Predicting the remaining useful life (RUL) of energy storage lithium-ion batteries is a prerequisite for the stable operation of new energy storage systems. A lithium-ion battery RUL prediction method is proposed based on dual improved complete ensemble empirical mode decomposition with adaptive noise (DICEEMDAN) and hybrid deep learning. To address the global nonlinearity of the capacity data, DICEEMDAN is used for denoising and reconstructing the training set into multiple subsequences. An improved bidirectional temporal convolutional network (IBiTCN) then extracts features from these subsequences, which serve as input to a differentiated deep neural network (DNN). An optimized weight integration method is proposed to superimpose the outputs of different networks to obtain the final prediction results. Finally, experimental validation using actual lithium-ion battery aging data demonstrates that the proposed DICEEMDAN-IBiTCN-DNN model achieves higher prediction accuracy than benchmark network models while significantly reducing model convergence time.
As an important protective device in power systems, the fault diagnosis of high-voltage circuit breakers directly impacts the stability and safety of the power system. Traditional fault diagnosis methods often rely on human experience or simple numerical analysis. However, these methods are often ineffective in diagnosing equipment faults under complex conditions. Therefore, improving the accuracy and efficiency of fault diagnosis has become one of the key issues in the intelligent operation and maintenance of power system equipment. A fault diagnosis method is prposed for high-voltage circuit breakers based on B-spline optimized Hilbert-Huang transform (HHT) and firefly algorithm (FA)-optimized radial basis function neural network (RBFNN). First, to address the boundary effects and overfitting problems that may arise in traditional signal processing methods, traditional cubic spline interpolation is replaced with B-spline interpolation and the signal envelope construction process is optimized in the Hilbert-Huang transform. B-spline interpolation effectively reduces boundary effects in the signal, improves the precision of intrinsic mode function (IMF) decomposition, and provides a more accurate foundation for subsequent feature extraction. Second, based on the vibration signal feature extraction of high-voltage circuit breakers, a FA-optimized RBFNN is designed. This method improves fault type classification performance by optimizing the structure and parameters of the RBFNN. The firefly algorithm, by simulating the mutual attraction behavior of fireflies, performs a global search for the optimal RBF center, avoiding the local optimal problems encountered in traditional methods, thus enhancing the robustness and accuracy of RBFNN in high-voltage circuit breaker fault diagnosis. Experimental results show that the proposed method achieves an average recognition accuracy of 96.8% and an average precision, recall, and F1-score of 97.0%, demonstrating its effectiveness and robustness in high-voltage circuit breaker fault diagnosis.
The thermal imbalance in modular multilevel converters (MMC) is investigated by this study. First, the mathematical relationship between thermal equilibrium and the control mechanism is analyzed. Second, an optimization control strategy based on the dynamic retention factor (DRF) is proposed. DRF consists of the temperature deviation coefficient η and the submodule capacitor voltage adjustment coefficient a. Through predictive calculations of temperature, the temperature deviation is determined. By setting submodule capacitor voltage thresholds (5%–10% of the rated value), capacitor voltage adjustment is set. Simulation and experiment results demonstrate that this approach reduces the thermal imbalance. Finally, the research reveals the correlation between DRF and submodule thermal equilibrium. The dynamic adjustment of switching priorities achieves temperature balance between different submodules. It reduces the maximum temperature difference to within 2C. These improvements enhance system reliability. It provides technical support for MMC engineering applications.
Driving conditions and styles significantly affect the energy economy and battery lifespan of hybrid energy storage system (HESS) for electric vehicles (EVs). An adaptive wavelet transform (WT)-based energy management strategy (EMS) that considers driving style is proposed to extend battery lifespan and reduce energy loss. First, the dynamic models of the battery, ultracapacitor, and DC/DC converter are established in detail. Second, driving style recognition (DSR) is implemented by analyzing the jerk based on driving cycle recognition (DCR), which is divided into conservative, normal and aggressive. Next, a WT decomposition level adaptive method is designed to separate the transient power online, which can improve the adaptability of the EMS to driving conditions and driving styles. Finally, fuzzy logic control (FLC) is employed to achieve optimal power allocation. Simulation results indicate that compared with the traditional WT-based EMS, the adaptive-level WT-FLC-based EMS has significant advantages in extending the battery lifespan and saving energy. It can effectively suppress the maximum battery current by 58.09%, extend the driving range by 1.49 km at a constant speed of 40 km/h and reduce power fluctuations by 38.04%.
Wireless power transmission technology can be used to solve the complicated wired charging and fixed station charging problems of unmanned aerial vehicles (UAVs). In this paper, a combined 5D transmitting coil structure consisting of four-cell D-shaped coils and one compensatory D-shaped coil is proposed. By combining different D coils according to the switching strategy, the rated power transmission can be achieved in the entire transmitting region; thus, the number and position of loads become flexible, which increases the redundancy of the system. Owing to the multiple loads with multiple-frequency characteristics, cross-coupling should be considered when accessing a system. Owing to the structure of the loose coupler, the power fluctuation between multiple loads is low in the charging region, which means that cross-coupling is negligible. An experimental prototype is constructed to verify its structural characteristics. The experimental results showed that the combined 5D transmitting coil structure can supply energy to a single load in the entire transmitting region by combining different coils with high efficiency or simultaneously supplying energy to multiple independent loads with different frequencies at different positions. This verifies the anti-offset ability and redundancy characteristics of the system.
