used in each operation mode. Additionally, a unified pulsewidth modulation generated method for three modes is given, which can be better applied in the digital controller. To validate the effectiveness of the proposed THB WPT converter under the hybrid modulation strategy, we construct a 1.8-kW experimental prototype. The experimental results demonstrate a peak efficiency of 95.26%. Furthermore, the efficiency under light load is up to 10.36% higher than the traditional full-bridge and half-bridge (FB&HB) converter under the mode switching strategy.

Hundreds of electric vehicle (EV) battery thermal runaway accidents resulting from untreated defects restrict further development of EV industry. Battery defect detection based on the abnormality of external parameters is a promising way to reduce this kind of thermal runaway accidents and protect EV consumers from fire danger. However, the influence of temperature and EV state i.e., charging and driving on battery characteristic will complicate the method establishment. Existing data-driven methods are continuously falsely judging normal batteries to be defected, which will cause panic of EV occupants. To cope with the issue, a Precision-concentrated battery defect detection method crossing different temperatures and vehicle states is constructed. The method only utilizes sparse and noisy voltage from existing onboard sensors. Firstly, a density-based semi-supervised cluster method (DBSSC) is proposed containing three novelties: The objective function is originally defined and a multi-layer L-shaped optimization method is proposed to improve the Precision; Six kernel-domains are proposed to cope with the arbitrary distribution of battery voltages; The soft boundary is designed to consider the random noise in real-world EV operation. Subsequently, the DBSSC is trained by real-world data of different EV states and temperatures to enhance the robustness. The training process only utilizes data of normal batteries to cope with the inadequacy of thermal runaway battery data. The results show that the method can detect defected batteries 13 days ahead the thermal runaway while achieve the Precision of 99.2%. By the three novelties and training by data of different conditions, the Precisions are improved by 40.9%, 3.4%, 7.0%, and 12.0% respectively.

To capture the charging station dynamics caused by uncertain user behavior and photovoltaic power generation(PV), we propose a deep reinforcement learning(DRL) approach for optimizing the energy flow paths of electric vehicle charging station(CS). By aggregating the charging state of individual electric vehicles(EVs) into characteristic parameters of CS, we aim to reduce the computational burden of time-varying dimensions of state space caused by the uncertain users charging behaviors. The proposed approach facilitates charging scheduling by allocating charging power to single EVs. Through numerical simulations using real-world data, we demonstrate that the proposed DRL approach achieves on average 13% higher profit for CS compared to the rule-based approach, while also reducing the impact of power fluctuation on grid.