The balancing technology is important for battery management systems. However, the widely used passive balancing technology has disadvantages such as a low efficiency and a long balancing time, while the active balancing technology which is more efficient has problems such as complex control and high cost. In response to the status quo, a double-layer balancing topology circuit based on an improved Buck converter is proposed. Its circuit parameters and balancing strategy are designed by taking into account the issue of matching, its working principle is analyzed in detail, and its economic cost is compared and analyzed, so as to verify the feasibility and economy of the proposed scheme. The simulation analysis of a model based on MATLAB/Simulink shows that the balancing time of the proposed scheme is reduced by 40.4% compared with that of the traditional adjacent cell balancing system. The results of a battery pack balancing test show that this scheme has advantages of a simple structure and simple control. It can quickly realize the balancing of 26650 lithium iron phosphate battery, and it only took 20.4 min to reduce the maximum differential voltage of four battery packs in series from 141 mV to within 50 mV.

The remaining useful life can monitor the instantaneous state of the performance and safety of lithium battery. Therefore, controlling the remaining useful life of lithium battery is of significance for the safety of battery use. In response to the low predictability of high-frequency oscillation components in the initial decomposition of lithium battery capacity, a nonlinear and non-stationary signal processing method based on the secondary decomposition of high-frequency signals combined with machine learning prediction models is proposed. First, the original signal is decomposed by the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) algorithm. Then, the high-frequency random disturbance signals are further decomposed using the variational mode decomposition algorithm, followed by the prediction of components using the long short-term memory (LSTM) network. Experimental verification based on four NASA datasets shows that the proposed method has high prediction accuracy. Compared with those obtained by the CEEMDAN-LSTM model for single decomposition prediction, the value of R² of the sequence prediction is improved by 1.25%, while the values of MAE, RMSE and MAPE are reduced by 39.34%, 49.43% and 41.44%, respectively. The proposed method is robust for the effective prediction of high-frequency signals, as demonstrated by experiments on two CALCE datasets.

Aimed at the poor estimation performance of single algorithms in estimating the state-of-charge (SOC) of lithium-ion batteries, a hybrid algorithm which combines the advantages of deep extreme learning machine optimized by the aquila optimizer (AO-DELM) and the extended Kalman filter (EKF) algorithm is proposed for the SOC estimation of lithium-ion batteries. First, the recursive least squares method with forgetting factor is used to identify the parameters of the battery model. Then, based on the data under a high acceleration cycle (US06) condition, the AO-DELM is trained using the current, voltage and various battery state features extracted by the EKF algorithm as input and the actual SOC as output. Finally, the trained AO-DELM model and the EKF algorithm are used to form a closedloop real-time SOC estimation system. Results show that compared with the EKF and DELM-EKF algorithms, the proposed AO-DELM-EKF algorithm significantly improves the SOC estimation accuracy under four dynamic conditions, with the root mean square error not more than 0.49%.

Aimed at the shortcomings of one single energy storage system in terms of power and energy demand, as well as the problems of low management efficiency and poor economy, the impact of energy storage filtering time constant on the working efficiency is analyzed, with a focus on the traditional hybrid energy storage crossover control method. In addition, a multi-mode coordinated control strategy based on the monitoring of state-of-charge (SOC) is put forward. By dynamically adjusting the filtering time constant and dividing the energy storage into different working areas, the flexibility of energy storage power allocation in DC microgrids is realized to improve the working efficiency. At the same time, according to the source-charge power difference and the SOC of energy storage, four operation modes are designed to optimize the source-charge-storage coordinated control strategy, so as to reduce the number of battery charging and discharging times and improve the system economy. The results of actual microgrid tests show that the proposed control strategy has a higher efficiency and better economy, indicating its value for engineering popularization and application.

The state-of-health (SOH) and remaining useful life (RUL) of batteries are important indicators for measuring their performance and conditions, and an accurate prediction of these indicators has high practical value. The electrochemical impedance spectroscopy (EIS) test of batteries can reveal the electrochemical reactions and processes inside them, providing important information for evaluating their SOH. EIS can be used to predict SOH and RUL, but it has high testing costs and limited data volume. The use of variational autoencoder (VAE) for data augmentation can expand the volume of EIS data, which helps improve the model’s generalization capability and reduce the risk of overfitting. Due to the fact that the EIS data is usually high-dimensional and may contain redundant information, principal component analysis (PCA) can be used to reduce the dimensionality of data while retaining the key information. Convolutional neural network (CNN) can recognize the spatial features in the EIS data, while bidirectional long short-term memory (BiLSTM) network can capture features within the time series data. The Attention mechanism can enhance the model’s understanding of data contextual relationships. The research results indicate that the model combining PCA, CNN, BiLSTM network and Attention mechanism can more effectively predict the SOH and RUL of batteries with the addition of augmented data. The proposed EIS data augmentation method can also be applied to other models to improve the predictive performance.

Dry-type core reactors are usually employedas key equipment to achieve reactive compensation in electric power systems on a large scale. However, the vibration noise caused by the magnetostriction effect of core becomes a dominant factor that restrict their applicationsto indoor substations in urban areas. Soft magnetic composites (SMCs) with low magnetostriction and silicon steel were used together as the core material of reactors based on the magnetic flux shunt principle. With the consideration of difference in the efficient permeability between two kinds of materials, the flux in the core is reasonably regulated, so that the larger portion of the flux is carried by the silicon steel component, while the smaller portion flows through the part of SMCs. Based on the design idea, a hybrid-magnetic-circuit (HMC) reactor prototype was designed and fabricated. The research results show that the core losses of HMC reactors were less than those of the pure silicon steel reactor, and thenoise of HMC reactors was less than 60 dB (A), which was lower than that of the silicon steelreactor.

To quickly and accurately calculate the winding losses of Litz-wire in a transformer with a CLLC resonant circuit, the orthogonal relationship between the skin effect and proximity effect is utilized to divide the winding losses into those caused by the skin effect and proximity effect, so that these losses can be calculated separately to avoid errors caused by end effects in the windings. First, the duty cycle is used to calculate the losses caused by the proximity effect. By analyzing the current expressions for the transformer with a CLLC resonant circuit and further combining them with the formulas for losses caused by the skin effect and proximity effect, an empirical formula for calculating the winding losses of the transformer with a CLLC resonant circuit is derived. Finally, a transformer with a CLLC resonant circuit was designed and fabricated, Maxwell simulations and experimental measurements were conducted for validation, and the simulated and measured values were also compared with those calculated by a modified Tourkhani model. Results show that the proposed model can accurately calculate the winding losses of Litz-wire in the CLLC transformer, enabling precise prediction of winding losses of Litz-wire in transformers with CLLC resonant circuits.

The AC-DC microgrid can effectively facilitate the applications of renewable energy sources. However, the strong electrical coupling between the AC and DC sides caused by an interconnected converter complicates the fault characteristics within the system, threatening the system’s safe operation. The traditional fault detection schemes based on the fault characteristics may not have enough sensitivity or rapidity to meet the system’s fault detection and protection requirements. Notably, the system exhibits significant fault characteristics before blocking the interconnected converter’s IGBTs in the AC-DC microgrid. Therefore, with the combination of grounding schemes for the system and the AC-DC microgrid’s electrical coupling characteristics, the feasibility of a novel fault detection scheme for the AC-DC microgrid based on the Hilbert-Huang transform (HHT) spectrum and energy amplitude is explored, which is aimed at the AC-DC system’s characteristics during the transient process before blocking the IGBTs under fault. Through simulation verification, it is found that the proposed HHT-based fault detection scheme can realize rapid and accurate identification of fault types for the AC-DC microgrid within 2 ms, which effectively improves the safety and reliability of the microgrid.

In a modular multilevel converter (MMC) system, there are numerous submodules, and switch faults and voltage sensor faults occur frequently. Some of these faults share similar characteristics, which can affect the system’s operation reliability. A joint diagnosis method for submodule IGBT open-circuit faults and voltage sensor faults based on a sliding mode observer is proposed. First, a fault detection strategy is designed using the actual and measured values of the MMC system’s arm voltage to detect faults in the MMC system and further achieve fault localization at the arm level. On this basis, the sliding mode observer is used to perform online estimation of the capacitor voltage within the faulted arm submodule, and the estimated values are compared with the measured values for submodule-level fault localization. Finally, simulation results verify the effectiveness and robustness of the proposed fault diagnosis algorithm.

Flyback converters in consumer and commercial products must adhere to strict regulatory standards for conducted and radiated electromagnetic interference (EMI). Managing EMI has become increasingly complex in modern power electronics, particularly with the integration of high-speed wide bandgap (WBG) devices into compact system layouts. A review of established modeling techniques and mitigation strategies for conducted EMI is presented, focusing on differential mode (DM) and common mode (CM) noise, alongside radiated EMI in flyback converters. The discussion encompasses solutions at both component-level design and converter system optimization.