Edge‐preserving denoising is important in image analysis. However, most existing methods suffer from smoothing and loss of high frequency detail at the edges. Aiming at this issue, a multi‐path residual network using edge prior information is proposed. Specifically, the edge is first recovered by a newly designed parallel edge extraction module. Then it is used as a priori to generate affine transform parameters to guide the weight distribution between noise and edge, so that the shallow feature extraction pays more attention to preserving edges. In the deep feature extraction stage, a multi‐scale attention unit is designed to provide the spatial neighbourhood information of the feature points to filter and activate the deep features, which further enhances the ability of the network to extract fine‐grained noisy features. Finally, by taking the difference between the noisy image and the noise feature obtained with residual learning, the denoised image is produced. Extensive experiments show that the PSNR is improved by 0.00–0.78 dB on grey image denoising and 0.00–2.69 dB on colour image denoising compared with others. The denoised image has a better ability to preserve image edges, high frequency details and visual perception.

With the rapid development of renewable energy, fluctuations in photovoltaic (PV) output power pose challenges to the power system. To address the issue of increased burden on hybrid energy storage system (HESS) due to the time delay associated with PV power smoothing methods and to optimize the power allocation of the HESS, this paper proposes a power allocation strategy of the HESS to smooth PV power fluctuations that combines a Hamming window-weighted optimized Savitzky-Golay (SG) filtering method with Variational Mode Decomposition (VMD). Additionally, a multi-objective fuzzy control approach is designed to regulate the compensating power of the HESS. Firstly, the Hamming window-weighted optimized SG filtering is applied to preprocess the PV output power, resulting in smoothed power and fluctuating power. Subsequently, VMD is employed to decompose the fluctuating power, separating high-frequency and low-frequency components, which are then allocated to supercapacitors and batteries respectively for storage and compensation. On this basis, this paper innovatively designs a multi-objective fuzzy controller that considers both the State of Charge (SOC) of the supercapacitors and batteries, as well as the mitigation effect of PV power, to achieve the regulation of compensating power for the HESS. Simulation analysis results show that the proposed method not only effectively smooths fluctuations in PV output power but also reduces the SG filter window by 24.39%, thereby reducing the burden on the HESS. Furthermore, while ensuring the safe operation of energy storage devices, it optimizes the SOC distribution of the energy storage devices and extends their service life.

To address the challenges of multi-source uncertainties and dynamic optimization faced by hybrid energy storage systems (HESS) in urban rail transit under complex operating conditions, this paper proposes a Robust Model Predictive Control (RMPC) method, which minimizes the system's performance upper bound under adverse disturbance scenarios to guarantee robustness. First, a state-space model of the controlled system is directly constructed using input-output data through the Multivariable Output Error State Space (MOESP) algorithm, reducing the complexity of direct system modeling. Subsequently, an infinite time-domain min-max optimization strategy is employed to establish robust performance guarantees, systematically minimizing the performance upper bound under adverse disturbance scenarios while effectively addressing the limitations of traditional MPC approaches that depend on deterministic assumptions. Furthermore, by integrating Linear Matrix Inequality (LMI) and invariant set theory, input/output constraints are explicitly handled to reduce optimization dimensionality and improve computational efficiency. The simulation results demonstrate that the proposed RMPC method delivers outstanding performance in multi-disturbance tests, achieving an average voltage recovery time below 0.3 seconds, an overshoot under 2.3%, and a steady-state error within 5V. In the multi-train operation scenario, the method maintains DC bus voltage fluctuations within the 1573–1750V range while reducing the SOC variance of the battery and supercapacitor to 0.94%² and 1.76%², respectively. Real-time HIL simulation further validates the computational efficiency of the proposed method, exhibiting a consistent execution time of 13 μs with zero CPU delay.