A mobile substation-based device operation state visual analysis system

Abstract

The present application relates to the technical field of visual analysis, in particular to a device operation state visual analysis system based on mobile substation. The system comprises a data acquisition and transmission unit, which collects device operation data from various devices of the mobile substation in real time, and transmits the device operation data to an edge computing and analysis unit; the edge computing and analysis unit extracts key features of the device operation data based on a topology-dynamic coupling feature extraction technology, identifies abnormal states and generates preliminary alarms; a state modeling and diagnosis unit constructs a trajectory-aware motion adaptive health degree evaluation model based on the key features, and diagnoses the device state. The present application designs and constructs a three-dimensional health degree evaluation model including electrical, mechanical and trajectory, combines spatial topological relationship and real-time motion information, and provides a stereoscopic perception display of device position-electrical state-motion influence.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual analysis, in particular to a device operation state visual analysis system based on a mobile substation. BACKGROUND

[0002] In modern power systems, mobile substations are widely used as a flexible solution in temporary power supply, emergency repair and power grid expansion scenarios. Compared with traditional fixed substations, mobile substations can be moved in short distances (a few meters to tens of meters) without interrupting power supply services to adapt to different operational requirements or environmental changes. This flexibility requires that the devices in the mobile substation must have high portability and rapid deployment capabilities, while ensuring the stability of electrical connections and mechanical structures is not affected.

[0003] However, during actual operation, mobile substations face several key challenges. First, since the devices in the substation are moved in the running state, this will increase the physical stress on the electrical connection points, which may cause poor contact or looseness, posing a potential threat to the safe and stable operation of the devices. Second, the vibration and displacement generated during movement not only affect the measurement accuracy of electrical parameters, but also may cause sensor readings to deviate, even misreport abnormal conditions. In addition, when multiple devices in the mobile substation work in parallel, the electrical interaction between them will change due to the change in position, which puts higher requirements on the real-time response speed and accuracy of the monitoring system. Therefore, a device operation state visual analysis system based on a mobile substation is designed. SUMMARY

[0004] The purpose of the present application is to provide a device operation state visual analysis system based on a mobile substation to solve the problems of electrical parameter abnormality misjudgment, inaccurate device state monitoring and difficulty in multi-device collaborative visualization caused by factors such as vibration and displacement when the mobile substation equipment is moved in the uninterrupted power supply situation.

[0005] To achieve the above purpose, the present application provides a device operation state visual analysis system based on a mobile substation, which comprises:

[0006] A data acquisition and transmission unit, which collects device operation data in real time from various devices in the mobile substation and transmits the device operation data to the edge computing and analysis unit;

[0007] An edge computing and analysis unit, which extracts key features of the device operation data based on a topology-dynamic coupling feature extraction technology, identifies abnormal states and generates preliminary alarms;

[0008] a state modeling and diagnosis unit, which constructs a trajectory-aware motion adaptive health degree evaluation model based on key features and diagnoses the device state;

[0009] a visual display unit, which displays the device operation state data and diagnosis results in a graphical interface form on a visual terminal.

[0010] As a further improvement of the technical solution, the data acquisition and transmission unit includes a data acquisition module and a data transmission module.

[0011] The data acquisition module collects device operation data in real time from various devices of the mobile substation and pre-processes the device operation data.

[0012] The data transmission module transmits the pre-processed device operation data to the edge computing and analysis unit using wireless communication.

[0013] As a further improvement of the technical solution, the edge computing and analysis unit identifies abnormal states and generates preliminary alarms, including the following steps:

[0014] S1.1, the received device operation data is divided into a sliding window of fixed length n according to the time axis;

[0015] S1.2, the original device operation data in the sliding window is processed, and key feature indicators are extracted based on topology-dynamic coupling feature extraction technology.

[0016] S1.3, based on the threshold comparison rule, it is judged whether the device operation data is abnormal, if the device operation data is abnormal, a preliminary alarm information is generated, and the abnormal operation data is transmitted to the state modeling and diagnosis unit.

[0017] As a further improvement of the technical solution, in S1.2, the original device operation data in the sliding window is processed, and key feature indicators are extracted based on topology-dynamic coupling feature extraction technology, including the following steps:

[0018] S1.21, the voltage and current of each measuring point in the sliding window are statistically analyzed and processed in frequency domain, and electrical characteristics are extracted.

[0019] S1.22, according to the device spatial position data collected in the sliding window, a spatial topology relationship diagram between devices is dynamically established, which is updated once in each sliding window and used to dynamically reflect the device layout structure.

[0020] S1.23, the rotation angle and displacement of the device are extracted from the IMU data in the sliding window, and a two-dimensional motion compensation matrix is constructed.

[0021] S1.24, based on the spatial topology relationship graph and the electrical state of the device, calculate the electrical-space distance ratio of each pair of adjacent devices to reflect whether the electrical difference between the devices changes with the spatial layout;

[0022] S1.25, difference analysis is performed on the voltage and current signals after applying the motion compensation matrix and the original uncompensated signals;

[0023] S1.26, for each device node, calculate its importance score in the topology graph, and filter the device nodes based on the importance score;

[0024] S1.27, the key features of each device in the current sliding window are uniformly organized into a four-dimensional feature representation matrix.

[0025] As a further improvement of the technical solution, the state modeling and diagnosis unit includes a state model establishment module and a state diagnosis module;

[0026] The state model establishment module constructs a device health degree model based on the historical operation data of the device and the key features of the device;

[0027] The state diagnosis module performs anomaly diagnosis on the current collected features and the established state model.

[0028] As a further improvement of the technical solution, the state model establishment module constructs a device health degree model based on the historical operation data of the device and the key features of the device, including the following steps:

[0029] S2.1, receiving the four-dimensional feature representation matrix transmitted by the edge computing and analysis unit;

[0030] S2.2, classifying and labeling the feature data in the four-dimensional feature representation matrix;

[0031] S2.3, using feature importance analysis method to quantify the influence degree of each dimension feature in different device state categories, outputting the feature weight distribution corresponding to each device state;

[0032] S2.4, based on the labeled feature data, using motion decoupling spatio-temporal Gaussian mixture model to probabilistically model the device state;

[0033] S2.5, analyze the natural evolution path of the key features in the time dimension, and construct the state transition probability combined with the sliding window sequence data;

[0034] S2.6, based on steps S2.4 and S2.5, construct a trajectory-aware motion adaptive health degree evaluation model.

[0035] As a further improvement of the technical solution, in S2.4, a motion decoupled spatio-temporal Gaussian mixture model is used to probabilistically model the device state, including the following steps:

[0036] S2.41, define a motion intensity factor According to the motion intensity factor, the feature data is dynamically divided into three categories of motion scenes: static state group, regular motion group, and intense motion group;

[0037] S2.42, analyze the acceleration vector of the device in the sliding window, construct a motion compensation projection matrix for different motion interference components, decouple and project the four-dimensional feature representation matrix to obtain the disturbance compensated feature representation;

[0038] S2.43, for each type of motion scene, the compensated feature representation is respectively Gaussian mixture modeled, and three independent state model subsets are established: static model, regular motion model and intense motion model;

[0039] S2.44, calculate the motion intensity of the current window, and construct a scene awareness weight vector based on the motion intensity;

[0040] S2.45, input the disturbance compensated feature representation into the static model, the regular motion model and the intense motion model respectively, calculate the state probability under each model, combine according to the fusion weight, and generate the final state probability.

[0041] As a further improvement of the technical solution, in S2.6, a trajectory-aware motion adaptive health degree evaluation model is constructed, including the following steps:

[0042] S2.61, select the corresponding health state reference vector group from the preset state space reference library based on the motion intensity data;

[0043] S2.62, difference operation is performed on the feature vectors of the current sliding window and the previous two sliding windows to construct a state trajectory second-order difference detection function;

[0044] S2.63, calculate the electrical health degree, mechanical health degree and trajectory health degree based on the state space reference library;

[0045] S2.64, dynamically adjust the weights of each health component in S2.63 according to the motion intensity, and calculate the fusion health degree;

[0046] S2.65, integrate the electrical health degree, mechanical health degree and trajectory health degree into a three-dimensional health vector output, and output the fusion health degree score.

[0047] As a further improvement of the technical solution, the state diagnosis module judges the equipment operation state according to the three-dimensional health vector and the fusion health degree score; wherein, the fusion health degree score is used to determine the overall health level of the equipment, and the components of the three-dimensional health vector are introduced for component analysis for tracing the abnormal reasons of the equipment.

[0048] As a further improvement of the technical solution, the visual display unit includes an equipment state display module, a topological space correlation display module, a motion-electric coupling dynamic mapping module, and a health degree score display module.

[0049] The equipment state display module displays the equipment operation data and the equipment state results based on the data acquisition module and the edge computing and analysis unit.

[0050] The topological space correlation display module represents the spatial topological relationship graph constructed based on the edge computing and analysis unit, and updates the equipment layout structure in real time.

[0051] The motion-electric coupling dynamic mapping module displays the abnormality contrast based on the state probability of the state model establishment module.

[0052] The health degree score display module displays the health degree of the equipment based on the diagnosis results of the state diagnosis module.

[0053] Compared with the prior art, the present application has the following advantages:

[0054] 1. In the equipment operation state visual analysis system based on the mobile substation, the topological-dynamic coupling feature extraction technology and the motion compensation mechanism are introduced, the system can effectively distinguish the real electrical fault of the equipment from the pseudo-abnormality caused by the vibration and displacement interference caused by short-distance movement, significantly reduce the false alarm rate, and ensure the accuracy of the equipment operation state monitoring during the movement.

[0055] 2. In the equipment operation state visual analysis system based on the mobile substation, a three-dimensional health degree evaluation model including electrical, mechanical and trajectory is constructed, combined with the spatial topological relationship and real-time motion information, a stereoscopic perception display of equipment position-electrical state-motion influence is provided, which helps the operation and maintenance personnel to quickly locate the abnormal reasons and make scientific decisions, and improves the intelligent and visual level of the operation and maintenance of the mobile substation. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The overall flowchart of the present application is shown in the figure.

[0057] The meanings of the various numbers in the figure are as follows:

[0058] 1, data acquisition and transmission unit; 11, data acquisition module; 12, data transmission module; 2, edge computing and analysis unit; 3, state modeling and diagnosis unit; 31, state model establishment module; 32, state diagnosis module; 4, visualization display unit; 41, device state display module; 42, topological space correlation display module; 43, motion-electric coupling dynamic mapping module; 44, health score display module. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0060] Embodiment: Please refer to Figure 1 As shown, a mobile substation-based device operating state visual analysis system is provided, comprising:

[0061] The data acquisition and transmission unit 1 collects device operating data (including current, voltage) from various devices of the mobile substation in real time, and transmits the device operating data to the edge computing and analysis unit 2;

[0062] In this embodiment, the data acquisition and transmission unit 1 includes a data acquisition module 11 and a data transmission module 12;

[0063] The data acquisition module 11 collects device operating data from various devices of the mobile substation in real time, and pre-processes the device operating data (including uniform collection, time synchronization, format standardization, and noise filtering, missing value filling, and abnormal value identification for preprocessing to ensure the data quality for subsequent processing);

[0064] The data transmission module 12 transmits the pre-processed device operating data to the edge computing and analysis unit 2 using wireless communication.

[0065] The edge computing and analysis unit 2 extracts key features of the device operating data based on topological-dynamic coupling feature extraction technology, identifies abnormal states and generates preliminary alarms;

[0066] In this embodiment, the edge computing and analysis unit 2 identifies abnormal states and generates preliminary alarms, including the following steps:

[0067] S1.1, split the received device operation data into a sliding window of fixed length n (10 seconds or 1 minute) according to the time axis, for data buffering and time alignment, to ensure the time continuity and integrity of the data, and to adapt to sensors with different sampling frequencies;

[0068] S1.2, process the original device operation data in the sliding window, and extract key feature indicators based on the topology-dynamic coupling feature extraction technology;

[0069] Among them, the core problem of mobile substation display scene: multi-device parallel monitoring leads to crowded visualization interface elements; data noise caused by device vibration / displacement is misidentified as electrical abnormality; electrical parameters and mechanical state (position / posture) are displayed separately; traditional feature extraction process cannot adapt to the real-time requirements of vehicle-mounted mobile scene;

[0070] The topology-dynamic coupling feature extraction technology dynamically couples the spatial topology relationship of the device with the motion state, constructs a four-dimensional feature space (time + electrical quantity + spatial position + motion vector), realizes the device cluster collaborative state visualization in the mobile substation scene through spatial topology coding and motion dynamics compensation, provides the device position-electrical state-motion influence stereoscopic perception ability for the operation and maintenance personnel, and completely solves the information fragmentation problem of traditional two-dimensional plane display;

[0071] Process the original device operation data in the sliding window, and extract key feature indicators based on the topology-dynamic coupling feature extraction technology, including the following steps:

[0072] S1.21, statistically analyze and process the voltage and current of each measuring point in the sliding window (perform basic statistical analysis on time domain signals (such as voltage and current sequence) in a sliding time window) and in the frequency domain, extract electrical features, including mean, extreme value, standard deviation, and change rate, and the extracted electrical features are used to construct the graph node attributes subsequently;

[0073] S1.22, dynamically establish the spatial topology relationship graph between devices according to the device spatial position data collected in the sliding window, and update the spatial topology relationship graph once every sliding window, to dynamically reflect the device layout structure;

[0074] Each device node in the spatial topology relationship graph is represented as , wherein, represents the current electrical feature of the device, represents the center position of the device in the sliding window;

[0075] The edge represents the spatial distance relationship between devices: , wherein, , device index, time, Euclidean distance;

[0076] S1.23, extract the rotation angle and displacement of the device from the IMU data in the sliding window (IMU data refers to the motion information of the device collected by the inertial measurement unit, including acceleration, angular velocity and attitude change, etc.), and construct a two-dimensional motion compensation matrix (the device on the mobile substation may be subject to physical disturbances such as shaking, turning, ramps, undulations, etc. during operation. These disturbances will cause cable stress, electrical connection disturbance, and interference with high-frequency sampled electrical signals, resulting in false current / voltage fluctuations. Therefore, the motion compensation matrix is constructed to correct the measured electrical fluctuations and determine whether the change is a physical anomaly or a motion disturbance);

[0077] The motion compensation matrix is: ;

[0078] The motion compensation matrix is used to eliminate the relative motion component of the device from the measured electrical changes and identify the true electrical fluctuations;

[0079] S1.24, based on the spatial topology relationship graph and the electrical state of the device, calculate the electrical-spatial distance ratio (ESR) of each pair of adjacent devices to reflect whether the electrical difference between the devices changes reasonably with the spatial layout (if the distance between the devices is very close but the electrical state difference is large, or the ESR changes dramatically during the motion process, there may be an anomaly);

[0080] Electrical-spatial distance ratio The calculation formula is: ;

[0081] S1.25, difference analysis is performed on the voltage and current signals after applying the motion compensation matrix and the original uncompensated signals (difference analysis specifically: first, calculate the residual sequence between the two, and extract its mean, standard deviation, peak deviation, skewness and kurtosis, etc. statistical characteristics, to determine whether the signal change has abnormal characteristics; second, analyze the differences in energy domain and frequency domain between the two signals to assess whether there are still significant spectral abnormalities or energy mutations after compensation; finally, when the residual standard deviation is less than the threshold d and the skewness feature is not significant, it can be determined as a pseudo-anomaly caused by motion disturbance, otherwise it is considered as a real electrical fault), analyze whether the fluctuations are mainly caused by real device anomalies or movement noise, if there are still significant fluctuations after compensation, it means a real physical anomaly, if the fluctuations are greatly weakened after compensation, it means a pseudo-anomaly caused by motion, this step improves the explainability of the system to the cause of the anomaly;

[0082] S1.26, for each device node, calculate its importance score in the topology graph, and filter the device nodes based on the importance score (keep the device nodes with high topology importance, select the features with ESR abnormal values, large real fluctuations, and obvious spectrum distortion);

[0083] Importance score is: ;

[0084] wherein, represents the eigenvector centrality based on the centrality (the eigenvector centrality of a node indicates that the importance of the node depends on the centrality scores of its neighbor nodes, the more important nodes it connects, the higher its score will be), represents the edge betweenness centrality of the node (the edge betweenness centrality of an edge indicates the frequency of its appearance in the shortest paths between all node pairs, the more shortest paths the edge appears on, the higher its score will be);

[0085] S1.27, organize the key features of each device in the current sliding window into a four-dimensional feature representation matrix ;

[0086] ;

[0087] wherein, is a feature fusion operator, which represents the integration of time, electrical state, spatial position and motion information into a unified multi-dimensional feature vector or feature tensor according to rules (feature splicing).

[0088] S1.3, determine whether the device operation data is abnormal based on the threshold comparison rule (the threshold comparison rule is to compare the real-time calculated feature value with the static threshold pre-configured on the edge device), if the device operation data is abnormal, generate a preliminary alarm information (the alarm information includes alarm type, alarm level, occurrence time, associated device, etc.), and transmit the abnormal operation data to the state modeling and diagnosis unit 3.

[0089] The state modeling and diagnosis unit 3 constructs a trajectory-aware motion adaptive health degree evaluation model based on the key features, and diagnoses the device state;

[0090] In this embodiment, the state modeling and diagnosis unit 3 includes a state model establishment module 31 and a state diagnosis module 32;

[0091] The state model establishment module 31 constructs a device health degree model based on the historical operation data of the device and the key features of the device;

[0092] The state model establishment module 31 constructs a device health degree model based on the historical operation data of the device and the key features of the device, including the following steps:

[0093] S2.1, receiving the four-dimensional feature representation matrix transmitted by the edge computing and analysis unit 2;

[0094] S2.2, classifying and labeling the feature data in the four-dimensional feature representation matrix, and identifying whether it belongs to normal operation, slight abnormality or fault state;

[0095] S2.3, using a feature importance analysis method (SHAP value interpretation model) to quantify the influence degree of each dimension feature in different device state categories, and outputting the feature weight distribution corresponding to each device state, which is used to evaluate the contribution of electrical features, spatial features and motion features in diagnosis;

[0096] S2.4, based on the labeled feature data, using a motion decoupled spatio-temporal Gaussian mixture model to probabilistically model the device state;

[0097] Further, the motion decoupled spatio-temporal Gaussian mixture model is used to probabilistically model the device state, which is aimed at the three core problems faced in the mobile substation operating environment: false positives caused by electrical-mechanical coupling interference caused by device motion, unstable diagnosis caused by fuzzy state boundary, and insufficient real-time performance of traditional modeling methods in complex motion scenarios. The model realizes the decoupling of electrical and motion features by constructing a motion compensation matrix, introduces a motion intensity driven scene modeling mechanism (stationary / normal / severe), and uses parallel computing and entropy fusion algorithm to greatly improve the diagnosis efficiency and stability. Compared with traditional methods, the model has five advantages: significantly reducing false positive rate, enhancing the sensitivity of the model to slight defects, realizing adaptive modeling accuracy in multiple scenes, compressing the diagnosis delay to milliseconds, and providing clear and intuitive visual output. The core purpose is to build a health assessment system suitable for mobile scenarios, realize continuous modeling and residual life prediction of device state evolution, support early warning and strategic intervention in intelligent operation and maintenance, and ultimately improve the safety, economy and self-automation level of substation operation and maintenance;

[0098] The motion decoupled spatio-temporal Gaussian mixture model for probabilistically modeling the device state includes the following steps:

[0099] S2.41, define the motion intensity factor According to the motion intensity factor, the feature data is dynamically divided into three types of motion scenes: a static state group (the device is in a stable state, and the displacement / rotation angle change is less than a, where a represents an empirical threshold of motion intensity, used to distinguish whether the device is in a static state, and a is calibrated through experiments according to the actual operation scene (such as short-distance moving characteristics) of the mobile substation and the sensor accuracy (such as IMU measurement error)), a regular motion group (the device is slightly shaken and slowly displaced), and a severe motion group (the device is in a turning, slope switching, strong jolt, and other disturbance environment);

[0100] S2.42, analyze the acceleration vector of the device in the sliding window, and construct a motion compensation projection matrix for different motion disturbance components Decouple and project the four-dimensional feature representation matrix to obtain a feature representation after disturbance compensation ;

[0101] ;

[0102] ;

[0103] In the formula, is linear acceleration, is angular velocity, , is a compensation coefficient, which is dynamically adjusted according to the motion intensity: , ;

[0104] S2.43, for each type of motion scene, the compensated feature representation is respectively modeled by Gaussian mixture, and three independent state model subsets (probability density functions) are established: a static model (used to describe the electrical and mechanical behavior pattern of the device in a static state), a regular motion model (used to describe the anti-disturbance behavior pattern in a stable state), and a severe motion model (used to describe the anti-impact behavior pattern of the device in severe motion (such as emergency stop, slope switching));

[0105] The probability density function of each sub-model is:

[0106] ;

[0107] wherein, , indicates the currently selected motion scene category, is severe motion, is regular motion, is a static state, indicates the feature representation In the first Gaussian Mixture Probability Density Function value corresponding to the motion class, indicating the possibility of the feature appearing in the motion state, Gaussian component index in the current mixture model, Gaussian component number used in the motion class (more components can be used in the stationary state to improve precision), Gaussian component index in the current mixture model, Gaussian component index in the current mixture model, Weight of the Gaussian distribution in the scene class, Weight of the Gaussian distribution in the scene class, Weight of the Gaussian distribution in the scene class, Multidimensional Gaussian distribution probability density function, Mean vector of the Gaussian distribution in the scene class, Mean vector of the Gaussian distribution in the scene class, Mean vector of the Gaussian distribution in the scene class, Mean vector of the Gaussian distribution in the scene class, Mean vector of the Gaussian distribution in the scene class, Mean vector of the Gaussian distribution in the scene class,

[0108] S2.44, calculate the motion intensity of the current window (by weighted synthesis of acceleration and angular velocity indicators), and construct a scene awareness weight vector based on the motion intensity, with the weight distribution order being: intense motion > regular motion > stationary, indicating the degree of influence of each scene model on the current sample;

[0109] The scene awareness weight vectors are respectively:

[0110] (intense motion);

[0111] (regular motion);

[0112] (stationary);

[0113] wherein, probability weight of the current window in the intense motion state, probability weight of the current window in the regular motion state, probability weight of the current window in the stationary state, natural constant;

[0114] S2.45, input the perturbation-compensated feature representation into the stationary model, the regular motion model and the intense motion model respectively, calculate the state probability of each model, combine according to the fusion weight (entropy fusion algorithm), and generate the final state probability;

[0115] The final state probability is:

[0116] .

[0117] S2.5, analyze the natural evolution path of the key features in the time dimension, construct the state transition probability combined with the sliding window sequence data, and identify the potential evolution trend from normal → slight abnormal → failure;

[0118] The state transition probability is

[0119]

[0120] In the formula, represents the transition probability from state to state , represents the state index, represents the number (or frequency) of occurrences of the transition from state to state in the historical data, represents the total number of categories of states, represents the category index;

[0121] S2.6, based on steps S2.4 and S2.5 (according to the distance, similarity, reconstruction error, etc. in the state space), a trajectory-aware motion adaptive health degree evaluation model is constructed (step S2.4 provides the trajectory position and metric basis for the probabilistic modeling of the device state; step S2.5 provides the time pattern and health trend of the trajectory evolution).

[0122] Further, the trajectory-aware motion adaptive health degree evaluation model is constructed, including the following steps:

[0123] When the device is subjected to motion disturbance, it is difficult for the operation and maintenance personnel to judge whether the health degree decline is caused by real failure, electrical fluctuation, or mechanical vibration. The traceability capability is missing. The health score can distinguish whether the fluctuation is caused by the ontology failure or the external disturbance. The trajectory health degree term smooths the instantaneous abnormality, identifies the real degradation trend, introduces the electrical, mechanical and trajectory three dimensions, constructs a unified health cube, and realizes precise visual evaluation. The present scheme solves the motion interference problem by dynamic reference switching, the trajectory continuity detection suppresses the instantaneous jump, and the three-dimensional health degree decomposition realizes precise tracing;

[0124] S2.61, based on the motion intensity data (static, regular motion, and intense motion), select the corresponding health state reference vector group from the preset state space reference library (the state space reference library is composed of three groups of state vectors: static state reference library, regular motion reference library, and intense motion reference library; each reference library is a set of state vectors);

[0125] ​​​​S2.62, difference operation between the current sliding window and the feature vectors of the previous two sliding windows to construct a state trajectory second-order difference detection function;

[0126] The state trajectory second-order difference detection function is:

[0127] ;

[0128] In the formula, indicates the trajectory continuity score, and the closer the value is to 1, the smoother the state change is, indicates time;

[0129] If , it indicates that there is a mutation in the state trajectory, triggering the credibility correction mechanism (the credibility correction mechanism is a dynamic judgment and regulation mechanism for the credibility of the health score based on the trajectory evolution trend, which realizes intelligent suppression of false positives caused by motion disturbance through: trajectory smoothness detection (S2.61), Sigmoid mapping correction factor (S2.62), fusion formula adjustment score (S2.63));

[0130] S2.63, calculate the electrical health degree, mechanical health degree and trajectory health degree based on the state space reference library respectively;

[0131] The electrical health degree is:

[0132] ;

[0133] In the formula, indicates the electrical feature importance factor, which is used to control the sensitivity of the health degree to deviation, indicates the reference vector in the electrical state space (i.e., the mean value of the features corresponding to the normal electrical state), indicates the variance of the electrical features, which controls the decay rate, indicates the dimension of the electrical features;

[0134] The mechanical health degree is:

[0135] ;

[0136] In the formula, indicates the change amount of the device posture (position + angle, specifically three position components (x, y, z coordinates of the device in three-dimensional space) and three angle components (angles of rotation of the device around the three coordinate axes, i.e., roll angle, pitch angle and yaw angle)) in the current sliding window, indicates the mechanical allowable disturbance threshold value, which is used to normalize the offset;

[0137] The trajectory health degree​ is:

[0138]

[0139] wherein, represents a trajectory sensitivity factor (positive value), which adjusts the influence degree of trajectory mutation on the score, and is usually in the range of 2-5, and the greater the value, the more sensitive to mutation, represents the base of natural logarithm;

[0140] S2.64, dynamically adjusting the weight of each health component in S2.63 according to the motion intensity, and calculating the fused health degree

[0141]

[0142] wherein, represents the weight coefficient of electrical health degree, represents the weight coefficient of mechanical health degree;

[0143] S2.65, integrating the electrical health degree, the mechanical health degree and the trajectory health degree into a three-dimensional health vector output, and outputting the fused health degree score, which is used for the state diagnosis module to further perform fault classification and visual display, and provides interpretable and multi-dimensional health evaluation basis for operation and maintenance.

[0144] The state diagnosis module 32 performs abnormal diagnosis on the current collected features and the established state model;

[0145] The state diagnosis module 32 judges the equipment operation state according to the three-dimensional health vector and the fused health degree score; wherein, the fused health degree score is used to determine the overall health grade of the equipment, and the state is divided into normal, mild abnormality, moderate abnormality and serious abnormality according to the grading threshold, and the component analysis of the three-dimensional health vector is introduced for tracing the abnormal reason of the equipment. If the electrical health degree decreases significantly, it is determined that there is an electrical fault, such as insulation deterioration and short-circuit fluctuation. If the mechanical health degree decreases, it indicates that there is a device structure loosening or position abnormality. If the trajectory health degree is low, the health decline is caused by pseudo-abnormality caused by motion disturbance.

[0146] The visual display unit 4 displays the equipment operation state data and the diagnosis result in the form of graphical interface on the visual terminal.

[0147] In this embodiment, the visual display unit 4 includes an equipment state display module 41, a topological space correlation display module 42, a motion-electricity coupling dynamic mapping module 43, and a health degree score display module 44.

[0148] ​​​The device state display module 41 displays the device running data and device state results (feature values, alarm states) based on the data acquisition module 11 and the edge computing and analysis unit 2.

[0149] The topology space correlation display module 42 represents the space topology relationship graph constructed based on the edge computing and analysis unit 2, and updates the device layout structure in real time.

[0150] The motion-electricity coupling dynamic mapping module 43 displays abnormal contrasts based on the state probability calculated by the state model establishment module 31, and highlights which abnormal fluctuations are caused by motion and which are caused by device itself abnormalities.

[0151] The health degree scoring display module 44 displays the health degree of the device based on the diagnosis results of the state diagnosis module 32.

[0152] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application.

Claims

1. A mobile substation-based device operation state visual analysis system, characterized by, The application relates to a mobile substation operation state monitoring system based on edge computing, which comprises the following units: a data acquisition and transmission unit (1) which collects device operation data in real time from various devices of a mobile substation and transmits the device operation data to an edge computing and analysis unit (2); the edge computing and analysis unit (2) which extracts key features of the device operation data based on a topology-dynamic coupling feature extraction technology, identifies abnormal states and generates preliminary alarms; a state modeling and diagnosis unit (3) which constructs a trajectory-aware motion adaptive health degree evaluation model based on the key features and diagnoses the device state; a visual display unit (4) which displays device operation state data and diagnosis results in a graphical interface form on a visual terminal; the edge computing and analysis unit (2) which identifies abnormal states and generates preliminary alarms comprises the following steps: S1.1, dividing the received device operation data into sliding windows with a fixed length n according to a time axis; S1.2, processing the original device operation data in the sliding window, and extracting key feature indexes based on a topology-dynamic coupling feature extraction technology; S1.3, judging whether the device operation data is abnormal based on a threshold comparison rule, and if the device operation data is abnormal, generating a preliminary alarm information and transmitting the abnormal operation data to the state modeling and diagnosis unit (3); in the step S1.2, the original device operation data in the sliding window is processed, and key feature indexes are extracted based on a topology-dynamic coupling feature extraction technology, which comprises the following steps: S1.21, statistically analyzing and processing the voltage and current of each measuring point in the sliding window to extract electrical features; S1.22, dynamically establishing a spatial topology relationship graph among the devices according to the device spatial position data collected in the sliding window, and updating the spatial topology relationship graph once in each sliding window to dynamically reflect the device layout structure; S1.23, extracting the rotation angle and displacement of the device from the IMU data in the sliding window to construct a two-dimensional motion compensation matrix; S1.24, calculating the electrical-spatial distance ratio of each pair of adjacent devices based on the spatial topology relationship graph and the electrical state of the devices to reflect whether the electrical difference between the devices changes with the spatial layout; S1.25, performing difference analysis on the voltage and current signals after applying the motion compensation matrix and the original uncompensated signals; S1.26, calculating the importance score of each device node in the topology graph, and screening the device nodes based on the importance score; S1.27, uniformly organizing the key features of each device in the current sliding window into a four-dimensional feature representation matrix. 2.The mobile substation-based equipment operating state visualized analysis system according to claim 1, characterized in that: The data acquisition and transmission unit (1) comprises a data acquisition module (11) and a data transmission module (12); the data acquisition module (11) collects device operation data in real time from various devices of a mobile substation and pre-processes the device operation data; the data transmission module (12) transmits the pre-processed device operation data to the edge computing and analysis unit (2) by using a wireless communication mode. 3.The mobile substation based equipment operating state visualized analysis system of claim 1, wherein: The state modeling and diagnosis unit (3) comprises a state model establishment module (31) and a state diagnosis module (32); The state model establishment module (31) establishes a device health degree model based on historical operation data of the device and key features of the device; The state diagnosis module (32) performs abnormal diagnosis on the current collected features and the established state model.

4. The mobile substation based equipment operating state visualized analysis system according to claim 3, characterized in that: The state model establishment module (31) establishes a device health degree model based on historical operation data of the device and key features of the device, comprising the following steps: S2.1, receiving the four-dimensional feature representation matrix transmitted by the edge computing and analysis unit (2); S2.2, classifying and labeling the feature data in the four-dimensional feature representation matrix; S2.3, using a feature importance analysis method to quantify the influence degree of each dimension feature in different device state categories, and outputting the feature weight distribution corresponding to each device state; S2.4, based on the labeled feature data, using a motion decoupling spatio-temporal Gaussian mixture model to probabilistically model the device state; S2.5, analyzing the natural evolution path of the key features in the time dimension, and combining the sliding window sequence data to construct the state transition probability; S2.6, based on steps S2.4 and S2.5, constructing a trajectory-aware motion adaptive health degree evaluation model.

5. The mobile substation based equipment operating state visualized analysis system according to claim 4, characterized in that: In S2.4, the motion decoupling spatio-temporal Gaussian mixture model is used to probabilistically model the device state, comprising the following steps: S2.41, defining the motion intensity factor According to the motion intensity factor, the feature data is dynamically divided into three categories of motion scenes: a static state group, a regular motion group, and an intense motion group. S2.42, analyzing the acceleration vector of the device in the sliding window, constructing a motion compensation projection matrix for different motion interference components, decoupling and projecting the four-dimensional feature representation matrix to obtain the disturbance compensated feature representation; S2.43, for each motion scene, the compensated feature representation is respectively modeled by Gaussian mixture, and three independent state model subsets are established: static model, regular motion model and intense motion model; S2.44, calculating the motion intensity of the current window, and constructing a scene-aware weight vector based on the motion intensity; S2.45, inputting the disturbance compensated feature representation into the static model, the regular motion model and the intense motion model respectively, calculating the state probability under each model, combining according to the fusion weight, and generating the final state probability. 6.The mobile substation based equipment operating state visualized analysis system according to claim 4, characterized in that: In S2.6, the trajectory-aware motion adaptive health degree evaluation model is constructed, comprising the following steps: S2.61, selecting the corresponding health state reference vector group from the preset state space reference library based on the motion intensity data; S2.62, performing difference operation on the feature vectors of the current sliding window and the previous two sliding windows to construct a state trajectory second-order difference detection function; S2.63, calculating the electrical health degree, the mechanical health degree and the trajectory health degree based on the state space reference library; S2.64, dynamically adjusting the weights of each health component in S2.63 according to the motion intensity, and calculating the fusion health degree; S2.65, integrating the electrical health degree, the mechanical health degree and the trajectory health degree into a three-dimensional health vector output, and outputting the fusion health degree score. 7.The mobile substation based equipment operating state visualized analysis system according to claim 3, wherein: The state diagnosis module (32) judges the equipment operation state according to the three-dimensional health vector and the fusion health degree score; wherein the fusion health degree score is used to determine the overall health level of the equipment, the state is divided according to the grading threshold, and the component analysis of the three-dimensional health vector is introduced for tracing the abnormal reason of the equipment. 8.The mobile substation based equipment operating state visualized analysis system of claim 1, wherein: The visual display unit (4) includes an equipment state display module (41), a topological space correlation display module (42), a motion-electric coupling dynamic mapping module (43), and a health degree score display module (44); The equipment state display module (41) displays the equipment operation data and the equipment state result based on the data acquisition module (11) and the edge computing and analysis unit (2); The topological space correlation display module (42) represents the space topological relationship diagram constructed based on the edge computing and analysis unit (2), and updates the equipment layout structure in real time; The motion-electric coupling dynamic mapping module (43) displays the abnormal contrast based on the state probability calculated by the state model establishment module (31); The health degree score display module (44) displays the health degree of the equipment based on the diagnosis result of the state diagnosis module (32).

Images