626 results about "Reconstruction error" patented technology

The invention discloses a blocked pipe section identification method of a drainage pipe network system based on dynamic characteristics, and relates to the technical field of drainage pipe network monitoring. Real-time hydraulic association between pipe network nodes is quantified through a dynamic adjacency matrix to generate a spatial topology matrix, and a time sequence is divided based on a sliding window; a space-time fusion model based on GCN and Transform is constructed to carry out multi-scale dynamic coding, and then a decoder is utilized to reconstruct normal working condition data of a pipe network. By calculating the deviation degree of the pipe section level reconstruction error and the threshold value, the blocked pipe section is accurately recognized, end-to-end modeling from'pipe network topology-drainage time sequence data-external rainfall 'multi-source data to blocked pipe section recognition is achieved, and the technical difficulties of a traditional method in the aspects of dynamic topology modeling, long time sequence dependence capture and multi-modal feature fusion are solved.

The invention provides a multi-modal neural causal inference micro-service fault positioning method and system, and the method comprises the steps: accessing observability data in a service operation process, and representing the tracking information of each request as a directed acyclic graph of a multi-modal feature; performing multi-modal feature coding and graph self-coding anomaly detection on the calling graph, and identifying an abnormal node through a reconstruction error; based on service topology prior, learning a sparse causal relationship graph between services by adopting a multi-scale neural causal inference method; calculating a node root cause score according to the causal relationship graph and the abnormal score, and executing causal path search to generate a fault propagation path; and marking the potential root cause according to the path weight of the propagation graph and the node popularity, and outputting a visual diagnosis result. According to the method, the system operation state is comprehensively described by fusing three kinds of micro-service system multi-modal data of logs, indexes and Trace in the micro-service system, and the structure-perceived causal diagram is constructed, so that accurate and explainable root cause positioning is realized.

The invention relates to the technical field of intelligent diagnosis, and discloses a micro-grid fault diagnosis method and system based on data driving and unsupervised learning. The method comprises the following steps: collecting current, voltage, temperature and power data of a micro-grid and constructing a time sequence matrix; inputting a time sequence prediction network and a time sequence reconstruction network, and performing parallel processing to obtain a prediction error and a reconstruction error; carrying out weighted fusion on the two errors and constructing a two-dimensional error space to judge normal fluctuation and fault abnormity; and extracting a state variable to generate a dynamic threshold to judge a fault. The false alarm rate and the missing report rate of fault diagnosis are reduced.

The invention belongs to the technical field of brain-computer interfaces, multi-modal feature alignment and information retrieval, and discloses a visual content retrieval method based on electroencephalogram signals. Cross-subject standardized electroencephalogram samples are obtained and input into an electroencephalogram encoder to extract low-dimensional electroencephalogram signal feature vectors; an image encoder is adopted to process the corresponding retrieval images to extract visual feature vectors; the low-dimensional electroencephalogram signal feature vector and the visual feature vector are jointly input into a prototype attention enhancement module to form a dynamic prototype pool, and prototype enhanced electroencephalogram signal representation is obtained through processing; the method comprises the following steps: performing classification training according to existing prototype enhanced electroencephalogram signal representation and image data pairs; for a plurality of time slices of the to-be-queried electroencephalogram signal sample, calculating a reconstruction error or a signal-to-noise ratio of each time slice to obtain a confidence coefficient; and adopting a weighted average or voting mechanism to fuse a plurality of time slice results, and outputting a stable and robust final visual retrieval classification result.

A time sequence anomaly detection method based on PatchTST dual reconstruction consistency constraint comprises the steps that multivariable time sequence data is acquired and preprocessed, the preprocessed data is divided into a plurality of patches, and each patch serves as a token to be input into a Transform encoder to acquire global context features; inputting the output features of the encoder into an MLP solution wharf, performing first reconstruction, performing weighted fusion on the original input and the first reconstruction result to obtain fusion data, performing convolution on the fusion data to obtain a convolution feature map, performing double-branch time sequence modeling to obtain two time sequence feature maps, and splicing the time sequence feature maps to obtain a second time sequence feature map; inputting into a decoder to obtain a final fusion reconstruction result, obtaining an abnormal score based on a secondary reconstruction error and a consistency error, and comparing the abnormal score with a threshold value to carry out time step anomaly detection; through global-to-local feature fusion and dual reconstruction constraints, the robustness and accuracy of multivariable time series data anomaly detection are improved, and the method is effectively suitable for high-noise and multivariable coupled complex time series scenes.

The invention relates to the technical field of data fusion, in particular to a multi-source electrocardiosignal correction method and system based on adaptive fusion, and the method comprises the following steps: constructing a multi-channel input tensor, extracting local features through a weight calculation network, carrying out the adaptive weight fusion and dimension reduction of multiple paths of signals, and carrying out the correction of the multi-source electrocardiosignal. A nonlinear mapping relation is established through a deep reconstruction network, a standard waveform is reconstructed, and network parameters are optimized based on reconstruction error reverse iteration. According to the method, local neighborhood features of multichannel signals are extracted by constructing a weight calculation network, a dynamic channel weight sequence reflecting the real-time contribution degree of a signal source is constructed, the amplitude intensity is adaptively adjusted according to the signal quality, unstable channel noise interference is effectively inhibited, and high-quality signal components are enhanced; a deep reconstruction network is used for carrying out nonlinear feature transformation on a fusion sequence, accurate mapping from non-standard input to standard lead waveforms is established, and weight distribution and optimization of signal reconstruction parameters are achieved in combination with an error back propagation mechanism.

The invention discloses a multi-source data fusion indoor positioning method and system based on deep learning, and relates to the technical field of indoor positioning, and the method comprises the steps: obtaining a wireless signal, inertial sensor data, a timestamp and an indoor map topology; processing the signal through a deep learning model to obtain a preliminary positioning coordinate, and mapping the preliminary positioning coordinate to a map path edge for topological constraint; calculating a displacement and a time interval in combination with the coordinate sequence and the timestamp, fusing the speed and the steering angle of the inertial data, and constructing a spatial-temporal feature vector; calculating a reconstruction error of the vector by using an auto-encoder model to obtain an abnormal confidence coefficient score; if the score exceeds a threshold value, calculating a moving direction according to sensor data, screening a candidate path edge through direction similarity matching, and projecting the coordinate to the edge for correction; otherwise, receiving the original coordinates; and finally outputting the corrected positioning track or the positioning track formed by the original coordinates. The method can ensure that the positioning track better conforms to the actual motion state, and the self-adaptive capability of the positioning system is enhanced.

The invention discloses a sheep abnormal behavior intelligent monitoring method and system based on multi-source data fusion and deep learning, and the method comprises the steps: collecting the multi-source data, such as the motion trail, limb joint angle, body temperature, sound and environmental parameters, of a sheep through multi-type sensing equipment, and carrying out the time-space alignment to form a feature matrix; encoding and reconstructing the feature matrix by using a variational auto-encoder to generate a reconstructed feature matrix, and calculating errors of the two feature matrixes to form an error sequence; and inputting the error sequence into a local abnormal factor analysis model to obtain an abnormal factor sequence, judging an abnormal behavior through a threshold value, and outputting related information. The system comprises a multi-dimensional parameter acquisition unit, a spatial-temporal feature fusion unit, a feature reconstruction unit, an error sequence generation unit, an abnormal factor analysis unit and an abnormal judgment output unit. The method and the system realize multi-source data fusion, improve anomaly identification accuracy, adapt to dynamic changes and are suitable for large-scale breeding.

The invention relates to an equipment state anomaly detection model processing method and an equipment state anomaly detection method. The method comprises the steps of obtaining a training sample set; inputting each training sample into a to-be-trained initial equipment state anomaly detection model to obtain reconstruction data corresponding to each training sample; obtaining a reconstruction error of each training sample according to each training sample and the reconstruction data corresponding to each training sample; according to the reconstruction error of each training sample, adjusting a loss function of a to-be-trained initial equipment state anomaly detection model to obtain a target loss function; according to the target loss function, optimizing model parameters of a to-be-trained initial equipment state anomaly detection model to obtain a target equipment state anomaly detection model; and the target equipment state anomaly detection model is used for identifying whether the equipment state of the to-be-detected equipment is abnormal or not, so that the equipment state anomaly detection precision is improved.

A power transformation equipment temperature anomaly identification and diagnosis method comprises the following steps: collecting state variables, performing cleaning, interpolation complementation and abnormal point elimination on multi-source data through a time synchronization mechanism, and constructing a unified data matrix; extracting statistical features and time sequence dynamic features in the time sequence based on the data matrix, and performing dimensionality reduction on redundant information in combination with a principal component analysis method to form a multi-dimensional fusion feature vector; an unsupervised learning model based on LSTM-AE is constructed, a normal working condition data learning feature reconstruction mode is utilized, and a reconstruction error is taken as a criterion to identify potential temperature anomaly; and calling a preset expert rule base and a knowledge graph, automatically analyzing dominant factors causing anomalies, and identifying typical anomaly types. According to the invention, automatic identification and classification diagnosis of the temperature abnormity of the power transformation equipment under an unsupervised condition are realized, the accuracy and response speed of fault identification are obviously improved, and the intelligence and practicability of equipment operation state monitoring are enhanced.

The invention relates to a multi-protocol compatible ultrasonic radar test method and system. According to the method, protocol configuration is automatically obtained by scanning a radar identification code, communication connection is established, a standardized test process is controlled and executed based on a parameter file to collect response data, and a reconstruction error of the test data is calculated in real time by using an LSTM auto-encoder model to realize anomaly detection and feature extraction. According to the method, a fault source is accurately positioned by combining causal reasoning of a fault knowledge graph, and a diagnosis report containing a solution is finally generated, so that full-process automatic closed loop from protocol adaptive configuration, intelligent anomaly recognition to root cause diagnosis is realized, the efficiency of multi-protocol radar testing and the accuracy of fault diagnosis are remarkably improved, and the fault diagnosis efficiency is improved. The technical defects that a traditional method depends on artificial experience, efficiency is low, and diagnosis results are one-sided are effectively overcome.

The invention relates to the technical field of data processing, in particular to a covariance inverse matrix recursive updating method for on-satellite hyperspectral anomaly detection, which comprises the following steps: acquiring a hyperspectral data stream; carrying out adaptive dimension reduction processing on the data stream and loading an initial background model; extracting local background statistics by adopting a sliding window mechanism, generating a background spectrum dictionary by utilizing online dictionary learning, calculating a reconstruction error between a current pixel and the dictionary and a Mahalanobis distance between the current pixel and a background model, and fusing to generate an abnormal score; performing abnormal confidence coefficient evaluation by combining the spatial context information and the spectral angle matching degree, and updating a spectral mean vector by using an exponential weighted moving average algorithm based on non-abnormal pixel data; and the updated model is injected into the next round of processing to form a recursive chain. According to the method, the high calculation complexity of direct inversion of a covariance matrix is avoided, real-time anomaly detection on a satellite is realized, the downloading amount of original data is remarkably reduced, and the congestion of a satellite-ground communication link is relieved.

The invention relates to the technical field of computer-aided modeling, in particular to a heating and ventilation equipment energy efficiency optimization method and system based on digital twinning, and the method comprises the steps: obtaining a real-time operation data set, and constructing a mechanism model; obtaining a model prediction data set by using the mechanism model, and training a data reconstruction model; inputting the real-time operation data set into the data reconstruction model to obtain a reconstruction error, and taking the reconstruction error as a state deviation data set; dynamically correcting an adjustable model parameter of the mechanism model based on the state deviation data set, and generating a dynamic calibration digital twinborn model; and finally, executing a multi-target collaborative optimization algorithm based on the dynamic calibration digital twin model, and generating an energy efficiency optimization control instruction. According to the method, the state deviation caused by the performance degradation of the physical equipment is quantified by utilizing the data reconstruction model, and the mechanism model is dynamically corrected by taking the state deviation as feedback, so that the problem that a static model is disjointed from a physical entity state is solved, and high-precision dynamic simulation and energy efficiency collaborative optimization are realized.

The invention discloses a power load time sequence anomaly detection method and system based on a Mama and LSTM hybrid network, and belongs to the technical field of power system data analysis and artificial intelligence, and the method comprises the steps: inputting the preprocessed power load and related factor time sequence data into a Mama-LSTM hybrid encoder; performing time sequence data reconstruction and anomaly probability prediction in parallel by using depth features output by an encoder, and performing model training by jointly optimizing reconstruction error loss and anomaly detection loss; calculating a comprehensive abnormal score based on the trained model, and judging an abnormal point by adopting a dynamic threshold value; and outputting an anomaly detection result and providing an analysis report containing an unsupervised evaluation index and multi-dimensional visualization. Through deep series fusion of Mama and LSTM, long-term dependence and complex modes in a power load sequence are effectively captured, and the accuracy, robustness and interpretability of anomaly detection are significantly improved in combination with a joint training strategy and an unsupervised evaluation system of the system.

The invention discloses an electric energy metering error evaluation method and system for a direct current charging pile, equipment and a medium, and the method comprises the steps: constructing an electric energy disturbance response function based on transient voltage and transient current, and effectively quantifying the transient metering deviation caused by measurement lag, voltage drop, current peak and other factors in the actual operation of the direct current charging pile. Error reconstruction modeling has higher dynamic adaptive capacity and physical rationality, the limitation that in the prior art, the influence of the high-variable-load working condition on metering cannot be reflected is broken through, and the accuracy and timeliness of direct-current charging pile error monitoring are greatly improved. Besides, a standard error is introduced to carry out fitting calibration on a reconstruction error and carry out energy conservation verification, thereby realizing verifiability on a physical level and controllability in engineering implementation, enhancing credibility and authority of an evaluation result, constructing a closed-loop analysis mechanism from error evaluation to metering correction and then to conservation verification, and improving reliability and reliability of the evaluation result. And the closeness and the adaptive capability of the overall error evaluation are improved.

The invention provides a channel state information feedback enhancement method and system based on bidirectional channel correlation, and the method is executed at a base station end, and comprises the steps: receiving a downlink channel state information semantic vector which is coded, compressed and quantized by a user end through a deep joint source channel; carrying out inverse quantization on the semantic vector to obtain an intermediate semantic vector; based on the intermediate semantic vector, retrieving a related historical uplink semantic vector from an uplink channel semantic knowledge base; performing fusion enhancement on the intermediate semantic vector and the retrieval result through an attention mechanism to generate an enhanced semantic vector; inputting the enhanced semantic vector into a decoder, and outputting a reconstructed downlink channel state information matrix; wherein the encoder, the enhancement module and the decoder perform end-to-end joint training optimization by minimizing a reconstruction error. According to the method, the semantics is enhanced at the receiving end by utilizing the correlation of the uplink and downlink channels, the feedback precision and the system robustness are remarkably improved, and the method is suitable for a frequency division duplex system.

The invention discloses an image vectorization method and device based on a neural diffusion curve. According to the method, geometric control point parameters of multiple sections of Bezier curves are predicted from end to end in an input grating image through a deep neural network, and the main contour and color structure of the image are accurately represented in a Bezier curve form; color control points on the two sides of the curve are automatically extracted through a micro color sampling module, and physically consistent diffusion curve representation is constructed; and performing neural approximation on the resolving process of the diffusion equation by adopting a rendering network based on a Fourier neural operator to realize end-to-end micro-reconstruction of the image generated by the prediction curve, thereby reversely optimizing the geometric prediction of the curve by using a reconstruction error. According to the method, manual intervention is not needed, image vector representation which is high in fidelity, infinite in scaling, lossless in editing and capable of supporting natural layering can be automatically generated while physical rendering consistency is kept, precision and efficiency are improved, and the method is suitable for the fields of image vectorization, digital art creation, micro rendering, graphic design and the like.

The invention discloses an electric power field operation safety monitoring method and system based on AI driving. The method comprises the following steps: synchronously acquiring voice, operation video and equipment state identification information, and de-noising to generate a multi-modal sequence; using a multi-modal cross reconstruction model to reconstruct other modals for any modal, and identifying suspicious fragments according to reconstruction errors; multi-layer consistency matching is executed on the suspicious fragments, and the consistency matching degree and mismatch attribution are calculated through time soft alignment, semantic embedding unification and semantic graph reasoning; dispersing the voice, action and equipment state elements into an event sequence, and matching the event sequence with a preset process causal graph to position an abnormal event; and finally, abnormal nodes and induced nodes are analyzed in combination with a Bayesian algorithm, and whole-process safety monitoring is realized. According to the method, password inconsistency, action out-of-order, object mismatching and equipment response abnormity in electric power operation can be intelligently identified and traced, and the real-time performance and accuracy of operation safety are improved.

The invention provides a multi-domain power grid data collaborative modeling method and system based on a tensor game diagram, and the method comprises the following steps: firstly constructing a six-dimensional enhanced tensor model, constructing a six-order tensor based on the number of nodes, timestamps and other six dimensions, and obtaining a kernel tensor and a factor matrix through CPD-Tucker mixed decomposition; a joint optimization objective function containing reconstruction errors, game equilibrium and privacy risks is constructed, and an optimization kernel tensor and factor matrix is solved; and finally, inputting an optimization result into the MGGCN, processing double targets through a leader branch (a physical topology adjacency matrix guarantees reliability) and a follower branch (a market transaction incidence matrix optimizes economic cost), and generating a collaborative decision result through multi-head attention fusion. According to the method, the problems of low multi-source heterogeneous data fusion efficiency and difficulty in considering cross-domain collaborative privacy security and dynamic optimization are solved, the new energy output prediction error can be reduced, and the data utility loss caused by global encryption is reduced.

The invention relates to a charging pile automatic detection operation and maintenance method and system based on the Internet of Things and a medium, and belongs to the technical field of intelligent power grid operation and maintenance. The automatic detection operation and maintenance method comprises the steps that multi-source sensor data of each charging pile is collected in real time, and a time-space aligned multi-source data set is generated; performing wavelet packet decomposition processing on the multi-source data set through an edge calculation node, generating a compressed feature vector, inputting the compressed feature vector into a pre-trained residual self-encoder model, calculating a reconstruction error, performing association reasoning in combination with a historical fault knowledge graph, and outputting a fault mode and a confidence coefficient set; and an operation and maintenance strategy instruction is dynamically generated in combination with the equipment health index attenuation rate, the charging pile to be operated and maintained is positioned, the digital twin is called for residual life prediction, a maintenance work order is generated through a path optimization algorithm, and an execution terminal is driven to complete operation and maintenance operation. The reliability and stability of the charging pile equipment can be improved, and the dual requirements of a modern charging system for high availability and low operation and maintenance cost are effectively met.

The embodiment of the invention provides an anomaly detection method and device, electronic equipment, a storage medium and a computer program product, and is at least applied to the field of artificial intelligence, and the method comprises the steps: carrying out the standardization processing of an index data sequence of a to-be-detected object in a preset time window, and obtaining a standardized index sequence; performing data reconstruction on the standardized index sequence to obtain reconstructed index data; determining a reconstruction error of the to-be-detected object in a preset time window based on the index data sequence and the reconstruction index data; and performing anomaly detection on the to-be-detected object based on the reconstruction error. According to the invention, the detection efficiency of the anomaly detection process can be improved, and the detection quality of the anomaly detection process is ensured.

The invention provides a power distribution communication network optical cable fault prediction method and system based on machine learning. The method comprises the following steps: collecting Rayleigh scattering signals of a plurality of monitoring points along an optical cable; performing unsupervised learning on the optical cable adaptive reconstruction model, and calculating a reconstruction error matrix and an abnormal cumulative metric; constructing an optical cable network diagram, calculating a fault propagation probability matrix, predicting a future optical cable health index, and obtaining a future health index matrix; establishing an environment influence matrix, and adopting an optical cable environment adaptive filtering model; and based on the health index and the fault confidence coefficient matrix after filtering optimization, optical cable maintenance priority scores are calculated and sorted, and a self-adaptive inspection and maintenance strategy is formulated. According to the invention, high-precision monitoring, fault propagation prediction and environmental adaptability optimization of the health state of the optical cable are realized, and the intelligent level of operation and maintenance of the optical cable is improved.

According to the wafer defect detection system and method based on polar coordinate transformation and the generative adversarial network, polar coordinate expansion, geometric position coding, a double-discriminator structure and double-space consistency reconstruction loss are introduced, so that the network can accurately model a circular geometric structure of a wafer while keeping pixel details; therefore, the significance of the defect in the reconstruction error is enhanced. According to the method, the wafer image is subjected to structural constraint in the Cartesian space and the polar coordinate space at the same time, the sensitivity of the model to annular defects, edge defects and radial anomalies is improved, the defects of a traditional method in the aspects of structural consistency, edge reconstruction and weak defect detectability are overcome, and higher accuracy and engineering deployability are achieved.

The invention belongs to the technical field of pavement detection, and discloses a highway pavement roadbed detection method and system based on laser three-dimensional scanning. By establishing an analytic optical or radar scattering physical model, a distance measurement deviation initial value conforming to a multi-path reflection physical mechanism is provided for each laser measurement point, and then refined correction is performed on the initial value through a residual error regression network fusing local point cloud context information, so that advantage complementation of physical prior and data driving is realized, and the accuracy of distance measurement is improved. And finally, in combination with adaptive filtering based on a confidence map, ranging errors caused by a multi-path effect can be effectively distinguished and corrected in the presence of complex road surface conditions such as slippery road surface, greasy dirt or thin water film, and meanwhile, geometric details such as real road surface pits, textures and the like are completely reserved. Therefore, the reconstruction error and the disease false alarm rate of the pavement digital elevation model are obviously reduced, and the precision and the reliability of highway pavement roadbed detection are improved.

A machine abnormal sound detection method based on feature enhanced dynamic graph convolution belongs to the field of machine abnormal sound detection, and comprises the following steps: firstly, generating a time-frequency spectrum feature representation for an original audio signal of a machine through a feature extractor; the method comprises the following steps: extracting multi-scale features by using an FE module through 1D-FFT, multi-scale trend period decomposition and a Token-Transform structure to obtain an enhanced feature spectrogram; sDA-GCN and DCA-GCN network structures are used, shared features and dynamic differences between devices are mined based on an enhanced feature spectrum, and feature differences under domain offset are reduced; through a coarse-grained label classifier, a fine-grained label classifier and a domain classifier, in combination with GRL and CORAL losses, fine-grained alignment of the feature space is realized; a three-stage training strategy is used, and the model is optimized through perception consistency pre-training, unsupervised contrast classification learning and reconstruction error-based anomaly detection. According to the invention, high-performance anomaly detection is realized.

The invention discloses a three-dimensional Gaussian splash reconstruction method for an underwater scene, and belongs to the technical field of computer vision and three-dimensional reconstruction. Comprising the following steps: acquiring a monocular video frame sequence of a target underwater scene, a corresponding camera pose sequence, an initial sparse point cloud, an initial three-dimensional Gaussian point set and learnable physical parameters of an underwater imaging model; in the training process, performing weighted evaluation on a reconstruction error based on a multi-view consistency mechanism of opacity weighting, calculating an importance score of each Gaussian point, and performing densification operation on a three-dimensional Gaussian point set; adopting a staged freezing strategy to cooperatively optimize the three-dimensional Gaussian point set and underwater imaging model parameters; and performing rendering and underwater image synthesis on any new view angle camera pose based on the optimized three-dimensional Gaussian point set and underwater imaging model parameters, and outputting a new view angle synthesized image to represent a reconstruction result. According to the method, the geometric compactness, the visual fidelity and the physical interpretability of an underwater three-dimensional reconstruction result are improved.

The invention provides a KPIs anomaly detection method based on MVMD decomposition. The KPIs anomaly detection method comprises the following steps: S1, obtaining multi-dimensional key performance index KPIs time sequence data in a micro-service system; s2, decomposing the normalized multivariable KPIs time sequence data into K intrinsic mode function (IMF) components; s3, dividing the time sequence of each IMF component into a plurality of fixed-length subsequences through a sliding window; s4, respectively constructing and training a variational auto-encoder VAE model, and calculating a reconstruction error of each IMF component at each time point; s5, obtaining a comprehensive abnormal score of each time point; s6, based on a grid search method, determining an optimal anomaly score threshold value for anomaly judgment; and S7, comparing the comprehensive abnormal score with an optimal abnormal score threshold value, and if the comprehensive abnormal score at a certain time point exceeds the threshold value, judging that the KPIs at the time point is abnormal. According to the invention, the modeling capability of abnormal modes of different frequency components can be enhanced, so that the accuracy of overall anomaly detection is improved.

The invention is suitable for the technical field of load detection, and provides a self-supervised learning charging abnormal load detection method based on a large model, and the method comprises the steps: collecting charging load time sequence data, environment parameters, and user behavior data; constructing a mask reconstruction task training model, a time sequence prediction task training model and a contrast learning task training model based on the preprocessed data; calculating the correlation of the extracted time domain features, frequency domain features and environment correlation features through a preset time sequence, an environment collaborative attention head and a frequency domain attention head, distributing feature weights according to preset power system exception categories, and outputting fused feature representation; generating a comprehensive anomaly score based on the reconstruction error, the embedding deviation and the output probability density; when the comprehensive anomaly score exceeds a preset threshold value, the anomaly type is judged in combination with the multi-dimensional feature conditions of the harmonic distortion rate, the temperature change rate and the phase deviation; and the model parameters are updated to adaptively generate a detection result, so that the false alarm rate is reduced and the operation and maintenance efficiency is improved.

The invention provides a new energy station abnormal data detection method and system, computer equipment and a medium, and belongs to the technical field of power system operation state monitoring and intelligent operation and maintenance, and the method comprises the steps: obtaining environment meteorological data, equipment operation state data and historical power output data; performing time alignment on the environmental meteorological data, the equipment operation state data and the historical power output data to obtain time sequence data; extracting a time pattern feature from the time series data; modeling a time relation and a sensor relation for the time mode characteristics based on a graph attention mechanism to obtain time sequence characteristics and sensor relation characteristics; fusion features are obtained through fusion; reconstructing according to the fusion features to obtain reconstructed data; and determining an abnormal data detection result of the to-be-evaluated new energy station according to the reconstruction data and reconstruction errors of the corresponding equipment operation state data and environment meteorological data. According to the method, the linkage abnormal mode between the sensors is captured, the limitation of single-point modeling is avoided, and the abnormal data detection accuracy is improved.

The invention discloses a complex fund chain intelligent management and anomaly detection method based on a deep learning model, and the method comprises the steps: capturing a fund flow network structure between accounts through building a fund topological relation, extracting the periodicity and trend characteristics of a transaction sequence through combining a time sequence neural network, and calculating a reconstruction error through an anomaly detection model, thereby achieving the intelligent management of a fund chain. Quantifying transaction deviations and identifying anomalies; if the abnormal score exceeds a threshold value, backtracking a risk conduction chain through a path tracking method, revealing an abnormal fund propagation path, simulating future fund transfer scale and direction by using a flow prediction model, comparing with a historical risk threshold value, judging potential fluctuation, and generating a real-time early warning signal; according to the method, full-chain analysis from anomaly detection to risk prediction is realized through multi-model fusion, the accuracy and timeliness of risk identification in financial transactions are remarkably improved, and effective technical support is provided for prevention of systematic financial risks.