Iet Electric Power Applications最新文献

Advancement of renewable energy technologies placed tidal, wave and wind energy systems at the forefront of sustainable power generation. Permanent magnet synchronous generators with high efficiency, modularity, and low power consumption, such as the lightweight and modular C-GEN design, have been applied successfully in these applications. The demagnetisation condition of nonadjacent magnets in direct-drive permanent magnet generators is investigated, and different diagnostic approaches are evaluated. It is shown that nonadjacent demagnetisation can produce false negative diagnostic alarms when familiar condition monitoring methods such as the MCSA are applied. It presents solutions for faulty magnet detection of permanent magnet generators for tidal, wave, and wind energy harvesting, supported by numerical analysis and experimental testing. It is particularly novel that demagnetisation in two nonadjacent magnets is investigated here, something not previously considered.

The interior permanent magnet integrated motor (IPMIM) for ball mill has complex eccentricity problems. In order to solve the difficult problem of calculating the strength and stiffness of IPMIM in eccentric state, a mathematical model under eccentricity was established. Analytical solutions for the rotor stress and deformation were respectively derived for parallel eccentricity, inclined eccentricity, and composite eccentricity. By using the analytical solutions and finite element method, the stress and deformation of the rotor for IPMIM with a power of 210 kW were analysed. The results show that the calculation results obtained by two methods are highly consistent, which proves that the analytical solutions can accurately predict the stress and deformation of the rotor of IPMIM for ball mill. In addition, based on the analytical modelling, the influence of magnetic tensile stress and structural parameters on the rotor stress and deformation were explored, aiming to summarise the variation laws of the stress and deformation of the rotor.

This paper presents a fault diagnosis method based on the error current for active power filter (APF) to enhance reliability. APF have significant practical significance and application value because they can be used to solve power quality problems such as harmonics, reactive power or negative sequence existing in the power system. First, the expression of error current is educed which is analysed by the influence of different open-circuit faults on the shunt three-phase three-level APF. Based on this, we analyse the differences in error current of the harmonic compensator when it is in normal operation and under different fault conditions, and then construct fault diagnosis variables. In addition, the relationship between the value of the fault diagnosis variable and fault type is built and the fault diagnosis method is proposed to detect the open-circuit fault and locate the faulty switches. Finally, the simulation model and the experimental platform are established to verify the fault diagnosis method. The results of simulation and experiment are given to validate the proposed fault diagnosis.

Bearing is very important for motors. When a bearing fails, if the problem can be discovered and solved as early as possible, it can not only reduce the cost of repairs, but also greatly improve safety. This study proposes a machine learning-based model for diagnosing bearing faults. Regarding this model, first, the Hilbert–Huang transform (HHT) and multi-resolution analysis (MRA) in feature extraction methods are used to derive relevant features from the original signal. Then, a feature selection method based on genetic algorithm (GA) and combined with the concept of expanded search scope is used to delete redundant features. Finally, the k-nearest neighbour algorithm (KNN) and feed-forward neural network (FFNN) in the classifier are used. In addition, the University of California Irvine (UCI) datasets, Case Western Reserve University (CWRU) bearing dataset, Mechanical Failure Prevention Technology (MFPT) bearing dataset, and motor fault current signal dataset were used to validate the fault diagnosis ability of the proposed model.

In industrial applications, motor operational status is crucial for production efficiency. However, timely detection and prediction of motor faults present significant challenges, often resulting in production incidents and substantial maintenance costs. This paper presents a novel approach for assessing motor equipment health based on knowledge-embedded machine learning and statistical data evaluation. Specifically, the methodology first employs mechanism-based motor operational models and statistical methods to identify key variable parameters associated with typical operational states from extensive monitoring variables, serving as input layers for machine learning algorithms. Subsequently, the study utilises machine learning algorithms to predict labels for normal operation, phase loss faults and overload faults, incorporating health degradation levels as knowledge-embedded foundations for the health state assessment. Finally, the Comprehensive Health Index (CHI) was evaluated, achieving 98.1% health assessment accuracy on test datasets in environments with data sampling frequencies below 1 Hz and relatively low data quality. This methodology establishes relationships between health states and actual fault records through a dynamic weight allocation strategy that provides quantified percentage values, reflecting actual equipment usage patterns and degradation trends. It bridges the gap between theoretical diagnostic accuracy and practical industrial implementation requirements, providing highly robust maintenance strategies for industrial scenarios.

In urban rail train traction motors, bearings serve as critical core components whose health status directly impacts traction motor operational performance and safety. Among various traction motor fault types, bearing faults have emerged as one of the most frequently occurring failure modes. However, the frequent start-stop operations and significant passenger capacity fluctuations characteristic of urban rail trains make stable operating condition data collection challenging, which has severely limited the engineering applicability of existing bearing fault diagnosis methods. This study proposes a bearing fault diagnosis method integrating SPWVD and YOLOv11: the method converts one-dimensional vibration signals into two-dimensional time–frequency maps using the SPWVD algorithm; these maps are then processed based on fault mechanisms and input into the YOLOv11 deep learning model learning and classification. Experimental results demonstrate that this method transcends the adaptability limitations of traditional time–frequency analysis under complex operating conditions and overcomes the multi-scale feature learning bottlenecks of CNN, achieving reliable bearing fault diagnosis under constant-speed conditions while maintaining over 90% accuracy in complex scenarios such as variable speed and strong noise, thereby significantly enhancing the robustness and universality of bearing fault diagnosis methods in engineering applications.

In the field of bearing fault diagnosis, effectively extracting critical information from raw motor signals while ensuring high accuracy and minimising computational resources remains a significant challenge. This study proposes a novel bearing fault diagnosis model consisting of three main stages: feature extraction, feature selection, and classification. In the feature extraction stage, empirical mode decomposition (EMD), Hilbert–Huang transform (HHT) and fast fourier transform (FFT) are utilised to extract features from raw motor signals. In the feature selection stage, a novel hybrid feature selection method combining genetic algorithm (GA) and binary state transition algorithm (BSTA) is proposed enhancing the model's performance. This research has also added a new memory function to the algorithm to avoid unnecessary computational waste. In the classification stage, k-nearest neighbours (k-NN) and support vector machine (SVM) are employed to evaluate the classification accuracy after feature selection. To validate the performance of the proposed model, experiments were conducted on four bearing fault datasets, including the University of California Irvine (UCI) benchmark dataset, Motor Bearing Fault Current Signal Dataset, Case Western Reserve University (CWRU) benchmark dataset and Mechanical Fault Prevention Technology (MFPT) benchmark dataset. In case study 1, using the UCI dataset for testing, GBSTA-M reduced computation time by up to 94% compared with traditional algorithms. In case study 3, GBSTA-M combined with SVM achieved an accuracy of 98.7% on the MFPT dataset. Experimental results demonstrate that, compared to conventional methods, the proposed model not only achieves higher fault diagnosis accuracy but also significantly reduces computational resource requirements in specific scenarios while exhibiting excellent robustness.

This paper presents an enhanced robust filtering algorithm designed for integrated SINS/GNSS navigation systems operating under nonGaussian noise conditions. To address the challenges posed by heavy-tailed noise distributions, a novel noise modelling framework based on Student's t-distribution is developed, which provides superior outlier resilience compared to conventional Gaussian assumptions. Furthermore, a Gaussian mixture model representation is employed for both the one-step predicted and likelihood probability density functions, enabling more accurate quantification of uncertainty. Additionally, a variational Bayesian-based adaptive mechanism is employed for dynamic scale matrix optimisation, effectively mitigating the impact of process noise outliers. Extensive experimental validation, including Monte Carlo simulations and vehicular tests, demonstrates the algorithm's superior performance in SINS/GNSS integration scenarios. Comparative results indicate significant improvements in positioning accuracy and robust convergence characteristics relative to a decent number of iterations.

The transformer is an important equipment in power systems. However, prolonged abnormal conditions can lead to significant damage of the transformer equipment. The current finite element analysis (FEA) method for calculating the internal physical field of transformers is time-consuming, limiting its practicality for fast simulation. This paper focuses on predicting the internal magnetic fields of transformers under overvoltage conditions, which cause irregular changes in the transformer magnetic fields due to overvoltage. Simulation datasets of transformer magnetic field under overvoltage conditions were acquired via the COMSOL software. Subsequent analysis elucidated the influence of overvoltage parameters on the electrical characteristics of transformers. Furthermore, the dimensionality of input features relevant to magnetic field prediction was expanded. Convolutional neural network (CNN) model was then employed to forecast the internal magnetic fields of transformers under overvoltage conditions. Experimental results were compared with Random Forest (RF), eXtreme Gradient Boosting (XGBoost) and deep neural network (DNN) models, demonstrating the efficiency of deep learning methods in predicting transformer magnetic fields under overvoltage conditions.

To design the permanent magnetic eddy current couplers (PMECCs), modelling the magnetic field is essential. Traditional equivalent magnetic circuit methods and analytical methods often rely heavily on expert experience, whereas finite element methods (FEM) demand significant computational resources and time. Recently, the physics-informed neural network (PINN) has emerged as a novel approach for modelling electromagnetic fields. To fully harness the potential of PINN, eliminate reliance on data sets, and enhance the generalisation ability of multi-scale physical systems, we simplify the physical model of PMECCs and analyse its inherent boundary conditions based on the fundamental properties of electromagnetic fields. A dimensionless and unsupervised PINN, employing dimensional analysis to reduce the dimensions of the physical variables in the system was proposed. The dimensionless PINN (DPINN) is trained through unsupervised learning to solve the magnetic field equations and predict PMECC performance. Furthermore, dimensional analysis and transfer learning method are applied to enable the network to address a broader class of problems, resulting in a 92% reduction in training cost. The solution results, compared with those from the finite element method and analytical solution, exhibit similar error magnitudes (10⁻⁴ Wb/m), confirming the method's high accuracy.