To address the scarcity of fault samples in photovoltaic (PV) arrays and the tendency of conventional data-driven models to overfit under small-sample conditions, this paper proposes a triplet-Siamese fault-diagnosis framework, termed triplet-Siamese-multi-scale temporal convolutional network (Tri-MSTCN). First, an operating model of the PV array is established to collect time-series samples. A triplet sampling strategy is then adopted to construct anchor-positive-negative tuples, which enlarges the effective training set and increases sample diversity. The proposed network employs a multi-scale temporal convolutional network (TCN) to extract features at multiple temporal resolutions, integrates a channel-attention module to emphasize informative channels, and introduces a gated residual fusion mechanism to adaptively combine the reweighted features with the original input, thereby enhancing feature representation. The experimental results show that from 45 samples to 135 samples, the accuracy of the Tri-MSTCN model significantly improved, increasing from 96.67% to 98.89%. In contrast, it is clearly higher than other models, such as GRU, CBG, and TCN. As the sample size increases, the accuracy of Tri-MSTCN increases significantly, further validating the model's effectiveness and generalization ability in PV array fault diagnosis.

In off-grid wind-to-hydrogen systems, economic scheduling of homogeneous alkaline electrolyzer arrays can result in dispersed low-load operation and abrupt power commands, increasing the burden on lower-level controllers. This paper develops a rolling mixed-integer linear programming (MILP) framework to coordinate electrolyzer operation, battery energy storage, and off-grid power balance. Hard ordered-loading constraints remove unordered power allocation, while a lexicographic objective sequentially maximizes economic performance, minimizes the number of production-state units, and smooths power commands. Case-study results show that ordered loading and production-state-unit minimization reduce the average number of production-state electrolyzers from 2.5556 to 2.1944 (14.13%), with a 0.46% decrease in the composite objective value relative to the economic-only baseline. Power-command shaping further reduces total absolute power variation from 462.00 to 73.05 MW (84.19%) without changing the first two metrics to the reported precision. The proposed framework therefore concentrates electrolyzer loading and substantially smooths upper-level power commands at a limited economic cost.

In off-grid wind-to-hydrogen systems, wind fluctuations may cause frequent electrolyzer start-stop cycles and equipment degradation. This paper proposes a prediction-based multi-objective rolling optimization strategy for alkaline electrolyzer arrays. CNN-LSTM forecasts short-term wind power, and the predicted average and fluctuation level are combined with a multi-state model to estimate operating units. A weighted fitness function coordinates start-stop suppression, operating-time balancing, standby buffering, and power smoothing. Case studies show that the proposed method reduces start-stop events by more than 60% and increases hydrogen production by about 2.6%–3.0%, improving stability, efficiency, and equipment lifetime.

Existing denoising methods for gas-insulated switchgear (GIS) partial discharge (PD) signals often rely on fixed decomposition parameters or single-index intrinsic mode function (IMF) selection, limiting their adaptability to low-SNR mixed noise and weakening PD pulse morphology. To address this issue, this paper proposes a denoising method combining sparrow search algorithm (SSA)-optimized variational mode decomposition (VMD) with adaptive IMF selection. A four-pulse PD simulation signal is tested under white noise and mixed noise containing narrow-band interference and envelope-modulated random disturbance. SSA jointly optimizes the VMD mode number and penalty factor to reduce under- and over-decomposition, while a multi-feature IMF score based on correlation, kurtosis, energy contribution, and front-to-tail energy ratio is used to select PD-related modes and suppress noise-dominated components. The method is compared with wavelet thresholding, Wiener filtering, and fixed-parameter VMD using normalized correlation coefficient, output signal-to-noise ratio, and mean squared error. Results show better denoising performance under low-SNR mixed noise while preserving the main time-domain characteristics of PD pulses.

The intermittent nature of solar irradiance introduces substantial variability into photovoltaic water electrolysis hydrogen production systems (PWEHPSs), posing significant challenges to stable power scheduling and hydrogen production. Accurate forecasting of global horizontal irradiance (GHI) is therefore essential for improving operational reliability; however, existing studies still offer insufficient predictive accuracy and inadequate system-level validation for reliable electrolyzer-coupled operation. To address this gap, this work develops an ESN-based GHI forecasting model and employs a multistrategy improved projection iteration optimization algorithm (MSPIMO) to tune key hyperparameters, including reservoir size, leakage rate, spectral radius, and input scaling coefficient. An integrated evaluation framework is further established to link irradiance forecasting with photovoltaic power output prediction and hydrogen production assessment in a proton-exchange-membrane-electrolyzer-based PWEHPS. Using multisource meteorological data from the National Solar Radiation Database for Lanzhou, Gansu Province, China, the proposed model was trained, validated, and evaluated over 30 independent runs. The model achieved mean MAE, RMSE, MAPE, and R2 values of 15.075 W/m2, 20.620 W/m2, 13.82%, and 0.993, respectively. Compared with the WOA-EESN, the best-performing benchmark model, the proposed model reduced MAE and RMSE by 40.6% and 39.6%, respectively. In the downstream PWEHPS application, the relative error for monthly cumulative hydrogen production was 0.50%, while the daily and weekly relative errors remained within 0.66% and 0.28% to 0.72%, respectively. These results demonstrate that the proposed approach improves the GHI forecasting accuracy and enables more reliable downstream hydrogen production estimation under the studied scenario.

Short-term load forecasting (STLF) is crucial for energy management. Accurate and robust predictions remain challenging because load series exhibit non-stationarity, multi-scale dependencies, and stochastic variations. We propose ICEEMDAN-DAGT, a hybrid multi-component framework designed to tackle these complexities. The framework employs an improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) to decompose complex load signals. It incorporates an attention-enhanced dual-stream prediction module (DAGT), which uses a Gated Temporal Convolutional Network (TCN) to capture long-term dependencies and a Temporal-Feature Attention (TFA)-enhanced Gated Recurrent Unit (GRU) to model short-to-medium-term patterns. Dendritic Neuron Models (DNM) are used to strengthen nonlinear mapping, and an error-driven dynamic weighting strategy adaptively fuses subnetwork outputs. When validated on two real-world datasets (Panama City and Australia), ICEEMDAN-DAGT demonstrates superior performance that is statistically significant. On the Australian dataset, it achieves MAE reductions of 11.94 %-19.10 % and RMSE reductions of 12.48 %- 19.22 % compared to advanced baselines such as PatchTST and DLinear. These results confirm the framework's robust generalizability, improved accuracy, and computational efficiency, providing a promising solution for challenging STLF tasks.

Wind power signals exhibit strong non-stationarity, high-frequency noise, and temporal coupling characteristics, which make it difficult for traditional models to effectively capture feature representations, temporal dependencies, and prediction reliability. To address the challenges of non-stationarity, multi-scale temporal dynamics, and uncertain interval estimation, this study proposes a unified deterministic-probabilistic deep ensemble forecasting framework for wind power prediction. First, the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is employed to decompose the original power series, followed by K-means clustering to reconstruct intrinsic mode functions (IMFs). The high-frequency components are further processed using feature model decomposition (FMD) to reduce mode mixing and enhance feature separability. Then, a hybrid deep forecasting model is constructed by integrating a multi-scale temporal convolutional network (MSTCN), bidirectional gated recurrent unit (BiGRU), and non-stationary self-attention (NSA) mechanism to jointly model multi-scale dynamic and contextual dependencies, thereby improving prediction accuracy and stability. In addition, a multi-strategy improvement black-winged kite algorithm (MSBKA) is designed for multi-strategy optimization of model parameters, enhancing hyperparameter search efficiency and global convergence capability. Finally, an improved non-crossing quantile regression (INCQR) method is introduced by combining H-Pinball loss and Softplus mapping to ensure structural consistency and monotonicity of prediction intervals, thus improving the reliability of uncertainty quantification. Experimental results demonstrate that the proposed model achieves the highest fitting accuracy across four seasonal datasets (R2spr.= 0.9533, R2 sum. = 0.9628, R2aum. = 0.9802, and R2win. = 0.9810) and exhibits superior consistency and coverage of prediction intervals (PICP = 92.65 %, PINAW = 0.1044).

To address the impact of arsenic accumulation on production stability during copper electrolyte purification and achieve precise optimization and dynamic regulation of process parameters, this paper develops a dynamic modelling and parameter estimation framework integrating mechanism and data. Based on the continuous stirred tank reactor (CSTR), a coupled dynamic model incorporating mass, energy, and electrochemical principles is established for the arsenic electrowinning removal process. Six key kinetic parameters in the model are efficiently inverted using the sequential quadratic programming (SQP) algorithm. Dynamic response tests based on the optimized model further reveal the influence mechanism of operating variables on arsenic removal efficiency. The model is comprehensively validated using industrial data from three typical periods, demonstrating favourable operational adaptability and predictive robustness. Parameter sensitivity analysis clarifies the influence degree of each kinetic parameter and identifies the relative contributions of key parameters. The model and optimization method developed in this study provide reliable theoretical and tool support for real-time optimization and precise control of copper electrolyte purification.

This paper addresses the speed regulation performance degradation in surface-mounted permanent magnet synchronous motor (SPMSM) drive systems caused by model uncertainties, parameter perturbations, and external load disturbances. To enhance both dynamic response and robustness, this paper propose a composite control strategy integrating a radial basis function (RBF) neural network, a fixed-time integral terminal sliding mode controller (FITSMC), and a super-twisting extended state observer (STESO). The FITSMC guarantees fixed-time error convergence with a bounded settling time independent of initial states, while the integral action eliminates steady-state errors. The RBF network online approximates internal parameter variations and unmodeled dynamics, whereas the STESO estimates external load disturbances for feedforward compensation. This coordinated disturbance rejection mechanism reduces the required switching gain, effectively suppressing sliding mode chattering without sacrificing robustness. Rigorous Lyapunov analysis establishes the fixed-time stability of the closed-loop system. Comparative simulations under no-load startup, sudden load change, and full-load operation demonstrate that the proposed RBF-FITSMC-STESO achieves faster convergence, smaller speed drops, and lower steady-state fluctuations than conventional PI control, standard sliding mode control, and FITSMC-STESO without RBF adaptation, confirming its effectiveness for high-performance SPMSM speed regulation.

High-precision trajectory tracking control of space flexible manipulator represents a significant research focus of contemporary research and poses great challenges in both academia and engineering. To address the issue of low control precision in space flexible manipulator, which arises from highly nonlinear dynamics in complex spacecraft environments, the LuGre friction model is incorporated into the dynamic equation to improve the accuracy of frictional dynamic behaviour modelling. Subsequently, a reinforcement learning-based sliding mode control (RL-SMC) method is developed to achieve precise approximation and compensation of uncertain nonlinearities within the space flexible manipulator system. The employed RL framework is based on the actor-critic architecture, where the actor neural network generates the control policy, whereas the critic neural network evaluates the policy and continuously provides feedback regarding the system state. This control method uses a radial basis function neural network (RBFNN) combined with the SMC to minimise approximation error. In complex space environments, the actor-critic framework enhances the approximation of nonlinear dynamics for a space flexible manipulator and facilitates more efficient adaptation to variations in system dynamics. In addition, joint angle output constraints are implemented to manage the restricted motion of the space flexible manipulator in confined workspaces, aiming to prevent collisions during operation and avoid structural damage. Finally, the stability of the closed-loop system is rigorously established using the Lyapunov stability theory. Numerical simulations demonstrate the efficacy of the proposed approach in improving both control precision and environmental adaptability of the space manipulator.

Deep learning has shown substantial progress in fault diagnosis in recent years. Nevertheless, its application is hindered by excessive computational requirements and challenges in addressing data scarcity and multicondition adaptability in practical settings. Notably, the problem of feature redundancy stands out, as it not only inflates computational demands but also obscures vital features, reducing the generalizability and precision of diagnostic models. To address these challenges, this study proposes an efficient transfer fault diagnosis algorithm based on a multi-scale redundant feature selection network enhanced by parameter optimization. First, the proposed algorithm combines short-time Fourier transform (STFT) and synchrosqueezing wavelet transform (SWT) to comprehensively extract deep-level features from the data. Secondly, a channel and spatial redundancy feature selection unit is employed to effectively reduce channel and spatial redundancy in the features, thereby focusing on the learning of highly representative features. Furthermore, a transformer feature extraction (TFE) module is integrated at the backend of the model to accurately capture global contextual information and detailed features in the data. Finally, a multi-strategy improved Harris Hawk optimization algorithm (MS-HHO) was designed to achieve efficient hyperparameter tuning and global optimal configuration. Experimental results demonstrate that the proposed model achieved peak diagnostic accuracies of 98.63%, 98.83%, and 98.9% on three validation datasets using only 20% of the samples. Even under severe noise conditions, the model maintained an average diagnostic accuracy of 73.75%. Additionally, it exhibited superior diagnostic performance and robustness compared to competing models in diverse transfer tasks.

In the realm of Post-combustion CO2 Capture (PCC) regulation, Model Predictive Control (MPC) has emerged as a dominant standard. However, conventional MPC faces challenges when addressing the intrinsic multi-variable coupling within PCC units. To overcome this, the present study initially establishes a multivariable MPC framework tailored to the system's three-input/two-output configuration. Moreover, to tackle the significant nonlinearity arising when PCC systems operate across broad load ranges, a novel control scheme integrating gain scheduling techniques with MPC is introduced. This methodology entails constructing transfer function models at representative operating points and synthesizing a global system model via a gain scheduling strategy. By leveraging predictive control algorithms, global optimization of control performance is realized. Simulation results reveal that, in contrast to traditional MPC approaches, the proposed gain scheduling-based predictor exhibits enhanced agility in tracking load fluctuations over wide operating conditions. The coordinated control system achieves faster setpoint tracking with minimized dynamic deviations, ensuring swifter, smoother actuator movements and superior stability.

Currently, most mainstream gearbox fault diagnosis methods rely heavily on deep learning techniques. However, such deep learning-based diagnostic models often face challenges such as high computational resource requirements, poor real-time performance, and low diagnostic accuracy when applied to other operating conditions. To address the aforementioned issues, a novel lightweight cross-operating-condition gearbox transfer fault diagnosis framework is proposed, incorporating parameter optimization and adaptive feature attention enhancement. The framework utilizes two-dimensional Gramian angular field images to extract deep gearbox fault information. An improved lightweight mechanism integrating multi-scale channel and spatial adaptive attention enhancement is first adopted to allocate feature weights. This mechanism amplifies critical features while effectively mitigating the influence of redundant or noisy features. Subsequently, a local maximum mean discrepancy loss is employed to achieve cross-domain feature alignment across different operating conditions. Finally, a multi-strategy improved triangular topology optimization algorithm is applied to fine-tune the model's hyperparameters, boosting its cross-condition diagnostic capabilities and ensuring more accurate fault detection across different operational scenarios. Experimental results indicate that the proposed model achieved accuracy rates of 99.85% and 99.94% on two validation datasets, while maintaining diagnostic accuracies of 74.6% and 79.6% in the presence of significant noise. The model's diagnostic performance and robustness are superior to those of other models, demonstrating its effectiveness in challenging conditions.

For the incipient faults of vehicle power supply under imbalanced datasets, the traditional shallow network has the problems of limited feature extraction ability and the insufficient generalization ability of a single network model. In this paper, an AdaBoost-SAE deep ensemble diagnosis method, which combines the Stacked Auto-Encoder (SAE) deep network and Adaptive Boosting (AdaBoost) algorithm, is proposed. First, SAE is used as a weak classifier to learn and extract incipient fault features from the monitoring date of vehicle power supply. Secondly, in the iterative training process of the model, the classification performance of a single SAE is improved step-by-step by constantly adjusting the weights of the misclassified samples in the training set. Finally, the multiple weak classifiers are combined into strong classifiers by linear weighting to achieve accurate identification of incipient faults under imbalanced datasets. The test results demonstrate that the proposed method can mine deeper features of incipient faults and effectively improve the adverse effects of sample imbalance. Compared with traditional fault diagnosis models and a single SAE, the accuracy of the incipient fault diagnosis can reach 96.6%. Furthermore, the F1-scores of the various working conditions also increased significantly.

In this article, a FCS-MPCC control algorithm with delay compensation for photovoltaic grid-connected inverters is proposed. This algorithm can reduce the delay problem caused by the calculation of switching time during the operation of the system through a two-step delay compensation strategy. Improve the delay problem in the algorithm’s running process as much as possible. In this paper, a new delay compensation algorithm is used, that is, the current i(k) measured at k time and the switching state applied to estimate the load current at k+1, and estimate the load current at k+1 to predict the load current at k+2. The corresponding switching state is applied in k+1 and k+2 time.Finally, the proposed algorithm is verified by MATLAB/Simulink simulation software, it is found that the performance of the FCS-MPCC algorithm with delay compensation is better than that of the traditional FCS-MPCC and the algorithm based on feedforward decoupling PI control, The THD is smaller and the power loss is lower.

To address the challenges of communication resource constraints and insufficient dynamic task adaptability in the cooperative control of second-order nonlinear multi-agent systems, this paper proposes an adaptive hybrid event-triggered fixed-time proportional consensus control method. First, a dynamic proportional factor generation mechanism is designed to adaptively adjust state proportion allocation parameters in real time, accommodating time-varying demands in scenarios such as formation reconfiguration and power distribution. Second, a hybrid event-triggered strategy is constructed by integrating time-triggered and dynamically thresholded event-triggered mechanisms, ensuring fixed-time convergence while reducing communication frequency. Theoretical analysis guarantees the fixed-time stability of the closed-loop system and establishes the existence of a lower bound on event-triggering intervals. Simulation results demonstrate that, compared to conventional static-threshold event-triggered methods, the proposed approach achieves reduced communication overhead under strong disturbances, maintains convergence of dynamic proportional tracking error, and ensures that the convergence time aligns with the predefined upper bound. Finally, simulation case studies validate the effectiveness of the proposed algorithm.

In order to solve the problem of low thermodynamic efficiency of traditional steam-enhanced reforming reaction systems, this paper uses Aspen Plus (R) software for chemical process simulation to build a complete process for hydrogen production from pyrolysis gas. It conducts detailed modeling of gas purification, enhanced reforming, calcination and regeneration, dehydration and drying, pressure swing adsorption(PSA) and other units. The model predictive control(MPC) is designed according to the optimal parameter ratio. The simulation results show that compared with the traditional proportion integration differentiation(PID) control, under the same feed condition, the response time of the MPC controller is reduced by 390s and 186s, respectively, under the step response and slope response tracking curves. The MPC controller's IAE(Integral Absolute Error) value is minor, and the control effect is better than the PID controller. The enhanced reforming hydrogen production system of pyrolysis gas combined with MPC control can realize the stable operation of the system and higher energy efficiency. It can be used as one of the ways to produce hydrogen from gas at present and has specific practical application value.

Aiming at the problems of long pre-training time-consuming and poor diagnostic accuracy during unsupervised training of traditional DBN, an adaptive deep belief network (ADBN ) is proposed for analog circuit fault diagnosis. The adaptive learning rate is proposed according to the similarity and difference of the parameter updating direction to improve the convergence speed of the network. The ADBN is applied to the diagnosis experiments of a two-stage four-op-amp dual second-order low-pass filter, and the experimental results show that the proposed ADBN can realize the classification and localization of difficult faults by guaranteeing the classification accuracy and higher diagnosis rate based on the fast convergence speed.

Aiming at the problem of difficult operation and maintenance of PV power plants in complex backgrounds and combined with image processing technology, a method for detecting hot spot defects in infrared image PV panels that combines segmentation and detection, Deeplab-YOLO, is proposed. In the PV panel segmentation stage, MobileNetV2 was introduced into the Deeplabv3+ model. Empty convolution in the atrous spatial pyramid pooling (ASPP) structure was improved, and established a relationship between layer-level features, the CBAM attention mechanism was combined, which achieved fast and accurate segment of PV panels and avoided false detection of hot spots. In the hot-spot recognition stage, a lightweight MobileNetV3 network was designed to replace the YOLO v5 backbone network, a small defect prediction head was added, and EIOU was used as a loss function, which improved the speed and accuracy of hot-spot detection and enhanced the performance of the YOLO v5 model. The experimental results show that the optimized Deeplabv3+ model and YOLO v5 model improve the accuracy of segmenting PV panels in images and identifying hot-spot defects by 2.61% and 0.7%, respectively, compared with the original model. This proposed method can accurately segment the PV panels and then identify different sizes of hot-spot defects on the PV panels.

This paper addresses the problem of registering unorganized point clouds without prior information and proposes a robust automatic point cloud registration method. This method is designed to overcome the sensitivity of the classical Iterative Closest Point (ICP) algorithm to initial poses and its tendency to get trapped in local optima. The algorithm utilizes local information to filter feature points and construct feature descriptors, and then combines with the Random Sample Consensus (RANSAC) algorithm for coarse point cloud registration. The classical ICP algorithm is improved by using bidirectional search and constraints based on distance and curvature to obtain correct correspondences, achieving precise point cloud registration. Experimental results of point cloud registration demonstrate that this algorithm has significant advantages in terms of robustness and registration accuracy.