Telecom base stations rely on a continuous and stable power supply for efficient operation. Batteries are the preferred choice to provide backup power during outages. With recent advances in the Internet of Things (IoT) and communication networks, telecom base stations have become complicated cyber-physical systems. However, the sharing of control signals and security-critical information over communication channels makes battery energy storage systems (BESS) vulnerable to cyberattacks. To solve this vulnerability, this study explores the role of digital twins (DT) in detecting and mitigating cyberattacks. To this end, we propose a data-driven battery digital twin architecture to characterize battery performance over a wide range of operating temperatures and state-of-charge (SOC) levels. The digital twin leverages a feedforward neural network to estimate SOC from current, voltage, and temperature measurements. Stealthy attacks are formulated using a Soft Actor-Critic (SAC) reinforcement learning agent to corrupt current and voltage sensors, which affect the estimation of SOC. The attacks are detected by comparing the SOC estimated by the data-driven model and the Coulomb counting method.

In electric vehicles and stationary storage applications, the increasing demand for Lithium-ion battery (LIB) systems highlights the need for precise state-of-health (SOH) estimation. Traditional approaches often rely on resource-heavy computations or impractical setups, such as controlled charging or high-frequency sampling, misaligned with the constraints of large-scale battery management systems (BMS) that operate with limited computational power and low-frequency data collected over seconds. This paper proposes a computationally efficient algorithm, using a nonlinear input-output neural network (NIONN) for state-of-charge (SOC) estimation and internal resistance derived from voltage drops during current transients for SOH estimation. Furthermore, temperature compensation through polynomial fitting enhances the accuracy of the estimate. Validated on the NASA battery aging dataset, the proposed approach achieves an average error below 3%. With only 41 arithmetic operations and adaptability to diverse conditions. The proposed method with real-time SOH monitoring for electric vehicles and large-scale energy storage offers a practical, robust solution tailored to BMS limitations.

Electric drives featuring high-speed operation use a combination of modulation techniques; each is suitable for specific operating conditions. Furthermore, a smooth transition between different modulation schemes is necessary to avoid over-current and over-voltage issues during transitions. This paper proposes a multi-mode scheduler that combines space vector pulse-width modulation (SVPWM), generalized discontinuous PWM (GDPWM) with edge- and center-aligned triple-, five-, and seven-pulse modulation approaches. Moreover, the proposed scheduler includes a transition handler designed to ensure seamless transitions while minimizing current and torque fluctuations. We validate the performance of the proposed method on an oil pump motor used in automotive transmission systems, tested on a dry motor bench. The results demonstrate that the proposed approach can decrease DC-link current fluctuations by up to 54% of the maximum deviations. In situations where direct transitions are impractical, a tentative transition strategy enables steady and controlled mode transitions.

Incorporated across diverse applications ranging from traditional energy utilization to cutting-edge instrumentation, multilevel converters encounter a fundamental trade-off between fidelity and switching loss or module quantity. Amidst this dilemma, asymmetrical multilevel circuits emerge as a compelling solution, offering precision with a reduced number of modules. However, prevailing asymmetrical patterns often exhibit significant voltage deviation between modules, posing implementation challenges. In this study, we propose a new framework of asymmetrical multilevel circuits, which encompasses enhanced geometric schemes and novel patterns such as arithmetic distribution and optimization-based irregular schemes. We evaluated the output fidelity of all considered asymmetrical schemes across various output profiles. We critically examine the rationale behind existing patterns, particularly the binary setup. The newly proposed framework guides the design of asymmetrical multilevel circuits towards achieving high fidelity with acceptable voltage differences.

Modular multilevel converters have developed into a door opener for a variety of energy applications. Enabling parallel connectivity between modules can offer several advantages to systems, including sensorless balancing, module-to-module energy exchange, reduced impedance, low-frequency operation down to dc output, and a larger effective capacitance. Existing module circuits, however, require many individual switches or offer only very limited parallel functionalities. We propose a new module topology, which we name the direction-selective parallel structure, which requires only a minimum of transistors. A combination of only four active switches and four diodes enables bidirectional equilibration, bipolar module output, and inter-module switched-capacitor features. The topology is highly attractive for existing converters working with cascaded H-bridge elements as the addition of only four diodes enables the key features of more complicated module topologies, particularly their superior balancing, such as sensorless estimator-free balancing, and module-to-module energy exchange. The diodes furthermore well match the insulated gate bipolar transistors implemented in most larger systems. We elaborate on the working principle of the direction-selective parallel module and demonstrate its operation.

Magnetic windings, in general, and small drives, in particular, are typically associated with thin round copper wires. This group of small drives includes electrical machines for automotive applications, ranging from ancillary units to traction machines for both hybrid electric vehicle (HEV) and battery electric vehicle (BEV)[1],[2]. Wire-wound machines can refer to well-established techniques for widely automatic manufacturing except for traction machines with distributed windings, which still contain manual steps in most assembly lines, particularly after the insertion process[3]. Machines typically wind the loops of continuous wires on a bobbin with a linear or flyer-winding technique outside the stator and pull them from one side of the stator to the other into the slots. The overhang on both ends of the stator, the so-called end turns, forms automatically from the continuous loops.

This paper proposes and implements a stepless frequency regulation method for wireless power transfer (WPT) with a simple LC compensation network to provide load-independent transmitter current. It should be mentioned that the capacitor in LC compensation network is controlled by a switch. Different from previous switched capacitor compensation networks, which can only select a limited number of resonant frequency points, this model can choose the resonant frequency arbitrarily in a wide frequency range. In theory, the frequency range can be controlled from a few kilohertz to hundreds of kilohertz. The key is controlling the turn-on time of the switch and adjusting the activation time length of the controlled capacitor in every WPT cycle. Therefore, the proposed system is very suitable for multi-frequency WPT, as the system can work at full resonance with practically arbitrary frequency. A Simulink simulation serves for verification, and the results prove the system can change the resonant frequency from 100 kHz to 300 kHz continuously while keeping the transmitter current load-independent.

Electric vehicle (EV) charging infrastructure will play a critical role in decarbonization during the next decades, energizing a large share of the transportation sector. This will further increase the enabling role of power electronics converters as an energy transition technology in the widespread adoption of clean energy sources and their efficient use. However, this deep transformation comes with challenges, some of which are already unfolding, such as the slow deployment of charging infrastructure and competing charging standards, and others that will have a long-term impact if not addressed timely, such as the reliability of power converters and power system stability due to loss of system inertia, just to name a few. Nevertheless, the inherent transition toward power systems with higher penetration of power electronics and batteries, together with a layer of communications and information technologies, will also bring opportunities for more flexible and intelligent grid integration and services, which could increase the share of renewable energy in the power grid. This work provides an overview of the existing charging infrastructure ecosystem, covering the different charging technologies for different EV classes, their structure, and configurations, including how they can impact the grid in the future.

The paper proposes an Asymmetrical Modular Multilevel Converter (AMMC) suitable for low/medium-voltage dc-ac conversions with very high output quality. The modules' dc-links of the AMMC are charged to a binary exponential sequence to produce a large number of output levels using only a few modules.The concept of using asymmetrical dc-links for high-quality output is not entirely new. However, the practicality of existing approaches is relatively low and challenged by the difficulties in maintaining the required dc-link voltages as well as suppressing their interaction with the output, which often requires multiple isolated dc/dc converters. We solve this problem by aligning the modules in the Marquardt MMC inverter module configuration that offers more control freedom, hence the term AMMC. Furthermore, we introduce a highly effective switched-inductor charge transfer and balancing mode between modules and even across arms. We accordingly modify the underlying conventional chopper modules so that the dc-link voltage control can be completely sensorless. The proposed AMMC is tested in a lab setup with four modules per arm reaching 32 output levels. In contrast to the low benefit of an additional module in MMC due to only linear improvement of the output granularity, each further module halves the finest voltage step. The components to maintain the graded voltage sequence and the underlying inductive charge transfer only a fraction (< 10%) of the load current so that relatively low-power devices can be used.

The electricity system, particularly the low-voltage grid is under immense pressure due to rapid growth of high-power loads and sources, such as vehicle chargers and photovoltaic installations, which can locally overload lines and generate both over- and under-voltage conditions in the same grid section so that conventional voltage adjustment through tap changers fails. Soft-open points (SOP) with either full ac–ac conversion or conventional series voltage injection through transformers with universal power flow and/or quality controllers (UPFC, UPQC) are studied and promoted but are large due to either the transformer or the full-power electronics, in case of a transformer injection also limited in dynamics, and costly. We present a direct-injection circuit for soft open points, universal power flow and quality control, which exploits the recent developments in high-current low-voltage transistors. The centre piece is a low-voltage (e.g., 48 V) module connected in series with each phase, which is floating with the grid voltage so that it only needs to deal with a small voltage between its outputs and can maintain that condition also under fault currents of several kiloampères and worst-case enforce 0 V. Due to the low-impedance of the distribution grid, only a few volts are sufficient to control several hundred ampères in active and reactive current in a loop. Thus, the circuit only needs to control a fraction of the controlled power in contrast to full ac–ac concepts. The direct injection eliminates the bulky transformer with its dynamic limitations. In contrast to commercially available transformer-injection-based SOP solutions with often considerable additional loss, the injection modules have almost negligible additional impedance (< 4 mΩ), which is particularly important in the low-voltage grid.

Lithium-ion (Li-ion) batteries have attracted significant attention in terms of technical features, but to use them within their specific operating region and prevent undue degradation, it is crucial to constantly monitor their state of charge (SOC). However, most available techniques are either too complex, too computationally demanding, or require a large measurement window. This paper proposes a general framework for the data-driven SOC estimation and demonstrates its effectiveness via a very simple Neural Network (NN) classifier. Accordingly, a nonlinear input–output neural network (NIONN) is developed based on the proposed network and then further optimized. Additionally, we present an input engineering process using available data to provide a good accuracy while considering the practical aspects such as complexity as well as the update speed for a low-cost online application. The numerical results from the measurements validate the performance of the proposed general framework. Per extensive optimizations, it is proven that a shallow network structure with 13 neurons in the first hidden layer and 12 neurons in the second layer can achieve <98.3% accuracy and a mean squared error of <10 −5 which is on par with more complex networks with higher neurons and/or hidden layers. Additionally, the optimum observation window was determined to be 5 seconds.

Modular reconfigurable batteries, also known as smart batteries, are gaining significant traction, mainly due to the large environmental incentives and falling price of electronic components. Although they have many advantages compared to a hard-wired battery pack, complex monitoring circuit and numerous sensor requirement make it harder to compete with conventional systems in a cost-driven application. This paper proposes a novel approach to estimate parameters of each individual battery module without any direct measurement at their terminals. The proposed algorithm uses the output voltage and current of the load combined with the exact knowledge of the modules’ states to estimate the open-circuit voltage, ohmic resistance, and polarization resistance in the electric circuit model for each battery module. The method combined with Kalman filter demonstrates the feasibility of this method through simulations, where the proposed method achieves above 98% and 96% accuracies for estimation of the open-circuit voltage and equivalent resistance of the battery, respectively. Additionally, the method can decouple the two resistances with

Modular multilevel converters (MMCs) are a well-known solution in high-voltage applications. Recently, new topologies such as MMCs with series/parallel (MMSPC) connectivity are attracting large amount focus, due to their improved efficiency as well as self-balancing capability. Although MMSPCs can operate without direct measurement of modules’ voltages in battery-integrated modules, monitoring of the voltage and resistance of each module can still provide useful information to assess the state of each battery. However, including a voltage and current sensor in each module is not a cost-effective approach. This paper proposes a method to estimate the voltage and resistance of each battery-module using only the voltage and current measurement at load terminal. The proposed technique can reduce the number of sensors from 2N + 2 to only 2. Additionally, due to the self-balancing capability, the estimation technique does not require fast convergence and can operate in the background during the idling intervals of the processor. MATLAB simulations confirm the applicability of the proposed approach. The result show a 99 % and also 96 % accuracy for balanced and imbalanced systems.

This paper proposes a motor drive suitable for electric vehicles (EVs) with notably reduced losses as compared to conventional drives. The proposed drive feeds the dc-link with a dynamically reconfigurable battery formed by cascaded half-bridges containing battery units. The reconfigurable battery concurrently forms a modular multilevel dc-link able to synthesize variable dc-link voltages, which can adjust the dc link voltage to run a motor to operate at the optimum load point or even to contribute to modulation and greatly reduce the switching loss of the main inverter—in this case a standard two-level three-phase inverter. Up to ⅔ switching actions are avoided in the main inverter, which can run at maximum duty cycle and fundamental-frequency commutation. The saved switching effort in the main inverter is shifted to the modular multilevel dc-link, but with much less loss due to the fractionized switching voltage and the use of field-effect transistors (FETs).At high drive speeds, the converter halves the total loss compared to a standard inverter only; at lower speeds and thus smaller modulation indices, the advantage is even more pronounced because of the dynamically lowered dc-link voltage. Other benefits include alleviated insulation stress for motor windings and direct battery balancing. The proposed motor drive is verified on a down-scaled setup with eight modules constituting the dc-link.

Modular multilevel converters (MMC) show great potential in high-voltage as well as emerging applications such as grid storage and electro mobility. The main advantages over the two−level converters include flexibility, scalability, and higher degrees of freedom. MMCs with series/parallel connectivity (MMSPC) are a more recent developed technology which can provide stable and efficient sensorless operation. However, although the additional parallel connection between the modules provides many advantages, it can change the equivalent resistance of the system and hence complicate the power loss analysis. This paper presents a simplified conduction loss calculation method based on equivalent resistance analysis for MMSPC. The presented method has the advantage of simplicity with relative accuracy that can speed up the process of loss analysis. Simulation results show that the accuracy of the presented method is above 95% percent for modulation indices above 0.5.

Reducing the computational burden and improving the accuracy are in the two opposite sides of every electrical equivalent circuit model (EECM) for batteries. In this paper, a novel identification method is developed to estimate EECM parameters for any Li-ion battery with the aim of reducing computational burden and improving the accuracy of estimation under high C-rates of charge/discharge cycles for electric vehicle (EV) applications. The proposed parameter identification method is implemented on the second-order EECM. A step by step execution of the new identification method is presented which is based on circuit analysis and Particle Swarm Optimization (PSO). A comparison is carried out between obtained and the Pseudo-Two-Dimension (P2D) electrochemical model results. Moreover, experiments are carried out on two Li-ion battery cells, NCR18650 Panasonic and Mp176065 to verify the accuracy of the proposed method. The included results demonstrate the performance of the proposed method regarding improving accuracy and applicability as well as reducing required memory.

The advantages of modular multilevel converters (MMC) over conventional two-level converters have lead to new emerging topologies and applications. The MMC with series/parallel connectivity (MMSPC) is one of the more recent topologies that can provide stable and efficient sensorless operation by introducing an additional parallel connection state between modules. However, while the parallel functionality has great potential in simplifying control and monitoring, it complicates the analysis of the system. The equivalent resistance of the MMC can be helpful in analyzing the conduction losses and also in designing appropriate heat management systems. Estimating the equivalent resistance of an MMC structure with half-bridge submodules is a straightforward procedure. However, performing a similar analysis for MMSPCs is more difficult, mainly because of the parallel state. In this paper, a simplified semi-analytical equivalent-resistance analysis method for MMSPC with full-bridge submodules (SM) is presented that can increase the speed of analyzing the system. Furthermore, the derived equations are compared to MMC with half-bridge SMs to provide better insight into the effect of varying parameters of the system.

The Modular Multilevel Series Parallel Converter (MMSPC) not only allows dynamic interconnection among the modules in series connection, like the traditional Modular Multilevel Converter (MMC), but also in parallel connection. This extra switching state among the modules allows new features and operation modes. The parallel connection between modules enabled by the MMSPC, is not only limited to those modules belonging to the same arm, but can also be extended to the parallel connection of modules from different arms or even from different phases. All these different interconnections brings operational benefits as well as some constraints. In this paper, the power topology, operation and comparison of different MMSPC configurations are presented. Specifically, this paper describes in detail the traditional MMC (without parallel interconnection), the MMSPC with parallel connection of modules of the same arm only, the MMSPC with the option of paralleling connection among the phases of the system, the MMSPC with parallel connection between arms of the same phase and finally, the MMSPC with full-parallel capability.

The modular multilevel converter with series and parallel connectivity was shown to provide advantages in several industrial applications. Its reliability largely depends on the absence of failures in the power semiconductors. We propose and analyze a fault-diagnosis technique to identify shorted switches based on features generated through wavelet transform of the converter output and subsequent classification in support vector machines. The multi-class support vector machine is trained with multiple recordings of the output of each fault condition as well as the converter under normal operation. Simulation results reveal that the proposed method has high classification latency and high robustness. Except for the monitoring of the output, which is required for the converter control in any case, this method does not require additional module sensors.

The star-configured modular multilevel converter is a promising topology for grid-connected storage as well as power-quality applications and motor drives. During operation, the ac output power generates substantial current pulsation at twice the output frequency in the modules, which causes additional loss and requires an increase of the module capacitance. This paper presents a common-mode voltage injection method that reduces the ripple current by 30%-50% compared to conventional methods under a wide range of load conditions. We further propose to replace the conventional H-bridge modules with double-H-bridge modules, which allow direct power exchange between the arms through the neutral point to smoothen power pulsations. The double-H bridge module can suppress the ripple current by 70%-80%.