Zhuo Long

To address the challenges of insufficient peak response and inadequate characterization of spatial dependencies in short-term power grid load forecasting, this paper proposes a peak-enhanced spatio-temporal modeling framework. First, the method utilizes Markov Random Field (MRF) to transform time series data into a graph structure that explicitly reflects state transition patterns. Subsequently, by integrating an improved adaptive spatio-temporal mechanism (based on AGCRN), the model dynamically captures spatial dependencies. Furthermore, a temperature-scaled calibration module and a peak-weighted hybrid optimization strategy are introduced to significantly refine prediction accuracy during high-load periods. Using 15-minute sampling data from Yuxi City (2022–2023) as a case study, experimental results demonstrate that without incorporating external features the proposed method achieves comprehensive error metrics of MAE ≈ 0.035, RMSE ≈ 0.057, and MAPE ≈ 0.193. Moreover, substantial improvements are observed in peak intervals, validating the effectiveness and practicality of the proposed approach.

Wind turbine condition monitoring is crucial for sustainable energy systems, yet it faces challenges from complex environmental disturbances and multi scale temporal variations. Conventional State Reconstruction methods often rely on historical data for trend prediction, lacking nonlinear response adaptability and real time capability. To address this, this paper proposes a Multi branch Spatio temporal Feature Extraction module to extract multi scale spatio temporal features, enhancing comprehensive perception of local and global dynamic variations in multi variables. Further, a Temporal Multi Scale Attention Aggregation module is proposed for adaptive multi scale feature fusion, removing confusion from redundant features and strengthening feature causal reliability. Referring to nonlinear real time response systems, a Multi Scale Feature Aggregation Transformer is proposed to construct a real time response equivalent model of wind turbine states based on external environmental excitation and initial states, enabling reconstruction of standard reference healthy states. A state monitoring and anomaly detection framework is built based on this model. Experiments on public datasets show that the proposed model significantly outperforms benchmark models, achieving a Mean Absolute Error of 0.2671 and providing weeks earlier fault warnings. This paper proposes a multi scale feature fusion framework that enhances model accuracy and interpretability and provides a feasible solution for wind turbine health management.

In this study, a robust speed-stator flux tolerant predictive control (RS-SFTPC) strategy is proposed for the permanent magnet synchronous motor (PMSM) drives with parameters disturbance faults. The parameters disturbance sensitivity of speed external loop and stator flux internal loop is conducted for the PMSM drives system. Then, the inherent structure of predictive control model is modified by utilizing ultra-local model theory, and the improved tolerant predictive controller is applied to both the speed external loop and the stator flux internal loop. Moreover, a composite discrete extended state observer (ESO) is designed to diagnose parameters disturbance faults and observe input states, which can enhance the robustness of tolerant predictive control systems. The RS-SFTPC strategy are proposed for parameters disturbance faults diagnosis by utilizing the composite discrete ESO, which can eliminate the impact of parameters disturbance on the PMSM drives system. Finally, the comparison of simulation and experimental results show that the proposed RS-SFTPC method does not depend on any model parameters and always has better stator flux/ torque control performance than conventional PI-MPC method.

Harmonic drives (HDs) are key components of industrial robots, and their malfunction or breakdown can cause robot operational mistakes. Therefore, accurately diagnosing faults of HDs is of great significance for their applications. In this paper, fault diagnosis features of HDs were extracted from vibration signals using Short-Time Fourier Transform (STFT) by creating spatiotemporal graphs, and a Deep Reinforcement Learning (DRL) framework was employed to enhance diagnostic process, wherein agents were trained to assign structure of graph neural networks and aggregation strategies. Meanwhile, the ChebGCN was used to classify the fault characteristics of HDs under various fault states and working conditions. This was the first time that DRL and ChebGCN were jointly used for HD fault diagnosis. Comprehensive experiments were run to validate effectiveness of the proposed methods on two separate datasets, wherein diagnostic accuracy rates of 99.51% and 100% were achieved,respectively, indicating its great potential as a backbone for HD fault diagnosis.

Since the parameters mismatch fault caused by the mismatch between motor parameters and model predictive controller parameters can seriously affect the performance of the drive system, it is necessary to investigate parameters mismatch fault diagnosis and tolerant control. In this study, a novel robust cascaded model-free tolerant control(RCM-FTC) method with parameters mismatch fault diagnosis is proposed for permanent magnet synchronous motor(PMSM) drive system, which can effectively improve the control performance of stator flux and torque under parameters mismatch fault. The parameters mismatch fault sensitivity of the PMSM drive system are analyzed, and a novel ultralocal model is established under parameters mismatch fault. The multivariable integral terminal sliding mode observer (MIT-SMO) with improved sliding mode reaching law is designed to diagnose parameters mismatch fault and identify operating states in PMSM drive system. The robust cascaded speed loop and stator flux loop control strategy is proposed for parameters mismatch fault by utilizing the model-free sliding mode control algorithm, which can eliminate the impact of parameters mismatch fault on the PMSM drive system. The test results demonstrate the effectiveness of the proposed RCM-FTC method in significantly eliminating stator flux/torque ripple and improving fault tolerance capabilities in the case of parameters mismatch fault.

Short-term load forecasting is a crucial task within the power system. However, existing studies have overlooked the spatial-temporal relationships between multiple series loads. Accounting for this spatial-temporal adjacency can lead to more accurate forecasting in certain scenarios. In this paper, we propose a short-term load forecasting model named Parallel Spatial-Temporal Graph Attention Network (PST-GAT). The method encodes the multi-sequence loads as nodes and constructs the adjacency matrix using the Dynamic Time Warping (DTW) technique to form a fully connected graph of the load data. Combining the sliding window concept, the constructed fully connected graph is partitioned into a series of subgraphs, and the Graph Attention Network (GAT) performs feature extraction on each subgraph individually. To realize learning at multiple scales, PST-GAT adopts a parallel multi-branching approach, where each branch splits the subgraphs with varying lengths. Finally, the feature vectors extracted by each branch are concatenated, and forecasting is accomplished using a fully connected layer. Moreover, the unique structural design of PST-GAT allows for the simultaneous prediction of multiple sequence loadings. Experimental results based on real-world load data validate the superior prediction accuracy of this method compared to existing algorithms.

The frequent occurrence of motor faults has been a great disturbance to the development of production in various fields. Traditional fault diagnosis methods primarily use 1-D data or 2-D data. However, 3-D data holds significant promise for motor fault diagnosis due to its voluminous data and unique spatial information. This paper aims to explore a motor fault diagnosis method leveraging 3-D data. Nonetheless, motor fault 3-D data exhibits the limitation of lacking geometric structure. To address this limitation, This paper proposes a fault diagnosis method named SP-PointCNN. This method utilizes the proposed spherical projection (SP) method to mitigate the limitations present in the 3D data of the motor. Meanwhile, a neural network based on PointCNN is constructed to enable the utilization of 3D data in fault diagnosis. A series of experiments demonstrated the validity of the proposed method. After tested, the diagnosis accuracy of SP-PointCNN can reach 99.36%.

Existing fault diagnosis methods for rolling bearings are mostly applied in static domains and cannot adapt to continuous and dynamic industrial data in practice. In domain incremental scenarios, achieving continuous learning and avoiding catastrophic forgetting is a major challenge for fault diagnosis. In this article, a novel domain incremental learning method based on filtering columns and knowledge base (F&K-DIL) is proposed to achieve bearing fault diagnosis in dynamic domains. It consists of filtering columns, knowledge bases, and parameter update modules. In the filtering columns, two L1-norm are used to select representative samples, which can protect previously learned knowledge in each domain while learning new knowledge. The correctness of knowledge delivery is guaranteed. Sparsity parameters that can adjust the number of representative samples flexibly are introduced to improve storage efficiency. Representative samples in different domains are stored in the knowledge base to realize knowledge sharing. In the parameter update module, cross entropy and elastic weight consolidation loss function are used to realize constraints on important parameters in old domains. It can further avoid catastrophic forgetting. Through experiments on three rotating machinery datasets, F&K-DIL is used for fault diagnosis in domain incremental scenarios with an accuracy of up to 99.6% and 37% improvement in memory rate.