With the continuous expansion and increasing complexity of power system scales,the binary classifica-tion for transient stability assessment in power systems can no longer meet the safety requirements of power system ...With the continuous expansion and increasing complexity of power system scales,the binary classifica-tion for transient stability assessment in power systems can no longer meet the safety requirements of power system control and regulation.Therefore,this paper proposes a multi-class transient stability assessment model based on an improved Transformer.The model is designed with a dual-tower encoder structure:one encoder focuses on the time dependency of data,while the other focuses on the dynamic correlations between variables.Feature extraction is conducted from both time and variable perspectives to ensure the completeness of the feature extraction process,thereby enhancing the accuracy of multi-class evaluation in power systems.Additionally,this paper introduces a hybrid sampling strategy based on sample boundaries,which addresses the issue of sample imbalance by increasing the number of boundary samples in the minority class and reducing the number of non-boundary samples in the majority class.Considering the frequent changes in power grid topology or operation modes,this paper proposes a two-stage updating scheme based on self-supervised learning:In the first stage,self-supervised learning is employed to mine the structural information from unlabeled data in the target domain,enhancing the model’s generalization capability in new scenarios.In the second stage,a sample screening mechanism is used to select key samples,which are labeled through long-term simulation techniques for fine-tuning the model parameters.This allows for rapid model updates without relying on many labeled samples.This paper’s proposed model and update scheme have been simulated and verified on two node systems,the IEEE New England 10-machine 39-bus system and the IEEE 47-machine 140-bus system,demonstrating their effectiveness and reliability.展开更多
As technologies related to power equipment fault diagnosis and infrared temperature measurement continue to advance,the classification and identification of infrared temperature measurement images have become crucial ...As technologies related to power equipment fault diagnosis and infrared temperature measurement continue to advance,the classification and identification of infrared temperature measurement images have become crucial in effective intelligent fault diagnosis of various electrical equipment.In response to the increasing demand for sufficient feature fusion in current real-time detection and low detection accuracy in existing networks for Substation fault diagnosis,we introduce an innovative method known as Gather and Distribution Mechanism-You Only Look Once(GD-YOLO).Firstly,a partial convolution group is designed based on different convolution kernels.We combine the partial convolution group with deep convolution to propose a new Grouped Channel-wise Spatial Convolution(GCSConv)that compensates for the information loss caused by spatial channel convolution.Secondly,the Gather and Distribute Mechanism,which addresses the fusion problem of different dimensional features,has been implemented by aligning and sharing information through aggregation and distribution mechanisms.Thirdly,considering the limitations in current bounding box regression and the imbalance between complex and simple samples,Maximum Possible Distance Intersection over Union(MPDIoU)and Adaptive SlideLoss is incorporated into the loss function,allowing samples near the Intersection over Union(IoU)to receive more attention through the dynamic variation of the mean Intersection over Union.The GD-YOLO algorithm can surpass YOLOv5,YOLOv7,and YOLOv8 in infrared image detection for electrical equipment,achieving a mean Average Precision(mAP)of 88.9%,with accuracy improvements of 3.7%,4.3%,and 3.1%,respectively.Additionally,the model delivers a frame rate of 48 FPS,which aligns with the precision and velocity criteria necessary for the detection of infrared images in power equipment.展开更多
Real-time prediction of the rock mass class in front of the tunnel face is essential for the adaptive adjustment of tunnel boring machines(TBMs).During the TBM tunnelling process,a large number of operation data are g...Real-time prediction of the rock mass class in front of the tunnel face is essential for the adaptive adjustment of tunnel boring machines(TBMs).During the TBM tunnelling process,a large number of operation data are generated,reflecting the interaction between the TBM system and surrounding rock,and these data can be used to evaluate the rock mass quality.This study proposed a stacking ensemble classifier for the real-time prediction of the rock mass classification using TBM operation data.Based on the Songhua River water conveyance project,a total of 7538 TBM tunnelling cycles and the corresponding rock mass classes are obtained after data preprocessing.Then,through the tree-based feature selection method,10 key TBM operation parameters are selected,and the mean values of the 10 selected features in the stable phase after removing outliers are calculated as the inputs of classifiers.The preprocessed data are randomly divided into the training set(90%)and test set(10%)using simple random sampling.Besides stacking ensemble classifier,seven individual classifiers are established as the comparison.These classifiers include support vector machine(SVM),k-nearest neighbors(KNN),random forest(RF),gradient boosting decision tree(GBDT),decision tree(DT),logistic regression(LR)and multilayer perceptron(MLP),where the hyper-parameters of each classifier are optimised using the grid search method.The prediction results show that the stacking ensemble classifier has a better performance than individual classifiers,and it shows a more powerful learning and generalisation ability for small and imbalanced samples.Additionally,a relative balance training set is obtained by the synthetic minority oversampling technique(SMOTE),and the influence of sample imbalance on the prediction performance is discussed.展开更多
In contemporary medium-voltage distribution networks heavily penetrated by distributed energy resources(DERs),the harmonic components injected by power-electronic interfacing converters,together with the inherently in...In contemporary medium-voltage distribution networks heavily penetrated by distributed energy resources(DERs),the harmonic components injected by power-electronic interfacing converters,together with the inherently intermittent output of renewable generation,distort the zero-sequence current and continuously reshape its frequency spectrum.As a result,single-line-to-ground(SLG)faults exhibit a pronounced,strongly non-stationary behaviour that varies with operating point,load mix and DER dispatch.Under such circumstances the performance of traditional rule-based algorithms—or methods that rely solely on steady-state frequency-domain indicators—degrades sharply,and they no longer satisfy the accuracy and universality required by practical protection systems.To overcome these shortcomings,the present study develops an SLG-fault identification scheme that transforms the zero-sequence currentwaveforminto two-dimensional image representations and processes themwith a convolutional neural network(CNN).First,the causes of sample-distribution imbalance are analysed in detail by considering different neutralgrounding configurations,fault-inception mechanisms and the statistical probability of fault occurrence on each phase.Building on these insights,a discriminator network incorporating a Convolutional Block Attention Module(CBAM)is designed to autonomously extract multi-layer spatial-spectral features,while Gradient-weighted Class Activation Mapping(Grad-CAM)is employed to visualise the contribution of every salient image region,thereby enhancing interpretability.A comprehensive simulation platform is subsequently established for a DER-rich distribution system encompassing several representative topologies,feeder lengths and DER penetration levels.Large numbers of realistic SLG-fault scenarios are generated—including noise and measurement uncertainty—and are used to train,validate and test the proposed model.Extensive simulation campaigns,corroborated by field measurements from an actual utility network,demonstrate that the proposed approach attains an SLG-fault identification accuracy approaching 100 percent and maintains robust performance under severe noise conditions,confirming its suitability for real-world engineering applications.展开更多
Decentralized machine learning frameworks,e.g.,federated learning,are emerging to facilitate learning with medical data under privacy protection.It is widely agreed that the establishment of an accurate and robust med...Decentralized machine learning frameworks,e.g.,federated learning,are emerging to facilitate learning with medical data under privacy protection.It is widely agreed that the establishment of an accurate and robust medical learning model requires a large number of continuous synchronous monitoring data of patients from various types of monitoring facilities.However,the clinic monitoring data are usually sparse and imbalanced with errors and time irregularity,leading to inaccurate risk prediction results.To address this issue,this paper designs a medical data resampling and balancing scheme for federated learning to eliminate model biases caused by sample imbalance and provide accurate disease risk prediction on multi-center medical data.Experimental results on a real-world clinical database MIMIC-Ⅳ demonstrate that the proposed method can improve AUC(the area under the receiver operating characteristic) from 50.1% to 62.8%,with a significant performance improvement of accuracy from 76.8% to 82.2%,compared to a vanilla federated learning artificial neural network(ANN).Moreover,we increase the model’s tolerance for missing data from 20% to 50% compared with a stand-alone baseline model.展开更多
With tunnel boring machine being used in underground engineering,accurate geological indicators have been the important basis for tunnel boring machine(TBM)construction.Back propagation neural network(BPNN)has been us...With tunnel boring machine being used in underground engineering,accurate geological indicators have been the important basis for tunnel boring machine(TBM)construction.Back propagation neural network(BPNN)has been used to predict the geological indicators of tunnels in previous studies.Nevertheless,these studies ignored the imbalance proportion of surrounding rock grades,leading to the indiscriminate use of data,thus affecting the predictive effect of BPNN.In order to prove the importance of the proportion of surround-ing rock grade in geological prediction,we mainly attempt to utilize particle swarm optimization(PSO)to optimize the proportion of sample data,and integrate with BPNN to establish a PSO-BPNN theoretical model to predict geological indicators.At the same time,combined with the actual engineering data,5 tunneling indicators were selected as input and 4 geological indicators were selected as out-put by a variety of dimensionality reduction methods.The geological indicators are density,uniaxial compressive strength,internal fric-tion angle(u)and Poisson’s ratio(e).On this basis,the PSO-BPNN prediction model was established in detail.By comparing the prediction of traditional BPNN,PSO-BPNN and other optimization-integrated models,the result shows that optimized proportion of surrounding rock grades reduces the prediction error and improves the interpretability of the prediction model.Meanwhile,we com-bined the theory of surrounding rock partition to illustrate the rationality of surrounding rock proportion in PSO result,that is,the proportion of complex surrounding rock should be increased appropriately to improve the prediction result.Ultimately,based on the optimization-integrated models with engineering data and the surrounding rock classification theory,the importance of proportion of surrounding rock grades for tunnel geological prediction is confirmed.展开更多
基金the National Natural Science Foundation of China(5227-7084).
文摘With the continuous expansion and increasing complexity of power system scales,the binary classifica-tion for transient stability assessment in power systems can no longer meet the safety requirements of power system control and regulation.Therefore,this paper proposes a multi-class transient stability assessment model based on an improved Transformer.The model is designed with a dual-tower encoder structure:one encoder focuses on the time dependency of data,while the other focuses on the dynamic correlations between variables.Feature extraction is conducted from both time and variable perspectives to ensure the completeness of the feature extraction process,thereby enhancing the accuracy of multi-class evaluation in power systems.Additionally,this paper introduces a hybrid sampling strategy based on sample boundaries,which addresses the issue of sample imbalance by increasing the number of boundary samples in the minority class and reducing the number of non-boundary samples in the majority class.Considering the frequent changes in power grid topology or operation modes,this paper proposes a two-stage updating scheme based on self-supervised learning:In the first stage,self-supervised learning is employed to mine the structural information from unlabeled data in the target domain,enhancing the model’s generalization capability in new scenarios.In the second stage,a sample screening mechanism is used to select key samples,which are labeled through long-term simulation techniques for fine-tuning the model parameters.This allows for rapid model updates without relying on many labeled samples.This paper’s proposed model and update scheme have been simulated and verified on two node systems,the IEEE New England 10-machine 39-bus system and the IEEE 47-machine 140-bus system,demonstrating their effectiveness and reliability.
基金Science and Technology Department of Jilin Province(No.20200403075SF)Education Department of Jilin Province(No.JJKH20240148KJ).
文摘As technologies related to power equipment fault diagnosis and infrared temperature measurement continue to advance,the classification and identification of infrared temperature measurement images have become crucial in effective intelligent fault diagnosis of various electrical equipment.In response to the increasing demand for sufficient feature fusion in current real-time detection and low detection accuracy in existing networks for Substation fault diagnosis,we introduce an innovative method known as Gather and Distribution Mechanism-You Only Look Once(GD-YOLO).Firstly,a partial convolution group is designed based on different convolution kernels.We combine the partial convolution group with deep convolution to propose a new Grouped Channel-wise Spatial Convolution(GCSConv)that compensates for the information loss caused by spatial channel convolution.Secondly,the Gather and Distribute Mechanism,which addresses the fusion problem of different dimensional features,has been implemented by aligning and sharing information through aggregation and distribution mechanisms.Thirdly,considering the limitations in current bounding box regression and the imbalance between complex and simple samples,Maximum Possible Distance Intersection over Union(MPDIoU)and Adaptive SlideLoss is incorporated into the loss function,allowing samples near the Intersection over Union(IoU)to receive more attention through the dynamic variation of the mean Intersection over Union.The GD-YOLO algorithm can surpass YOLOv5,YOLOv7,and YOLOv8 in infrared image detection for electrical equipment,achieving a mean Average Precision(mAP)of 88.9%,with accuracy improvements of 3.7%,4.3%,and 3.1%,respectively.Additionally,the model delivers a frame rate of 48 FPS,which aligns with the precision and velocity criteria necessary for the detection of infrared images in power equipment.
基金funded by the National Natural Science Foundation of China(Grant No.41941019)the State Key Laboratory of Hydroscience and Engineering(Grant No.2019-KY-03)。
文摘Real-time prediction of the rock mass class in front of the tunnel face is essential for the adaptive adjustment of tunnel boring machines(TBMs).During the TBM tunnelling process,a large number of operation data are generated,reflecting the interaction between the TBM system and surrounding rock,and these data can be used to evaluate the rock mass quality.This study proposed a stacking ensemble classifier for the real-time prediction of the rock mass classification using TBM operation data.Based on the Songhua River water conveyance project,a total of 7538 TBM tunnelling cycles and the corresponding rock mass classes are obtained after data preprocessing.Then,through the tree-based feature selection method,10 key TBM operation parameters are selected,and the mean values of the 10 selected features in the stable phase after removing outliers are calculated as the inputs of classifiers.The preprocessed data are randomly divided into the training set(90%)and test set(10%)using simple random sampling.Besides stacking ensemble classifier,seven individual classifiers are established as the comparison.These classifiers include support vector machine(SVM),k-nearest neighbors(KNN),random forest(RF),gradient boosting decision tree(GBDT),decision tree(DT),logistic regression(LR)and multilayer perceptron(MLP),where the hyper-parameters of each classifier are optimised using the grid search method.The prediction results show that the stacking ensemble classifier has a better performance than individual classifiers,and it shows a more powerful learning and generalisation ability for small and imbalanced samples.Additionally,a relative balance training set is obtained by the synthetic minority oversampling technique(SMOTE),and the influence of sample imbalance on the prediction performance is discussed.
基金supported by the Science and Technology Program of China Southern Power Grid(031800KC23120003).
文摘In contemporary medium-voltage distribution networks heavily penetrated by distributed energy resources(DERs),the harmonic components injected by power-electronic interfacing converters,together with the inherently intermittent output of renewable generation,distort the zero-sequence current and continuously reshape its frequency spectrum.As a result,single-line-to-ground(SLG)faults exhibit a pronounced,strongly non-stationary behaviour that varies with operating point,load mix and DER dispatch.Under such circumstances the performance of traditional rule-based algorithms—or methods that rely solely on steady-state frequency-domain indicators—degrades sharply,and they no longer satisfy the accuracy and universality required by practical protection systems.To overcome these shortcomings,the present study develops an SLG-fault identification scheme that transforms the zero-sequence currentwaveforminto two-dimensional image representations and processes themwith a convolutional neural network(CNN).First,the causes of sample-distribution imbalance are analysed in detail by considering different neutralgrounding configurations,fault-inception mechanisms and the statistical probability of fault occurrence on each phase.Building on these insights,a discriminator network incorporating a Convolutional Block Attention Module(CBAM)is designed to autonomously extract multi-layer spatial-spectral features,while Gradient-weighted Class Activation Mapping(Grad-CAM)is employed to visualise the contribution of every salient image region,thereby enhancing interpretability.A comprehensive simulation platform is subsequently established for a DER-rich distribution system encompassing several representative topologies,feeder lengths and DER penetration levels.Large numbers of realistic SLG-fault scenarios are generated—including noise and measurement uncertainty—and are used to train,validate and test the proposed model.Extensive simulation campaigns,corroborated by field measurements from an actual utility network,demonstrate that the proposed approach attains an SLG-fault identification accuracy approaching 100 percent and maintains robust performance under severe noise conditions,confirming its suitability for real-world engineering applications.
基金supported by Hubei Provincial Development and Reform Commission Program"Hubei Big Data Analysis Platform and Intelligent Service Project for Medical and Health"。
文摘Decentralized machine learning frameworks,e.g.,federated learning,are emerging to facilitate learning with medical data under privacy protection.It is widely agreed that the establishment of an accurate and robust medical learning model requires a large number of continuous synchronous monitoring data of patients from various types of monitoring facilities.However,the clinic monitoring data are usually sparse and imbalanced with errors and time irregularity,leading to inaccurate risk prediction results.To address this issue,this paper designs a medical data resampling and balancing scheme for federated learning to eliminate model biases caused by sample imbalance and provide accurate disease risk prediction on multi-center medical data.Experimental results on a real-world clinical database MIMIC-Ⅳ demonstrate that the proposed method can improve AUC(the area under the receiver operating characteristic) from 50.1% to 62.8%,with a significant performance improvement of accuracy from 76.8% to 82.2%,compared to a vanilla federated learning artificial neural network(ANN).Moreover,we increase the model’s tolerance for missing data from 20% to 50% compared with a stand-alone baseline model.
基金supported by the National Natural Science Foundation of China,China(Grant No.52075370).
文摘With tunnel boring machine being used in underground engineering,accurate geological indicators have been the important basis for tunnel boring machine(TBM)construction.Back propagation neural network(BPNN)has been used to predict the geological indicators of tunnels in previous studies.Nevertheless,these studies ignored the imbalance proportion of surrounding rock grades,leading to the indiscriminate use of data,thus affecting the predictive effect of BPNN.In order to prove the importance of the proportion of surround-ing rock grade in geological prediction,we mainly attempt to utilize particle swarm optimization(PSO)to optimize the proportion of sample data,and integrate with BPNN to establish a PSO-BPNN theoretical model to predict geological indicators.At the same time,combined with the actual engineering data,5 tunneling indicators were selected as input and 4 geological indicators were selected as out-put by a variety of dimensionality reduction methods.The geological indicators are density,uniaxial compressive strength,internal fric-tion angle(u)and Poisson’s ratio(e).On this basis,the PSO-BPNN prediction model was established in detail.By comparing the prediction of traditional BPNN,PSO-BPNN and other optimization-integrated models,the result shows that optimized proportion of surrounding rock grades reduces the prediction error and improves the interpretability of the prediction model.Meanwhile,we com-bined the theory of surrounding rock partition to illustrate the rationality of surrounding rock proportion in PSO result,that is,the proportion of complex surrounding rock should be increased appropriately to improve the prediction result.Ultimately,based on the optimization-integrated models with engineering data and the surrounding rock classification theory,the importance of proportion of surrounding rock grades for tunnel geological prediction is confirmed.
