The operational stability of isolated industrial and mining microgrids is critically challenged by the diverse response characteristics of heterogeneous flexible resources (FRs), such as varying ramp rates and response times. Conventional scheduling methods, which often overlook these multi-granularity attributes, result in significant power imbalances, mismatched execution of the dispatch commands, and degradation of industrial product quality. To address this gap, this paper proposes a novel dispatch command disaggregation strategy for the precise coordination of these resources. The strategy's core innovation begins with establishing a pioneering quantitative flexibility model for lithium mine loads (LMLs), mechanistically exploring the thermo-electric dynamics of the salt-lake lithium extraction process. This specific model is then integrated within a unified framework designed to systematically characterize and quantify the multi-granularity flexibility attributes of diverse resources. Finally, a Discrete Choice Model (DCM) is employed to optimize the dispatch priority of FRs, effectively translating the microgrid's aggregate dispatch command into individualized and feasible setpoints. Validated on a real-world microgrid in Xizang, the proposed strategy reduces system power deviation by 89.05% and decreases the total operating cost by 25.25%. This work provides a practical framework for enhancing the control precision and economic efficiency of isolated microgrids through effective coordination of diverse flexibility assets.

Accurate detection of forced oscillations is important for the stable operation of power systems. The method based on prior knowledge relies on manual feature extraction, which has limited ability to characterize non-stationary signals. While deep learning (DL) methods can automatically learn features, they may overlook the physical mechanisms of power systems, potentially leading to misjudgments. We propose a Hybrid Knowledge-DL network (HKD-SVM) that utilizes Support Vector Machine (SVM) as the classifier. In our method, Discrete Wavelet Transform (DWT) is used to represent the time–frequency structure of the input signals, and DL features are extracted by Convolutional Neural Network (CNN) from this time–frequency representation. These learned features are subsequently fused with prior knowledge features that carry explicit physical interpretations, thereby constructing a more discriminative feature representation space. Finally, SVM is adopted as the classifier, making the network well-suited for nonlinear, high-dimensional classification scenarios with limited training samples, which are common in power system applications. Experiments on both simulated and real-world phasor measurement unit (PMU) data demonstrate that HKD-SVM outperforms purely data-driven and purely knowledge-driven methods. The proposed method provides an effective solution for power system oscillation detection.

Deep learning (DL) holds significant potential for distinguishing forced oscillations (FOs) from natural oscillations (NOs) and resonances in power systems. However, most existing studies treat DL models as opaque system classifiers, and these models are often susceptible to challenges such as class imbalance among the three oscillation patterns (i.e., FOs, NOs, and resonances) and the limited size of real-world training datasets. These issues can lead to incorrect diagnoses of oscillation patterns in practical applications. To address these challenges, we propose a FO oscillation recognition framework that integrates balanced data processing, an interpretable recognition network, and a fine-tuning learning strategy. In this framework, DeepSMOTE1D is introduced to address the issue of class imbalance by oversampling features extracted from an encoder-decoder network. The interpretable DWT-CNN-LSTM recognition network incorporates multiple discrete wavelet transforms (DWT) into each convolutional and LSTM layer to enhance both interpretability and performance of the model. Finally, a fine-tuning transfer learning strategy is employed, where the recognition network is first pretrained on simulation data and subsequently retrained on limited real-world data. The effectiveness of DeepSMOTE1D and the fine-tuning strategy in the proposed framework is also evaluated. Experimental results show that the proposed recognition model not only learns effectively from simulated oscillation data but also outperforms existing related networks in real-world scenarios, achieving superior accuracy, precision, recall, and F1-score.

Forced oscillation detection represents a critical challenge in power system stability monitoring. Most conventional detection methods exhibit certain limitations when dealing with complex real-world scenarios under time-varying and non-stationary conditions, whereas deep learning methods typically rely on sufficient labeled data and suffer from a lack of interpretability. To address these limitations, this article proposes a novel framework for forced oscillation detection, which integrates Discrete Wavelet Transform (DWT), Dilated Convolutional Neural Networks (DiCNN), and a Model-Agnostic Meta-Learning (MAML) training strategy. The proposed DWT+DiCNN architecture enhances model interpretability by leveraging DWT's time-frequency analysis capabilities. Concurrently, DiCNN extends the receptive field to effectively capture multi-scale temporal features. To overcome the challenge of limited labeled data, the MAML strategy is implemented to enable the proposed model to rapidly adapt to new detection tasks with minimal samples in various operating scenarios. Comprehensive experimental evaluations demonstrate the model's superior performance on both simulated and real-world Phasor Measurement Unit (PMU) data, achieving accurate discrimination between forced and natural oscillations while exhibiting adaptability in complex power system environments.

Photovoltaic (PV) systems are highly susceptible to various faults during long-term outdoor operation, while severe class imbalance, noise interference, and strong similarity among fault patterns significantly degrade diagnostic performance. To address these challenges, this paper proposes a PV fault diagnosis framework that integrates DeepSMOTE-edited nearest-neighbor (ENN) with a lightweight two-dimensional convolutional neural network (lightweight-2DCNN). An DeepSMOTE-ENN module is first developed to construct a balanced dataset by generating high-quality minority-class samples in the latent space and eliminating ambiguous boundary samples. Wavelet-based time-frequency analysis is then employed to transform one-dimensional electrical signals into discriminative two-dimensional representations. Based on these representations, lightweight-2DCNN is designed to extract multi-scale spatial features with low computational complexity. Experimental results on a simulated PV array demonstrate that the proposed framework achieves superior diagnostic performance for both single and composite faults. With only 0.96 M parameters, the proposed method attains an overall accuracy of 98.35 +/- 0.40%, outperforming traditional resampling strategies and existing lightweight models, thereby providing an efficient and robust solution for practical PV fault diagnosis.

Infrared (IR) imaging system is widely used in military, medical, and other fields due to its high sensitivity and anti-interference ability. However, IR small targets pose challenges for the detection task in complex environments, due to their blurry shapes, small sizes, and being overwhelmed by background clutter easily. In this paper, we propose an innovative Progressive Multi-Level Feature Fusion network (PMFFNet) to solve the related problems in IR small target detection, i.e., feature loss and confusion with the background. Specifically, we design a Global-Local Feature Enhancement module with an attention mechanism to enhance and combine the global semantic and local context information, which improves the discriminability of IR small targets and complex surroundings. Further, during the upsampling process, a Progressive Multi-Level Feature Fusion module is proposed to achieve cross-layer interaction among high-level semantic and low-level fine details, and implement the fusion of different-level enhanced features. The experimental results demonstrate that the proposed PMFFNet achieves satisfactory performance and outperforms other representative models on three public datasets including various complex scenarios.

The proportion of new energy sources, such as wind, photovoltaic and hydropower, in the power grid is increasing year by year. In addition, a large number of nonlinear loads are connected to the grid, resulting in frequent power quality disturbances (PQDs), which pose challenges to the stability and reliability of the power system. Accurate identification of these disturbances is crucial for effective grid management and protection. Although deep learning methods have high accuracy, their lack of interpretability can limit their acceptance in engineering applications. Traditional signal analysis has a good physical foundation, but it is not integrated with deep learning to a sufficient degree. To address these issues, we propose the DWT-1DCNN-LSTM network as an interpretable model for PQD classification. This method effectively decomposes the time-domain signals into sub-signals in different frequency bands by employing the Discrete Wavelet Transform (DWT), which enhances the anti-interference capability of the classification model. This approach enhances the interference resilience of the classification model through the incorporation of the Discrete Wavelet Transform (DWT), which effectively decomposes time-domain signals into sub-signals across different frequency bands. The one-dimensional Convolutional Neural Network (1DCNN) then extracts local features, while the Long Short-Term Memory network (LSTM) analyzes temporal dependencies of the transformed sub-signals. Experimental validation with simulated datasets demonstrates that the DWT-1DCNN-LSTM model achieves an accuracy of 99.27%, outperforming the DWT-1DCNN, 1DCNN-LSTM, LSTM, and CNN models by 1.59%, 1.13%, 1.44%, and 6.48%, respectively. The robustness provided by the DWT module makes the model well suited for PQDs in environments with large disturbances, helping to detect and mitigate PQDs in a timely manner and ultimately contributing to improved power quality and system reliability.

The detection of forced oscillations, especially distinguishing them from natural oscillations, has emerged as a major concern in power system stability monitoring. Deep learning (DL) holds significant potential for detecting forced oscillations correctly. However, existing artificial neural networks (ANNs) face challenges when employed in edge devices for timely detection due to their inherent complex computations and high power consumption. This paper proposes a novel hybrid network that integrates a spiking recurrent neural network (SRNN) with long short-term memory (LSTM). The SRNN achieves computational and energy efficiency, while the integration with LSTM is conducive to effectively capturing temporal dependencies in time-series oscillation data. The proposed hybrid network is trained using the backpropagation-through-time (BPTT) optimization algorithm, with adjustments made to address the discontinuous gradient in the SRNN. We evaluate our proposed model on both simulated and real-world oscillation datasets. Overall, the experimental results demonstrate that the proposed model can achieve higher accuracy and superior performance in distinguishing forced oscillations from natural oscillations, even in the presence of strong noise, compared to pure LSTM and other SRNN-related models.

Partial discharge (PD) signals are used to evaluate the insulation condition of stators in electrical machines. Their measurements are often heavily corrupted by ambient noise, making denoising essential for effective detection and analysis of PD signals. Wavelet thresholding techniques are widely applied to denoise PD signals. However, existing hard and soft thresholding functions introduce oscillation or deviation into PD signals after wavelet reconstruction, particularly under high-noise conditions. This paper proposes an improved thresholding function for the wavelet threshold denoising method that effectively overcomes the oscillation issue associated with the hard thresholding function and the constant deviation of the soft thresholding function. Additionally, wavelet basis selection based on the correlation coefficient and an adaptive threshold value is integrated with the improved thresholding function to implement the wavelet threshold denoising method. The proposed technique is applied to both simulated and real-world measured PD signals to evaluate its performance across different signal-to-noise ratio (SNR) levels. Compared with traditional soft and hard thresholding functions, simulation results confirm the superiority of the improved thresholding function, especially under high-noise conditions. At an input Gaussian noise level of −10 dB, the proposed method yielded an SNR that was 1.20 dB and 2.66 dB higher than those of the hard and soft thresholding functions, respectively.

To realize the daily detection of human respiratory frequency effectively and conveniently, a non-contact respiratory frequency detection method based on depth image information obtained by RealSense camera was proposed. The proposed maximum distance limit method and the maximum external contour extraction algorithm were used to eliminate the influence of cavity noise and close-range interference objects respectively, and the depth image of human body was obtained. The human skeleton was extracted using Zhang-Suen refinement algorithm, and the location of the thoracic breathing area was realized by combining the skeleton joint information. The principal component analysis algorithm（PCA） was used to extract the respiratory signal data contained in the depth image of the chest area, and the human respiratory frequency was further solved using the peak detection method. Experimental results show that the proposed method can effectively non-contactually realize the detection of human respiratory frequency with high accuracy.

Due to nonlinear components in the charging piles of electric vehicles, harmonics and nonstationary signals in the electric vehicle charging load bring voltage and current distortion, seriously affecting the accuracy of the power-related calculation in nonsinusoidal environments. This paper proposed a new approach to calculate the active power and root mean square values from decomposed components using the adaptive chirp mode decomposition (ACMD) method on voltage and current. The advantage of the ACMD-based method is that it correctly provides the power-related quantities of harmonics or nonstationary components for the electric vehicle charging load. The performance of the proposed method was verified using synthetic signals and simulation tests. The experimental results presented better estimations for each quantity defined in IEEE Standard 1459-2010, compared with the discrete wavelet transform approach.

In recent years, subsynchronous resonance (SSR) has frequently occurred in DFIG-connected series-compensated systems. For the analysis and prevention, it is of great importance to achieve wide area monitoring of the incident. This paper presents a Hankel dynamic mode decomposition (DMD) method to identify SSR parameters using synchrophasor data. The basic idea is to employ the DMD technique to explore the subspace of Hankel matrices constructed by synchrophasors. It is analytically demonstrated that the subspace of these Hankel matrices is a combination of fundamental and SSR modes. Therefore, the SSR parameters can be calculated once the modal parameter is extracted. Compared with the existing method, the presented work has better dynamic performances as it requires much less data. Thus, it is more suitable for practical cases in which the SSR characteristics are time-varying. The effectiveness and superiority of the proposed method have been verified by both simulations and field data.

Load disaggregation determines appliance-level energy consumption unintrusively from aggregated consumption measured by a single meter. Deep neural networks have been proven to have great potential in load disaggregation. In this article, a temporal convolution network, mainly consisting of residual blocks with bidirectional dilated convolution, the GeLu activation function, and multihead attention, is proposed to improve the prediction accuracy of individual appliances. Bidirectional dilated convolution is applied to enlarge the receptive field and effectively extract load features from historical and future information. Meanwhile, GeLU is introduced into the residual structure to overcome the "dead state" issue of traditional ReLU. Furthermore, multihead attention aims to improve the prediction accuracy by giving different weights according to the importance of different-level load features. The proposed model is validated using the REDD and UK-DALE datasets. Among six existing neural networks, the experimental results demonstrate that the proposed algorithm achieves the least average errors when disaggregating four appliances in terms of mean absolute error (MAE) and signal aggregate error (SAE), respectively, reduced by 22.33% and 60.58% compared with the model with the second-best performance on the REDD dataset. Additionally, the proposed algorithm shows superior results in identifying the on/off state in four appliances from the UK-DALE dataset.

In recent years, subsynchronous control interaction (SSCI) has frequently taken place in renewable-connected power systems. To counter this issue, utilities have been seeking tools for fast and accurate identification of SSCI events. The main challenges of SSCI monitoring are the time-varying nature and uncertain modes of SSCI events. Accordingly, this paper presents a simple but effective method that takes advantage of intrinsic time-scale decomposition (ITD). The main purpose is to improve the accuracy and robustness of ITD by incorporating the least-squares method. Results show that the proposed method strikes a good balance between dynamic performance and estimation accuracy. More importantly, the method does not require any prior information, and its performance is therefore not affected by the frequency constitution of the SSCI. Comprehensive comparative studies are conducted to demonstrate the usefulness of the method through synthetic signals, electromagnetic temporary program (EMTP) simulations, and field-recorded SSCI data. Finally, real-time simulation tests are conducted to show the feasibility of the method for real-time monitoring.

In recent years, subsynchronous control interaction (SSCI) frequently occurred due to the rapid development of inverter based resources. Online monitoring and identifying SSCI are thus of great importance for the safe operation of power systems. This paper proposes a data-driven method for SSCI identification using synchrophasors generated by phasor measurement units (PMUs). The challenges of this work mainly result from two factors: 1) SSCI may simultaneously involve a sub-synchronous oscillation (Sub-SO) mode and a super-synchronous oscillation (Sup-SO) mode. The spectra of these two modes would be aliased due to the limited reporting rate of PMUs. 2) The reporting filter in PMUs significantly weakens the oscillation signal as the frequency of SSCI is generally far away from the nominal frequency. To tackle these challenges, the influence of Sub-SO and Sup-SO on spectra of synchrophasors is analyzed first. It is revealed that the positive and negative spectra are linearly correlated with Sub-SO and Sup-SO modes. Several features are then extracted via the analysis of the spectra which can reflect the SSCI features and a data-driven identification method is further proposed through a fast reduced kernel extreme learning machine (RKELM) model. Compared with the traditional classification techniques, the RKELM algorithm based on kernel function has stronger generalization performance and higher accuracy while ensuring shorter training time. This significantly improves the accuracy and efficiency of SSCI identification, especially under noise conditions. Both simulations and field tests demonstrate the effectiveness and usefulness of the proposed method.

为解决电能质量扰动的分类问题,利用多分辨奇异值分解(Singular Value Decomposition,SVD)的信号逐层分解方式,提出基于多分辨SVD包与随机森林(Multi-Resolution SVD and Random Forest,MRSVD-RF)的电能质量扰动分类方法.通过实验证明了该算法对单一和复合电能质量信号的分类效果明显优于分解结构相似的基于的小波包的信号分解方式,比较了分类器模型的选择和特征提取数量对算法性能的影响.

Subsynchronous resonance (Sub-SR) is traditionally caused by turbine generators. In recent years, a new type of Sub-SR has occurred that is related to inverter-based resources (IBRs). In those events, supersynchronous resonance (Sup-SR) accompanied Sub-SR, a phenomenon commonly called frequency coupling. It is thus of great significance to detect whether there is a supersynchronous component, which can help to perform postincident analysis and design a mitigation strategy. This paper uses synchrophasor data provided by phasor measurement units (PMUs) to detect the Sup-SR. Due to the limitation of the reporting rate of PMUs, spectrum aliasing occurs between Sub-SR and Sup-SR, which imposes great challenges for detection. To address this issue, the influence of Sup-SR and Sub-SR on the positive and negative spectra of synchrophasors is analyzed. The results show that Sup-SR primarily affects the negative spectra of synchrophasors. Based on this finding, the difference between the measured spectra and the theoretical spectra with pure Sub-SR is utilized to detect Sup-SR. An adaptive threshold is further proposed using a histogram-based statistical approach to enhance the detection accuracy. Case studies demonstrate that the proposed algorithm can accurately detect the presence of Sup-SR from synchrophasor data.

Abstract Identifying the parameters of sub/super‐synchronous oscillation (SubSO/SupSO) signals is challenging, because (1) the mode numbers of the oscillation signals are unknown, and (2) the oscillation signals are non‐stationary and noisy. The non‐stationary components need to be accurately identified in order to efficiently suppress SubSO/SupSO. Here, an adaptive chirp mode decomposition (ACMD) method joining a synchroextracting transform (SET) is exploited. First, for detecting the initial instantaneous frequencies (IFs) of SubSO and SupSO, SET is employed to provide time–frequency representations with satisfied energy concentration to ACMD. Second, ACMD is applied to decomposing the main components from the oscillation signal without needing to use the mode numbers, and directly compute the final IFs and instantaneous amplitudes (IAs) for SubSO and SupSO without Hilbert transform. And third, the damping factors of the SubSO and SupSO are further identified with the IAs. The proposed method is evaluated over synthetic signals, simulation data, and real field data from a wind system. Experimental results demonstrate that, compared with the classical signal decomposition‐based identification methods, the proposed method more accurately detects frequencies, amplitudes, and damping factors of SubSO and SupSO components for non‐stationary and noisy signals.

电力电缆早期故障严重威胁用电安全且难以准确识别,在基于特征提取与特征选择的识别方法中,一旦不能准确获得关键特征信息会直接导致识别精度下降.鉴于此,提出一种基于S变换特征提取和最大相关最小冗余(mRMR)特征选择的电缆早期故障识别方法.对故障相电流进行S变换,提取一些具有相关性、冗余性的统计量、熵和能量等构成初始特征集;采用mRMR选择出具有最佳分类效果的特征子集;利用带核函数的SVM分类器对多种电缆故障进行识别.仿真结果表明,在不同噪声环境下该方法在识别精度和鲁棒性方面都优于同类算法.