The operation of the electricity network has grown more complex due to the increased adoption of renewable energy resources, such as wind and solar power. Using energy storage technology can improve the stability and quality of the power grid. One such technology is flywheel energy storage systems (FESSs). Compared with other energy storage systems, FESSs offer numerous advantages, including a long lifespan, exceptional efficiency, high power density, and minimal environmental impact. This article comprehensively reviews the key components of FESSs, including flywheel rotors, motor types, bearing support technologies, and power electronic converter technologies. It also presents the diverse applications of FESSs in different scenarios. The progress of state-of-the-art research is discussed, emphasizing the use of artificial intelligence methods such as machine learning, digital twins, and data-driven techniques for system simulation, fault prediction, and life-assessment research. The article also addresses the challenges related to current research and the application of FESSs. It concludes by summarizing future directions and trends in FESS research, offering valuable information for further advancement and improvement in this field.

The popularity of permanent magnet synchronous motors (PMSMs) has increased in recent years due to their high efficiency, compact size, and low maintenance needs. Calculating iron loss in PMSMs is crucial for designing and optimizing PMSMs to achieve high efficiency and a long lifespan, as this can significantly affect motor performance. However, multiple factors influence the accuracy of iron loss calculations in PMSMs, including the intricate magnetic behavior of the motor under different operating conditions, as well as the influence of the motor’s dynamic behavior during the operation process. This paper proposes a method based on particle swarm optimization (PSO) and a recurrent neural network (RNN) to estimate the iron loss in PMSMs, independent of the empirical iron loss formula. This method establishes an iron loss calculation model considering high-order harmonics, rotating magnetization, and temperature factors. Accounting for the multifactor influence, the model studies the law of loss change under different magnetic flux densities, frequencies, and temperature conditions. To avoid the deviation problem caused by conventional polynomial fitting, a multilayer RNN and PSO are used to train and optimize the neural network. Iron loss in complex cases beyond the measurement range can be accurately estimated. The proposed method helps achieve a PMSM iron loss calculation model with broad applicability and high accuracy.

With the improvement of the digitalization of electrical machines and their drive systems, it is an inevitable trend to realize accurate and fast physical simulation and behavior modeling methods. The realization of digitalization is also helpful to the calculation of performance and coupling analysis in other systems. It is of great significance to monitor and control system operating conditions. This paper presents an electromagnetic performance analysis method for fast dynamic simulation of a bearingless permanent magnet motor (BPMSM) based on model order reduction (MOR). The electromagnetic characteristics and dynamic behavior model of the rotating machine are constructed by using the MOR based on proper orthogonal decomposition. First, the q-axis and d-axis currents and rotor angle of the motor are scanned to construct the magnetic field distribution snapshot matrix. Then singular value decomposition of the snapshot matrix is used to calculate the basis vectors. Finally, the order of the system is reduced by selecting an appropriate truncation order of the basis vectors. Compared with the finite element method (FEM), the calculation time of MOR for the BPMSM is significantly reduced. At the same time, it can provide accurate results of motor magnetic field distribution, electromagnetic force, levitation force and dynamic response calculations, and achieve fast and accurate state feedback. Through analysis and comparison, the results of the proposed method match well with the FEM simulations.

The data-driven machine-learning approach has significantly advanced the development of computational electromagnetics. This study introduces the Kolmogorov-Arnold network (KAN) as a novel method to overcome the limitations of traditional multilayer perceptron-based physics-informed neural networks (MLP-PINNs), which often struggle with fixed activation functions and high computational costs. In terms of accuracy, KAN outperforms traditional PINNs with a 13.47% improvement in estimated accuracy. For efficient convergence, KAN achieves stable training with only 175 steps, significantly reducing computational overhead compared to PINNs, which require over 15000 steps. Additionally, KAN demonstrates superior generalizability and flexibility, achieving an average 5.03% accuracy improvement over PINN in transfer learning scenarios. These results highlight KANs potential in computational electromagnetics.

Sensorless control of high-speed permanent magnet synchronous motors (HSPMSMs) using the conventional extended state observer (ESO) exhibits limitations in accurately estimating the back electromotive force (BEMF). Such phase delays adversely affect the performance and dynamic response of HSPMSMs. To address this issue, a cascade high-order extended state observer (CHESO) is proposed to enhance position estimation and speed control of HSPMSMs. The effectiveness of the CHESO in extending the observation bandwidth and improving control accuracy is demonstrated. In the proposed two-level CHESO architecture, the BEMF is accurately estimated at the first level, while position, load torque, and uncertain disturbances are concurrently estimated at the second level. Subsequently, the estimated disturbances and load torque are employed in an adaptive feedforward control strategy to enhance the dynamic performance and anti-disturbance capability of the HSPMSM control system. Compared to the conventional ESO, the proposed CHESO minimizes position estimation errors and improves the estimation accuracy of high-speed regions, as verified by experimental results.

In the optimization of high-speed permanent magnet motors (HSPMMs), electromagnetic characteristics, rotor stress, rotor dynamics, and temperature characteristics must all be considered simultaneously, and there are numerous optimization parameters for both the stator and rotor. These factors pose significant challenges to the multiphysics optimization of HSPMMs. Therefore, this paper presents a multiphysics optimization process for the HSPMM of 60 kW 30,000 rpm by combining subspace strategy and sequential strategy to mitigate the issues of high training volume and mutual coupling. Ten optimization parameters of stator and rotor are determined firstly. Then, using finite element analysis of rotor stress and rotor dynamics, the range of values for critical parameters of the rotor is established. Next, in the electromagnetic optimization, the process is divided into rotor parameter subspace and stator parameter subspace according to the subspace optimization strategy. The temperature field is also checked based on the optimization results. Finally, a prototype is manufactured and the comprehensive performance is tested to validate the multiphysics optimization process.