Accurate wind power prediction can increase the utilization rate of wind power generation and maintain the stability of the power system. At present, a large number of wind power prediction studies are based on the mean square error (MSE) loss function, which generates many errors when predicting original data with random fluctuation and non-stationarity. Therefore, a hybrid model for wind power prediction named IVMD-FE-Ad-Informer, which is based on Informer with an adaptive loss function and combines improved variational mode decomposition (IVMD) and fuzzy entropy (FE), is proposed. Firstly, the original data are decomposed into K subsequences by IVMD, which possess distinct frequency domain characteristics. Secondly, the sub-series are reconstructed into new elements using FE. Then, the adaptive and robust Ad-Informer model predicts new elements and the predicted values of each element are superimposed to obtain the final results of wind power. Finally, the model is analyzed and evaluated on two real datasets collected from wind farms in China and Spain. The results demonstrate that the proposed model is superior to other models in the performance and accuracy on different datasets, and this model can effectively meet the demand for actual wind power prediction.

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.

Physics-informed neural networks (PINNs) have significant potential for modeling and parameter design in engineering field. While most existing PINNs research focuses on fluid mechanics and thermodynamics, few studies explore its application in electromagnetic field modeling of electromagnetic devices. Modeling the permanent magnet eddy current coupler (PMECC) to predict its performance characteristics based on geometric parameters and material properties is crucial for its design and optimization. An unsupervised modeling method for PMECC based on physics-informed radial basis neural networks (PIRBFNNs) is presented in this work. The modeling and solving of static magnetic field for devices excited by permanent magnets (PMs) is realized, which solves the problem of the traditional PINN fully-connected structure with many parameters and difficult training. We use the magnetic vector potential as the solution objective without providing the magnetic field boundary parameters and without labeling data, which is an unsupervised learning paradigm. The magnetic field distribution and performance of the PMECC can be computed using only the structural parameters. The experimental results show that the proposed PIRBFNN method is basically consistent with the results of the finite element numerical method and the analytical method. Additionally, a transfer learning experimental study was conducted to validate the effectiveness of the network components and training methods proposed in this paper. Furthermore, the proposed method can be applied to the modeling and analysis of various devices utilizing PM excitations.