Lithium-ion batteries (LIBs) serve as core energy components in various applications, including new-energy vehicles, unmanned aerial vehicles, and energy-storage systems. In application, precise state-of-health (SOH) estimation is crucial for ensuring safe, reliable operation and improving system efficiency. However, the limitations of traditional SOH estimation methods in terms of interpretability and data generalization remain to be addressed. To tackle these challenges, a solution is proposed using a self-adaptive physics-informed neural network (SA-PINN) approach. The method extracts highly-correlated features from electronic current and temperature profiles during the constant-voltage charging phase while simultaneously incorporating solid electrolyte interphase (SEI) film growth kinetics and the Arrhenius temperature effect as dual physical constraints to guide network learning. A collaborative mechanism combining adaptive weighting and transfer learning enables SOH monitoring across different battery types and operating conditions, enhancing physical consistency without compromising data-fitting accuracy. Experimental results indicate that the proposed method achieves a mean absolute percentage error (MAPE) below 0.81
Addressing the severe torque ripple and output torque loss in five-phase interior permanent magnet synchronous motors (FP-IPMSMs) during single-phase open-circuit faults, a novel virtual signal injection (VSI)-based maximum torque per ampere (MTPA) control strategy integrated with fault-tolerant current optimization is proposed in this paper. First, the post-fault reduced order decoupling transformation matrix is constructed, and the single-phase open-circuit mathematical model of FP-IPMSM in synchronous rotating coordinate system is obtained. Second, based on the principle of constant magnetomotive force before and after the fault, the current constraint condition for minimizing torque ripple is solved to construct fault-tolerant vector control. On this basis, the fault-tolerant current that maximizes the permanent magnet torque component is solved. Then, the VSI-MTPA approach dynamically exploits the reluctance torque component in salient-pole motors, enabling higher torque density during fault-tolerant operation. Finally, simulation and experimental validation demonstrate that the proposed strategy reduces torque ripple while increasing the output torque under single-phase fault conditions.
The randomness and intermittency of wind speed cause severe fluctuations in the output power of wind farms (WFs). To minimize battery lifetime cost and smooth power output, a coordinated power control (CPC) strategy is proposed for the energy storage systems (ESSs) of WFs based on the model predictive control (MPC). First, to linearize the aerodynamics, drive trains, and energy storage converters of wind turbines, the small-signal analysis method is employed to derive the linear incremental equations of WF-ESSs. Then, to quantify the nonlinear life degradation process of ESSs, a degradation cost function incorporating the coupling relationship between depth of discharge (DOD) and temperature is established. A MPC-based model of wind turbines (WTs) and ESSs is developed to optimize the active power output for minimizing battery lifetime cost and smoothing power fluctuations. Finally, a WFs-ESSs simulations is established to validate the effectiveness of the proposed method.
Lithium-ion batteries are widely used because of their high energy density. However, mechanical, thermal, and electrical abuse can cause internal short-circuit (ISC) faults, which may trigger thermal runaway and threaten the operational safety of battery energy storage systems. To address this issue, an adaptive multi-scale MRC-BiLSTM-based ISC fault diagnosis method is proposed. First, a RIME optimization algorithm is integrated into a multi-scale convolutional neural network to adaptively optimize the convolution kernel size, number of kernels, and learning rate, constructing a multi-scale RIME CNN (MRC). The MRC is used to extract multi-scale features from current, voltage, and temperature signals before and after ISC occurrence, while a channel attention mechanism is introduced to enhance informative feature selection. Subsequently, a self-attention mechanism is incorporated to strengthen the ability of the model to capture critical temporal information, and the extracted features are further input into a BiLSTM network to learn the temporal dependencies among characteristic variables. Finally, ISC simulation data are used to validate the proposed method. Results demonstrate that the proposed method can effectively improve the accuracy and reliability of ISC fault diagnosis.
A variable boundary layer adaptive sliding mode observer (ASMO) for low carrier ratios is proposed in this paper. By constructing an exact discrete sliding mode observer and redefining the back-EMF estimation based on quasi-sliding mode dynamics, the proposed method eliminates the low-pass filter and significantly reduces the discretization errors, non-ideal sliding mode effects, and filter delays that are common in conventional sliding mode observers under low carrier ratios. A saturation function is used instead of a sign function, and a method for selecting the variable boundary layers and adaptive sliding mode gains based on stable quasi-sliding mode operation is introduced. This approach considers both the sampling frequency and rotor speed to enable real-time parameter adjustment. Furthermore, a discretized phase-locked loop (PLL) with angle compensation is proposed to extract the motor speed and position, effectively reducing angle estimation errors under low carrier ratios. The ASMO presented in this paper features a simple computational structure and demonstrates superior estimation accuracy and disturbance rejection capability compared to the conventional sliding mode observer under low carrier ratios, as validated by experimental results.
