Power distribution network fault data preprocessing method and system based on multi-source information fusion

Abstract

The invention discloses a power distribution network fault data preprocessing method and system based on multi-source information fusion, and relates to the technical field of power systems, and the method comprises the steps: constructing a dynamic topology manifold based on the physical connection relation of power distribution network equipment and multi-source data; manifold deformation caused by line switching or sudden load change is responded, and abnormal data are recognized in combination with the manifold curvature and the equipment connection path distance; encoding the abnormal data, the meteorological data and the equipment state data into quantum states, and generating entanglement feature vectors through quantum gate operation; performing chaos sequence modulation on the basic features to generate modulation features; performing quantum fusion transformation on the modulation feature and the entanglement feature vector to generate an enhanced feature; calculating topological space correlation, chaos behavior similarity and quantum correlation based on the enhanced features, and constructing a multi-dimensional fusion weight; and performing weighted fusion on the enhanced features according to the multi-dimensional fusion weight, and outputting a preprocessed fault feature vector.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, more particularly to the technical field of intelligent power distribution system, power distribution switch control equipment manufacturing, and specifically to a power distribution network fault data preprocessing method and system based on multi-source information fusion. BACKGROUND

[0002] With the continuous improvement of the intelligent power grid and the power distribution network automation level, the data scale and complexity generated in the process of power grid operation continue to increase. As a key link of power system operation and maintenance, the core significance of power distribution network fault data preprocessing lies in that through systematic processing of multi-source data, characteristic information accurately reflecting the operation state of the power grid is extracted, providing reliable data support for subsequent fault diagnosis, positioning and disposal, so as to effectively guarantee the safe and stable operation of the power distribution network and improve the response efficiency and decision-making scientificity of the power system in dealing with faults.

[0003] The traditional data preprocessing method often analyzes various types of data independently, and it is difficult to fully consider the potential relationship between the data and the dynamic changes of the power distribution network topology structure. When dealing with data generated due to line switching, load mutation and other situations, there are problems such as inability to accurately identify abnormal data, which affects the accuracy and efficiency of fault diagnosis.

[0004] At present, there is no effective solution to the above problems. SUMMARY

[0005] The embodiments of the present application provide a power distribution network fault data preprocessing method and system based on multi-source information fusion to solve the above technical problems.

[0006] The present application provides a power distribution network fault data preprocessing method based on multi-source information fusion, comprising:

[0007] constructing a dynamic topological manifold based on the physical connection relationship of the power distribution network equipment and multi-source data, wherein the multi-source data includes electrical quantity data, meteorological data and equipment state data;

[0008] in response to the manifold deformation caused by line switching or load mutation, identifying abnormal data in combination with the manifold curvature and the equipment connection path distance;

[0009] encoding the abnormal data, the meteorological data and the equipment state data into quantum states, and generating an entangled feature vector through quantum gate operation;

[0010] extracting basic features from the local structure of the dynamic topological manifold, generating modulation features after chaotic sequence modulation of the basic features; performing quantum fusion transformation on the modulation features and the entangled feature vector to generate enhanced features;

[0011] The topology space correlation, chaotic behavior similarity and quantum correlation are calculated based on the enhanced features, and a multi-dimensional fusion weight is constructed.

[0012] The enhanced features are weighted and fused according to the multi-dimensional fusion weight, and a preprocessed fault feature vector is output.

[0013] Further, the dynamic topology manifold is constructed based on the physical connection relationship of the power distribution network equipment and the multi-source data, comprising:

[0014] The physical connection relationship between the switch nodes of the power distribution network is obtained, and an adjacency matrix for representing the equipment connection topology structure is formed;

[0015] The current phase information in the electrical quantity data is analyzed, and the dynamic impedance parameters between nodes are calculated;

[0016] Combined with the temperature and humidity distribution in the meteorological data, a manifold curvature field model is established;

[0017] According to the insulation aging degree in the equipment state data, the manifold metric tensor is determined;

[0018] The adjacency matrix, the dynamic impedance parameters, the manifold curvature field model and the manifold metric tensor are fused, and a basic topology manifold is constructed by a Riemann manifold embedding algorithm;

[0019] The dynamic response mechanism is configured for the basic topology manifold, so that the basic topology manifold has the characteristics of adaptive deformation with the running state;

[0020] The dynamic topology manifold that fuses the equipment connection relationship, electrical characteristics and environmental state is output based on the basic topology manifold.

[0021] Further, the manifold deformation caused by the response line switching or load mutation is combined with the manifold curvature and the equipment connection path distance to identify abnormal data, comprising:

[0022] The local curvature change rate of the dynamic topology manifold in the deformation process is monitored;

[0023] Based on the physical connection relationship, the shortest topology path distance between the switch nodes of the power distribution network is calculated;

[0024] Based on the electrical quantity data, a characteristic fluctuation sequence reflecting the data fluctuation characteristics is generated using a preset chaotic mapping rule, and the Lyapunov exponent of the characteristic fluctuation sequence is calculated;

[0025] The geodesic distance representing the actual connection relationship between the electrical quantity data points in the dynamic topology manifold is calculated;

[0026] The deviation rate of the geodesic distance and the shortest topology path distance between the corresponding nodes is calculated.

[0027] When the local curvature rate of change exceeds a first dynamic threshold, and the Lyapunov index exceeds a second dynamic threshold, mark the corresponding region as a candidate abnormal region;

[0028] In the candidate abnormal region, if the deviation rate continues to exceed a third dynamic threshold, determine that the corresponding electrical quantity data point is abnormal data.

[0029] Further, the encoding of the abnormal data, the weather data and the equipment state data into quantum states, and the generation of an entangled feature vector through quantum gate operation, comprises:

[0030] Encode the abnormal data, the weather data and the equipment state data into independent quantum states respectively to obtain abnormal data quantum state, weather data quantum state and equipment state quantum state;

[0031] Perform a double-bit entanglement gate operation on the abnormal data quantum state and the equipment state quantum state to generate an abnormal and equipment entangled state;

[0032] Take the weather data quantum state as a control bit, and perform a controlled quantum gate operation on the abnormal and equipment entangled state to generate a cross-modal entangled state;

[0033] Perform a quantum feature transformation operation on the cross-modal entangled state to output an entangled feature vector representing quantum correlation of multi-source data.

[0034] Further, the method further comprises:

[0035] Based on the local geometric properties of the dynamic topological manifold, extract a basic feature representing the device connection topology structure, the basic feature including local Gaussian curvature, geodesic line gradient and connection coefficient;

[0036] Adjust the control parameters of the chaotic system using the local Gaussian curvature to generate a chaotic modulation sequence associated with the manifold local structure of the dynamic topological manifold;

[0037] Fuse the basic feature and the chaotic modulation sequence to generate a primary modulation feature.

[0038] Further, the method further comprises:

[0039] Based on the value of the local Gaussian curvature, perform weighted normalization processing on the primary modulation feature to output a modulation feature with topological constraints and chaotic characteristics.

[0040] Further, the topological space correlation, chaotic behavior similarity and quantum correlation are calculated based on the enhanced features, and a multi-dimensional fusion weight is constructed, including:

[0041] A local Gaussian curvature component is extracted from the enhanced features, and a curvature covariance matrix between switch nodes of the power distribution network is calculated;

[0042] Based on the chaotic modulation sequence, a phase space is reconstructed and the maximum Lyapunov exponent of the current chaotic sequence is calculated;

[0043] The maximum Lyapunov exponent is compared with the fractal dimension difference of the fractal dimension benchmark value in the historical fault state;

[0044] Based on the entanglement feature vector and the typical fault mode quantum state, a quantum correlation strength parameter is calculated; and a device topology correlation weight is generated according to the curvature covariance matrix;

[0045] Based on the fractal dimension difference, a chaotic behavior similarity coefficient is calculated;

[0046] The device topology correlation weight, the chaotic behavior similarity coefficient and the quantum correlation strength parameter are fused to construct a multi-dimensional fusion weight.

[0047] Further, the enhanced features are weighted and fused according to the multi-dimensional fusion weight, and a preprocessed fault feature vector is output, including:

[0048] The enhanced features are mapped to the tangent space of the dynamic topological manifold;

[0049] In the device connection topology dimension, the device topology correlation weight is applied to perform weighted aggregation of adjacent nodes to generate a topology weighted feature;

[0050] In the data fluctuation characteristic dimension, the chaotic behavior similarity coefficient is used to enhance the amplitude of abnormal fluctuation characteristics to generate chaotic enhanced features;

[0051] In the cross-modal correlation dimension, the quantum correlation strength parameter weighted feature selection is implemented to generate quantum screening features.

[0052] Further, the method further includes:

[0053] The topology weighted feature, the chaotic enhanced feature and the quantum screening feature are fused to form a three-dimensional feature tensor;

[0054] Based on the manifold metric tensor, the Riemann distance between the three-dimensional feature tensor and the historical fault template feature is calculated;

[0055] When the Riemann distance is less than an adaptive threshold, the three-dimensional feature tensor is mapped to a fault feature space;

[0056] output the mapped three-dimensional feature vector as the preprocessed fault feature vector.

[0057] The application provides a power distribution network fault data preprocessing system based on multi-source information fusion, comprising:

[0058] a dynamic topological manifold construction module, configured to construct a dynamic topological manifold based on physical connection relationships of power distribution network equipment and multi-source data, wherein the multi-source data comprises electrical quantity data, meteorological data and equipment state data;

[0059] an abnormal data identification module, configured to identify abnormal data in combination with manifold curvature and equipment connection path distance in response to manifold deformation caused by line switching or load mutation;

[0060] an entangled feature vector generation module, configured to encode the abnormal data, the meteorological data and the equipment state data into quantum states, and generate an entangled feature vector through quantum gate operation;

[0061] an enhanced feature generation module, configured to extract basic features from a local structure of the dynamic topological manifold, generate modulation features after chaotic sequence modulation of the basic features, and perform quantum fusion transformation on the modulation features and the entangled feature vector to generate enhanced features;

[0062] a multi-dimensional fusion weight construction module, configured to calculate topological space correlation, chaotic behavior similarity and quantum correlation based on the enhanced features, and construct multi-dimensional fusion weights;

[0063] a fault feature vector output module, configured to perform weighted fusion on the enhanced features according to the multi-dimensional fusion weights, and output a preprocessed fault feature vector.

[0064] Based on the embodiments provided in the application, the dynamic topological manifold is constructed based on the physical connection relationships of the power distribution network equipment and the multi-source data, and the electrical quantity data, the meteorological data and the equipment state data are organically integrated, and the potential relationship between the data is fully considered. Compared with the traditional method of independently analyzing various types of data, the power distribution network operation state can be more comprehensively and truly reflected. When the line switching or load mutation causes the manifold deformation, the abnormal data is identified in combination with the manifold curvature and the equipment connection path distance, effectively solving the problem that the traditional method cannot accurately identify the abnormal data in such cases, and providing a more reliable data basis for subsequent fault diagnosis.

[0065] Abnormal data, weather data and equipment state data are encoded into quantum states, entangled feature vectors are generated through quantum gate operation, and meanwhile, basic features are extracted from the local structure of the dynamic topological manifold and modulated into chaotic sequences to generate modulation features, and then the modulation features and the entangled feature vectors are subjected to quantum fusion transformation to generate enhanced features, and this innovative data processing method can deeply mine data features and obtain more representative fault features. Based on the enhanced features, topological space correlation, chaotic behavior similarity and quantum correlation are calculated, multi-dimensional fusion weights are constructed, and the enhanced features are weighted and fused to output the preprocessed fault feature vector, fully considering the dynamic changes of the power distribution network topology, realizing effective fusion of multi-dimensional features, providing more accurate and comprehensive data support for subsequent fault diagnosis, positioning and disposal, effectively guaranteeing the safe and stable operation of the power distribution network, and significantly improving the response efficiency and decision-making scientificity of the power system when responding to faults. BRIEF DESCRIPTION OF DRAWINGS

[0066] The accompanying drawings, which are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the application without limiting it. In the drawings:

[0067] Figure 1 A flowchart of an optional power distribution network fault data preprocessing method based on multi-source information fusion according to an embodiment of the application;

[0068] Figure 2 A flowchart of another optional power distribution network fault data preprocessing method based on multi-source information fusion according to an embodiment of the application;

[0069] Figure 3 A structural diagram of an optional power distribution network fault data preprocessing system based on multi-source information fusion according to an embodiment of the application.

[0070] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings of the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0072] Optionally, as shown in Figure 1 The application provides a power distribution network fault data preprocessing method based on multi-source information fusion, which comprises:

[0073] S101, constructing a dynamic topological manifold based on a power distribution network equipment physical connection relationship and multi-source data, the multi-source data including electrical quantity data, meteorological data, and equipment state data;

[0074] S102, in response to a manifold deformation caused by line switching or load mutation, identifying abnormal data in combination with manifold curvature and equipment connection path distance;

[0075] The line switching refers to an operation in a power distribution network that changes the connection mode and on-off state of a power line by operating a switching device (such as a circuit breaker, a knife switch, etc.), thereby changing the power transmission path. For example, when a fault occurs in a certain line of the power distribution network, an operator remotely controls the circuit breakers at both ends of the fault line to be disconnected, and at the same time, closes the tie switch, thereby switching the power supply of the region originally supplied by the fault line to other normal lines, so as to ensure the power supply of the non-fault region.

[0076] The load mutation refers to a phenomenon that the power load in the power distribution network suddenly increases or decreases greatly in a short time, resulting in a sharp fluctuation of electrical parameters such as current and voltage. For example, when a large motor in an industrial park is started, a large amount of power is consumed instantaneously, causing the current in the region to suddenly increase; or in a commercial area, air conditioners are turned on at the same time during the power consumption peak, causing the power load in the local region to increase greatly in a short time.

[0077] The manifold curvature is used to describe the bending degree of the local region of the dynamic topological manifold, and the bending degree reflects the change of the connection relationship of the equipment in the power distribution network and the severity of the data fluctuation. For example, when a load mutation occurs in a certain region of the power distribution network, causing the data such as current and voltage in the region to change sharply, a larger curvature will be generated on the corresponding region of the dynamic topological manifold, like a wrinkle on a flat piece of paper.

[0078] The equipment connection path refers to the actual physical connection line between each device (such as a switch, a transformer, a line, etc.) in the power distribution network, and the length of the path can be measured by calculating the number of nodes on the connection path. For example, from switch A to switch B, switch C and switch D need to be passed through, so there are 3 connection lines on the connection path from A to B, and the path length is 3.

[0079] S103, encoding the abnormal data, the meteorological data, and the equipment state data into quantum states, and generating an entangled feature vector through quantum gate operation;

[0080] S104, extracting a basic feature from a manifold local structure of the dynamic topological manifold, generating a modulated feature after chaos sequence modulation of the basic feature; and performing quantum fusion transformation on the modulated feature and the entangled feature vector to generate an enhanced feature;

[0081] The local structure of the manifold refers to the geometric characteristics in a certain small area of the dynamic topological manifold, including the connection mode of the nodes in the area, the curvature distribution of the manifold, and the like. For example, the local structure of the corresponding area on the dynamic topological manifold near a feeder of a power distribution network contains the connection relationship of the switch nodes at both ends of the feeder and the curvature change of the manifold around the feeder due to load change or environmental factors.

[0082] The quantum fusion transformation is to combine the features modulated by the chaotic sequence and the entangled feature vectors generated by quantum state encoding and quantum gate operation, using the related operations in quantum computing, such as quantum state superposition and controlled operation, to generate enhanced features containing more information and better reflecting the data features. For example, the modulated features describing the connection topology of the device are regarded as a quantum state, and the entangled feature vectors reflecting the quantum correlation of multi-source data are regarded as another quantum state. Through the controlled non-gate operation in quantum computing, the two quantum states interact with each other, so that the generated enhanced features contain both the device connection relationship information and the quantum correlation characteristics between multi-source data.

[0083] S105, calculating the topological space correlation, chaotic behavior similarity, and quantum correlation based on the enhanced features, and constructing multi-dimensional fusion weights;

[0084] S106, weighting and fusing the enhanced features according to the multi-dimensional fusion weights, and outputting the preprocessed fault feature vector.

[0085] Based on the embodiments provided in the present application, by constructing a dynamic topological manifold based on the physical connection relationship of the power distribution network equipment and the multi-source data, the electrical quantity data, meteorological data, and equipment state data are organically integrated, and the potential relationship between the data is fully considered. Compared with the traditional way of independently analyzing various types of data, the power distribution network operation state can be more comprehensively and truly reflected. When the line switching or load mutation causes the manifold to deform, the abnormal data is identified in combination with the manifold curvature and the device connection path distance, effectively solving the problem that the traditional method cannot accurately identify abnormal data in such cases, and providing a more reliable data basis for subsequent fault diagnosis.

[0086] This innovative data processing method encodes abnormal data, meteorological data, and equipment status data into quantum states, generates entangled feature vectors through quantum gate operations, and simultaneously extracts fundamental features from the local structure of dynamic topological manifolds and modulates them with chaotic sequences to generate modulated features. These modulated features are then quantum-fused with the entangled feature vectors to generate enhanced features. This method enables in-depth mining of data features and the acquisition of more representative fault characteristics. Based on the enhanced features, topological spatial correlation, chaotic behavior similarity, and quantum correlation are calculated to construct multi-dimensional fusion weights. These weighted fusion weights are then used to output preprocessed fault feature vectors. This approach fully considers the dynamic changes in the distribution network topology and achieves effective fusion of multi-dimensional features. It provides more accurate and comprehensive data support for subsequent fault diagnosis, location, and handling, effectively ensuring the safe and stable operation of the distribution network and significantly improving the response efficiency and scientific decision-making of the power system when dealing with faults.

[0087] Furthermore, such as Figure 2 As shown, a dynamic topology manifold is constructed based on the physical connection relationships of distribution network equipment and multi-source data, including:

[0088] S201, Obtain the physical connection relationship between the distribution network switch nodes and form an adjacency matrix to characterize the equipment connection topology;

[0089] S202, analyze the current phase information in the electrical quantity data, and calculate the dynamic impedance parameters between nodes;

[0090] In this embodiment, calculating the dynamic impedance parameters between nodes includes: analyzing the current phase information of each node in the distribution network, and combining this with voltage conditions to calculate the impedance values ​​between nodes that change in real time as the power system operating state changes. For example, at a certain moment, if the current phase of node M lags behind that of node N by a certain angle, and the voltage amplitudes of the two nodes are simultaneously acquired, the current dynamic impedance between node M and node N is calculated based on the relationship between voltage and current. When the line load changes, the current phase and voltage are acquired again, and the dynamic impedance is recalculated; this value will change with the load change.

[0091] S203, combined with the temperature and humidity distribution in meteorological data, establish a manifold curvature field model;

[0092] In the embodiment, the manifold curvature field model is established, including: constructing a model reflecting the curvature size distribution of different regions on the dynamic topological manifold according to meteorological data (such as temperature, humidity, etc.), and the regions with different meteorological conditions correspond to different curvature values on the manifold. For example, in a high-temperature and humid region, the performance of the power distribution network equipment is easily affected by such an environment, so in the manifold curvature field model, the manifold curvature corresponding to the region is set to a higher value; while in a dry and cold region, the performance of the equipment is less affected by the environment, and the corresponding manifold curvature is set to a lower value. In this way, the manifold curvature field model reflects the influence of meteorological conditions on the operation of the power distribution network.

[0093] In S204, the manifold metric tensor is determined according to the insulation aging degree in the equipment state data.

[0094] In the embodiment, the manifold metric tensor is determined, including: determining the metric tensor describing the geometric properties of the dynamic topological manifold according to the equipment state data (such as the insulation aging degree of the equipment, the running time of the equipment, etc.), and the element values of the metric tensor reflect the influence of the equipment state on the geometric relationship of the topological manifold. For example, for a transformer with a high insulation aging degree, when determining the manifold metric tensor, the tensor element value of the dimension related to the insulation state is increased, so that the geometric metric relationship of the region corresponding to the transformer on the manifold can better reflect its aging state; and for a newly put into operation and well-conditioned equipment, the corresponding tensor element value is set to be low

[0095] In S205, the adjacent matrix, the dynamic impedance parameter, the manifold curvature field model and the manifold metric tensor are fused, and the basic topological manifold is constructed by the Riemann manifold embedding algorithm.

[0096] The Riemann manifold embedding algorithm is a method of mapping high-dimensional complex power distribution network multi-source data (such as equipment connection relationship, electrical parameters, meteorological data, etc.) to a low-dimensional Riemann manifold space, and in the mapping process, the original distance, angle and other geometric relationships between the data are preserved as much as possible. For example, there are a large number of equipment in the power distribution network, and the connection relationship between them and various operation data constitute high-dimensional data. Using the Riemann manifold embedding algorithm is like "folding" these complex high-dimensional data into a low-dimensional space (such as a three-dimensional space), in which the proximity relationship between the equipment, the similarity of the data and other information can still be reflected, facilitating subsequent analysis and processing of the data.

[0097] In S206, a dynamic response mechanism is configured for the basic topological manifold, so that the basic topological manifold has the characteristics of self-adapting deformation with the running state.

[0098] In the embodiment, the operating state refers to various conditions of the power distribution network during operation, including electrical parameters (such as current, voltage, power, etc.), device states (such as on-off of switches, whether the device is faulty, aging degree of the device, etc.), and environmental conditions (such as temperature, humidity, wind speed, etc.). For example, at a certain moment, the current of a line in the power distribution network is in the normal range, the switch is in the closed state, the device is not faulty but has a certain degree of aging, the environmental temperature is high, and the humidity is low. The comprehensive conditions of these electrical parameters, device states, and environmental conditions constitute the operating state of the power distribution network at this moment.

[0099] In S207, a dynamic topological manifold that integrates the device connection relationship, electrical characteristics, and environmental state is output based on the basic topological manifold.

[0100] In a specific implementation, the dynamic topological manifold is constructed based on the following formula:

[0101]

[0102] wherein M(t) is the dynamic topological manifold at time t, which integrates the device connection relationship, electrical characteristics, and environmental state and self-adapts to deformation with operation time; t0 is the initial construction time of the manifold; is the outer product calculation of the matrix; A is the device adjacency matrix; Z norm is the normalized dynamic impedance; is the outer product operation of the adjacency matrix A (0 / 1 Boolean value) and the normalized impedance Z norm , which couples the device connection relationship and the electrical characteristics, for example, when the sectional switch is open (A ij = 0), the outer product result forces the node impedance Z ij to be invalid, and the local topology of the manifold is broken; I is the unit matrix; R c is the meteorological curvature factor, example 1: when a typhoon passes, a temperature gradient of 0.3°C / km and a humidity gradient of 0.4g / m 3 / km are monitored, the cross product calculation obtains a meteorological curvature factor of 0.12 per kilometer, example 2: in winter icing weather, a temperature gradient of 0.5°C / km (strong cold air) and a humidity gradient of 0.1g / m 3 / km (dry), the cross product is calculated as a curvature factor of 0.05 per kilometer; G is the equipment state attenuation coefficient, example 1: a power distribution transformer in operation for 20 years, with an insulation aging degree of 90%, the calculated equipment state attenuation coefficient is 0.1, example 2: a newly installed switch cabinet, with an insulation aging degree of 5%, the calculated equipment state attenuation coefficient is 0.95, the manifold geometry representation maintains 95% integrity, marked as a low-risk node; K is the environmental sensitivity, for example, in a sandstorm weather, the environmental sensitivity can be taken as 0.8, in a sunny weather, the environmental sensitivity can be taken as 0.5; X is the time attenuation constant, in a high sampling frequency scenario (period of 0.1s), the time attenuation constant is 0.1 per second, in a low sampling frequency scenario (period of 0.5s), the time attenuation constant is 0.2 per second; R is the Riemann manifold embedding operator, used to map the input matrix to a differential manifold, for example, projecting the electrical impedance matrix to the geodesic line distribution on a three-dimensional manifold surface.

[0103] Based on the embodiments provided in the present application, when constructing a dynamic topology manifold, an adjacency matrix is formed by acquiring the physical connection relationship of the switch nodes of the power distribution network, the dynamic impedance parameters are calculated by analyzing the current phase information in the electrical quantity data, the manifold curvature field model is established in combination with the meteorological data, and the manifold metric tensor is determined according to the equipment state data, and finally the basic topology manifold is constructed through the Riemann manifold embedding algorithm. After configuring a dynamic response mechanism for the basic topology manifold, the manifold can adaptively deform with the change of the operation state of the power distribution network, realizing the deep fusion of multi-dimensional information such as equipment connection topology structure, electrical characteristics and environmental state. Unlike the traditional method which only relies on static topology structure analysis, the dynamic topology manifold constructed by this method can reflect the topology change and operation state of the power distribution network under different working conditions in real time, providing a more accurate and dynamic space carrier for subsequent fault data processing, and greatly improving the adaptability to complex operation scenarios.

[0104] Further, in response to the deformation of the manifold caused by line switching or load mutation, abnormal data is identified in combination with the manifold curvature and the distance of the equipment connection path, including:

[0105] Monitoring the local curvature change rate of the dynamic topology manifold in the deformation process;

[0106] Based on the physical connection relationship, the shortest topology path distance between the switch nodes of the power distribution network is calculated;

[0107] Based on the electrical quantity data, a characteristic fluctuation sequence reflecting the data fluctuation characteristics is generated using a preset chaotic mapping rule, and the Lyapunov exponent of the characteristic fluctuation sequence is calculated;

[0108] In this embodiment, the preset chaotic mapping rule is a mathematical transformation rule set in advance, which converts the power grid electrical quantity data (such as the fluctuation of current and voltage) into a sequence with chaotic characteristics. The sequence has the characteristics of sensitivity to initial value and ergodicity, and can better reflect the fluctuation characteristics of the data. For example, a simple chaotic mapping rule is used, and each monitored current value is taken as input, and a new numerical sequence is generated after a plurality of specific operations (such as repeated multiplication, subtraction and remainder operation). In this sequence, even if the initial current value has only a slight difference, the sequence obtained after a plurality of operations will be completely different, and the values in the sequence will change irregularly within a certain range, showing chaotic characteristics.

[0109] calculating the geodesic distance representing the actual connection relationship between the electrical quantity data points in the dynamic topological manifold;

[0110] calculating the deviation rate of the geodesic distance and the shortest topological path distance between the corresponding nodes;

[0111] When the local curvature change rate exceeds the first dynamic threshold and the Lyapunov index exceeds the second dynamic threshold, the corresponding region is marked as a candidate abnormal region;

[0112] In the candidate abnormal region, if the deviation rate continuously exceeds the third dynamic threshold, the corresponding electrical quantity data point is determined to be abnormal data.

[0113] In a specific implementation, the identification of abnormal data is based on the following formula:

[0114]

[0115] where Anomaly is the abnormal data determination result (Boolean value); ΔG is the local Gaussian curvature change of the dynamic topological manifold; Δt is the data sampling time interval; is the Gaussian curvature change rate; λ L is the Lyapunov index; δ d is the path deviation rate; is the current variance; deg ins is the insulation aging degree; SOI is the meteorological disaster index; ∧ is the logical AND operator (AND), indicating that the three conditions in the formula must be true at the same time, and Anomaly is determined to be true.

[0116] The three dynamic thresholds are standards for judging whether the power grid data is abnormal, and will be adjusted according to the running state of the power grid. The first dynamic threshold is used to judge whether the local curvature change of the manifold is abnormal; the second dynamic threshold is used to judge whether the chaotic characteristics of the data fluctuation are abnormal; and the third dynamic threshold is used to judge whether the distance deviation between the data points is abnormal.

[0117] Exemplarily, the first dynamic threshold can be set to consider that the region may have an anomaly when the local curvature of the manifold changes by more than 0.6 in one second; it can also be adjusted to 0.8 according to actual conditions. The second dynamic threshold is determined to be abnormal when a certain index value reflecting the data fluctuation characteristics is greater than 0.3; or it is considered to be abnormal when it is greater than 0.4. The third dynamic threshold is determined to be abnormal when the deviation rate of the geodesic distance and the shortest topological path distance between the corresponding nodes is greater than 12%; it can also be set to be greater than 18% to be determined to be abnormal.

[0118] Based on the embodiments provided in the present application, in the process of identifying abnormal data, by monitoring the local curvature change rate during the deformation of the dynamic topological manifold, calculating the shortest topological path distance between nodes based on the physical connection relationship, generating a characteristic fluctuation sequence using electrical quantity data and calculating the Lyapunov index, and comparing the deviation rate of the geodesic distance and the shortest topological path distance, a multi-dimensional abnormal data judgment system is formed. When the local curvature change rate and the Lyapunov index exceed the dynamic threshold and the deviation rate continues to exceed the standard, the corresponding data point is determined to be abnormal. This method breaks through the limitations of traditional single threshold judgment of abnormal data, fully considers the comprehensive influence of the change of the distribution network topology structure, the data fluctuation characteristics and the node connection relationship, and can accurately capture abnormal data in the scene of line switching and load mutation, providing a more reliable and pure data basis for subsequent fault diagnosis, and effectively reducing the risk of data misjudgment.

[0119] Further, the abnormal data, weather data and equipment state data are encoded into quantum states, and entangled feature vectors are generated through quantum gate operations, including:

[0120] The abnormal data, weather data and equipment state data are respectively encoded into independent quantum states to obtain abnormal data quantum states, weather data quantum states and equipment state quantum states.

[0121] In this embodiment, encoding into independent quantum states means that the abnormal data, weather data and equipment state data of the distribution network are respectively converted into the state of the basic unit in quantum computing, i.e., a quantum bit. Each data has a corresponding quantum state representation form, and these quantum states are independent of each other. For example, the equipment abnormal data monitored at a certain time is converted into a quantum state according to its numerical value. This quantum state can be represented in the form of "the probability of being in state 0 is 0.3, and the probability of being in state 1 is 0.7". Similarly, the weather data and equipment state data are also converted into similar quantum states. The quantum state of each data exists independently and does not interfere with each other.

[0122] A double-bit entanglement gate operation is performed on the abnormal data quantum state and the equipment state quantum state to generate an abnormal and equipment entangled state.

[0123] In this embodiment, the two-bit entanglement gate operation is an operation on two quantum states, so that a special correlation, i.e., an entanglement, is generated between the two quantum states. Once entanglement occurs, the measurement of one quantum state will instantly affect the state of the other quantum state, regardless of the distance between the two quantum states. For example, the quantum state corresponding to the abnormal data and the quantum state corresponding to the device state data are taken as two inputs, and the two quantum states are entangled through the two-bit entanglement gate operation. The originally independent two quantum states are correlated with each other after the operation, for example, when the abnormal data quantum state is in a 0 state, the device state quantum state will be in a specific state, and vice versa.

[0124] The meteorological data quantum state is taken as a control bit, and a controlled quantum gate operation is performed on the abnormal and device entangled state to generate a cross-modal entangled state.

[0125] A quantum feature transformation operation is performed on the cross-modal entangled state to output an entangled feature vector representing the quantum correlation between the multi-source data.

[0126] In this embodiment, the quantum feature transformation operation is to process the quantum state after entanglement and other operations by using the characteristics of quantum computing, and extract a feature vector that can reflect the correlation characteristics between multi-source data. For example, for the cross-modal quantum state that has been entangled, a set of numerical values is extracted from the quantum state through a series of specific quantum computing steps (such as quantum rotation, quantum measurement, etc.), and the set of numerical values constitutes an entangled feature vector, which contains the correlation information between abnormal data, meteorological data and device state data.

[0127] Based on the embodiments provided in this application, the abnormal data, meteorological data and device state data are respectively encoded into quantum states, and a cross-modal entangled feature vector is generated through a two-bit entanglement gate operation, a controlled quantum gate operation and a quantum feature transformation. This process uses the superposition and entanglement characteristics of quantum computing to break the relative independence between data in traditional feature extraction methods and deeply excavate the implicit correlation between different types of data. This way of converting multi-source data into quantum states and performing entanglement operations makes the feature vector not only contain information of a single data, but also contain quantum correlation between data, providing a more rich and deep feature expression form for fault feature analysis and enhancing the representation ability of data features in complex fault scenarios.

[0128] Further, the basic features are extracted from the local structure of the dynamic topological manifold, and the modulation features are generated after the basic features are modulated by chaotic sequences, including:

[0129] Based on the local geometric properties of the dynamic topological manifold, the basic features representing the device connection topology structure are extracted, and the basic features include local Gaussian curvature, geodesic gradient and connection coefficient.

[0130] The control parameter of the chaotic system is dynamically adjusted by using local Gaussian curvature, and a chaotic modulation sequence associated with the local structure of the dynamic topological manifold is generated.

[0131] In the embodiment, the control parameter of the chaotic system is a key value for adjusting the operation behavior of the chaotic system. Different values of the control parameter will cause the chaotic system to generate chaotic sequences with different characteristics, such as complexity and variation law. For example, in a certain chaotic system, there is a control parameter X. When X has a small value, the generated chaotic sequence changes relatively gently, and the numerical fluctuation range is small. When X increases, the variation of the chaotic sequence becomes more violent, and the numerical value fluctuates irregularly in a larger range.

[0132] The control parameter is dynamically adjusted by using local Gaussian curvature, including automatically adjusting the control parameter of the chaotic system according to the Gaussian curvature of the local area of the dynamic topological manifold, so that the sequence generated by the chaotic system matches the change of the power distribution network topology structure. For example, in a certain area of the power distribution network, if the local Gaussian curvature is large, it indicates that the topology structure of the area changes violently. At this time, the control parameter of the chaotic system is increased, so that the chaotic system generates a sequence with more complex changes, to better reflect the fluctuation characteristics of the data in the area. If the local Gaussian curvature is small, that is, the topology structure changes gently, the control parameter is reduced, and the chaotic sequence also changes relatively gently.

[0133] The basic features are fused with the chaotic modulation sequence to generate a primary modulation feature.

[0134] Based on the embodiments provided in the present application, the local Gaussian curvature, geodesic gradient and connection coefficient and other basic features are extracted based on the local geometric properties of the dynamic topological manifold, the chaotic modulation sequence is generated by dynamically adjusting the chaotic system parameter using the local Gaussian curvature, and the basic features are fused with the chaotic modulation sequence to generate a primary modulation feature. This method combines the geometric characteristics of the topology structure with the dynamic characteristics of the chaotic system, so that the basic features and the power distribution network topology structure establish a dynamic mapping relationship. The introduction of the chaotic modulation sequence uses the characteristics of the chaotic system, such as sensitivity to initial conditions and strong ergodicity, to give the feature a more complex variation law, which can more finely describe the subtle fluctuations of the power distribution network operation state, effectively improve the resolution and anti-interference ability of the fault feature, and make the feature more suitable for complex and variable fault scene analysis.

[0135] Further, the method further comprises:

[0136] The primary modulation feature is weighted and normalized based on the value of the local Gaussian curvature, and a modulation feature with topology constraints and chaotic characteristics is output.

[0137] The local Gaussian curvature is used to measure the "local bending degree" of the dynamic topological manifold, which is similar to the steepness of different regions of the earth's surface (for example, the curvature is large at the mountain peak and small at the plain). In the power distribution network, the region with large curvature usually corresponds to the position with complex device connection or large electrical parameter fluctuation (for example, near the hub substation).

[0138] The weighted normalization process includes: weighting: according to the size of the local Gaussian curvature, different weights are given to the primary modulation features. The larger the curvature, the more important the data in this region to fault identification, and the higher the feature weight (for example, the data of the hub substation is more critical than the data of the terminal line). Normalization: the weighted features are uniformly scaled to the range of 0-1 or -1 to 1 to avoid analysis deviation caused by different data dimensions (for example, the numerical range of current data and temperature data is different, and they can be compared fairly after normalization). According to the embodiments provided in the present application, the primary modulation features are weighted and normalized according to the value of the local Gaussian curvature. Since the local Gaussian curvature reflects the bending degree of the local region of the dynamic topological manifold, weighting based on this can make the features of different topological regions obtain reasonable weight distribution in the fusion process. This processing method avoids the feature distortion problem that may be caused by global uniform normalization, retains the constraint of the topological structure on the features, and fuses the dynamic characteristics brought by chaotic modulation, so that the modulation features can better adapt to the spatial distribution difference of the power distribution network topology structure, and ensure that the output features have more accurate expression in the topological space dimension, providing a more balanced and effective feature basis for subsequent multi-dimensional feature fusion.

[0139] Further, based on the enhanced features, the topological space correlation, chaotic behavior similarity and quantum correlation are calculated to construct multi-dimensional fusion weights, including:

[0140] The local Gaussian curvature component is extracted from the enhanced features to calculate the curvature covariance matrix between the switch nodes of the power distribution network.

[0141] The curvature covariance matrix is used to measure the "bending degree" correlation of different nodes in the power distribution network. The local Gaussian curvature values of all switch nodes (such as node A curvature 0.8 and node B curvature 0.3) are extracted; the covariance of the curvatures of any two nodes is calculated to form a matrix. Based on the chaotic modulation sequence, the phase space is reconstructed and the maximum Lyapunov exponent of the current chaotic sequence is calculated.

[0142] The chaotic modulation sequence (one-dimensional time sequence) is expanded into a high-dimensional phase space trajectory through the time delay embedding method to restore the original dynamic system characteristics. Specifically, the Takens embedding theorem is used, the optimal time delay τ is determined according to the autocorrelation function of the sequence, the embedding dimension m is determined by the false nearest neighbor method, and the m-dimensional phase space point set is constructed.

[0143] Based on the reconstructed phase space point set, the Wolf algorithm is used to calculate the adjacent trajectory divergence rate to obtain the maximum Lyapunov exponent λ. The value quantifies the chaos strength of the system: λ>0 indicates that the fault disturbance has chaotic characteristics (such as short-circuit current chaotic oscillation), and its numerical size directly reflects the data fluctuation sensitivity, which is used for subsequent fractal dimension difference analysis.

[0144] The fractal dimension difference between the maximum Lyapunov exponent and the fractal dimension reference value of the historical fault state is compared.

[0145] The fractal dimension reference value is the statistical eigenvalue of the fractal dimension of the chaotic sequence in the historical same fault state, and is usually the arithmetic mean or median of the fractal dimensions of all samples of a specific fault type. When calculating, it is necessary to ensure that the historical data and the current system have the same topological structure and operating conditions.

[0146] Based on the entangled feature vector and the typical fault mode quantum state, the quantum correlation strength parameter is calculated; wherein the typical fault mode quantum state is a standardized quantum state generated by the quantum encoding process of the historical fault data, representing the quantum feature template of typical faults such as short circuit, wire breakage, and ground fault. It is stored in the fault feature library as a reference for quantum correlation strength comparison. For example, the quantum state of single-phase ground fault: zero-sequence current feature encoding; the quantum state of lightning stroke wire breakage fault: sudden voltage waveform encoding.

[0147] The device topological correlation weight is generated according to the curvature covariance matrix;

[0148] The chaotic behavior similarity coefficient is calculated based on the fractal dimension difference;

[0149] The multi-dimensional fusion weight is constructed by fusing the device topological correlation weight, the chaotic behavior similarity coefficient, and the quantum correlation strength parameter.

[0150] Based on the embodiments provided in the present application, the local Gaussian curvature component is extracted from the enhanced features to calculate the curvature covariance matrix, the phase space is reconstructed based on the chaotic modulation sequence, and the fractal dimension difference under the historical fault state is compared, the quantum state similarity analysis is performed on the entangled feature vector to determine the quantum correlation strength parameter, and finally the multi-dimensional fusion weight is generated. This method comprehensively evaluates the importance of features from three dimensions of topological structure, data fluctuation characteristics, and quantum correlation, breaking through the limitations of traditional single-dimensional or few-dimensional weight setting. By quantifying the topological correlation through the curvature covariance matrix, analyzing the data fluctuation similarity using chaos theory, and determining the cross-modal correlation based on quantum state similarity, a complete and comprehensive weight construction system is formed, which can more scientifically and reasonably reflect the contribution of different features in fault diagnosis, and provide accurate weight basis for subsequent feature fusion.

[0151] Further, the enhanced features are weighted and fused according to multi-dimensional fusion weights, and a preprocessed fault feature vector is output, including:

[0152] The enhanced features are mapped to a tangent space of a dynamic topological manifold.

[0153] The dynamic topological manifold is a curved space (such as the surface of the earth), and the tangent space is a "tangent plane" at a point on the manifold (such as the horizontal plane under the feet when standing at a point on the ground, which is the tangent space of the point). Mapping to the tangent space is equivalent to "flattening" the features of the curved space into vectors in a linear space, facilitating weighted calculation.

[0154] In the device connection topology dimension, device topology correlation weights are applied to perform weighted aggregation of adjacent nodes, to generate topological weighted features.

[0155] In the data fluctuation characteristic dimension, chaotic behavior similarity coefficients are used to enhance the amplitude of abnormal fluctuation features, to generate chaotic enhanced features.

[0156] In the cross-modal correlation dimension, quantum correlation strength parameter weighted feature selection is implemented, to generate quantum screened features.

[0157] Based on the embodiments provided in the present application, the enhanced features are mapped to the tangent space of the dynamic topological manifold, device topology correlation weights are applied in the device connection topology dimension to perform weighted aggregation of adjacent nodes, chaotic behavior similarity coefficients are used in the data fluctuation characteristic dimension to enhance the amplitude of abnormal fluctuation features, and quantum correlation strength parameters are used in the cross-modal correlation dimension to perform feature selection. This feature processing method in different dimensions respectively strengthens the topological structure information, data fluctuation features, and cross-modal correlation features, according to different attribute characteristics of the power distribution network fault data. Compared with the traditional simple splicing or unified weighted feature fusion method, this method can avoid information confusion and loss in the feature fusion process, fully utilizes the advantages of features in different dimensions, and makes the generated feature vector more fully and accurately reflect the power distribution network fault state, effectively improving the fault representation capability of the features.

[0158] Further, the method further includes:

[0159] The topological weighted features, chaotic enhanced features, and quantum screened features are fused to form a three-dimensional feature tensor.

[0160] Based on the manifold metric tensor, a Riemann distance between the three-dimensional feature tensor and a historical fault template feature is calculated.

[0161] In an optional implementation, the Riemann distance is calculated based on the following formula:

[0162]

[0163] wherein D Riemannis the Riemannian distance, used to represent the geometric difference between the current feature and the historical fault template in the manifold space; T {3D} is the three-dimensional feature tensor; F {template} is the historical fault template feature; is the manifold metric tensor; T is the matrix transpose operation. The historical fault template feature is a preprocessed standard feature vector of the typical fault data stored in advance. The Riemannian distance is the shortest distance between two points on the Riemannian manifold, similar to the "great circle distance" between two points on the surface of the earth (not the straight-line distance, because the manifold is curved). In the power distribution network, it is used to measure the difference between the current feature and the historical template: the smaller the distance, the more likely it belongs to the same type of fault. When the Riemannian distance is less than the adaptive threshold, the three-dimensional feature tensor is mapped to the fault feature space;

[0164] wherein the adaptive threshold is dynamically set according to the statistical distribution of the Riemannian distance of the historical fault template feature. The specific value is as follows: for the single-phase ground fault type of the 10kV power distribution network, based on 300 historical fault samples, the distance mean is 1.15 and the standard deviation is 0.08, and the threshold is 1.27 by taking 1.5 times the standard deviation. For the 35kV transformer turn-to-turn short circuit fault, based on 200 samples, the mean is 2.4 and the standard deviation is 0.15, and the threshold is 2.58 by taking 1.2 times the standard deviation. The threshold is automatically widened to 1.5 times the standard deviation to reduce the misjudgment rate in the load fluctuation scenario, and is tightened to 1.0 times the standard deviation to improve the sensitivity in the lightning fault scenario. All calculation parameters come from the historical fault template feature library, and the adaptability is maintained through a rolling update mechanism.

[0165] The mapped three-dimensional feature vector is output as the preprocessed fault feature vector.

[0166] Based on the embodiments provided in the present application, the topological weighted feature, the chaos enhanced feature and the quantum screening feature are fused to form a three-dimensional feature tensor, the Riemannian distance between the tensor and the historical fault template feature is calculated based on the manifold metric tensor, and when the distance is less than the adaptive threshold, the fault feature vector is mapped to the fault feature space. The Riemannian distance is calculated using the manifold metric tensor, which fully considers the geometric characteristics of the dynamic topological manifold, and can accurately measure the similarity between the current feature and the historical fault mode from the spatial geometric point of view. The setting of the adaptive threshold can dynamically adjust the feature mapping standard according to the actual scene, so that the output fault feature vector is more consistent with the feature distribution law under the actual fault scene, provides a more targeted and reliable feature expression for subsequent fault diagnosis, and effectively improves the accuracy and applicability of fault diagnosis.

[0167] Optionally, as Figure 3 shown, the present application provides a power distribution network fault data preprocessing system based on multi-source information fusion, comprising:

[0168] The dynamic topological manifold construction module 301 is configured to construct a dynamic topological manifold based on physical connection relationships of power distribution network equipment and multi-source data, the multi-source data including electrical quantity data, meteorological data and equipment state data;

[0169] The abnormal data identification module 302 is configured to identify abnormal data in combination with manifold curvature and equipment connection path distance in response to manifold deformation caused by line switching or load mutation;

[0170] The entangled feature vector generation module 303 is configured to encode the abnormal data, the meteorological data and the equipment state data into quantum states, and generate an entangled feature vector through quantum gate operation;

[0171] The enhanced feature generation module 304 is configured to extract basic features from a local structure of the dynamic topological manifold, generate modulation features after chaotic sequence modulation of the basic features, and generate enhanced features through quantum fusion transformation of the modulation features and the entangled feature vector;

[0172] The multi-dimensional fusion weight construction module 305 is configured to calculate topological space correlation, chaotic behavior similarity and quantum correlation based on the enhanced features, and construct multi-dimensional fusion weights.

[0173] The fault feature vector output module 306 is configured to perform weighted fusion on the enhanced features according to the multi-dimensional fusion weights, and output a preprocessed fault feature vector.

[0174] Reference Figure 3 The power distribution network fault data preprocessing system based on multi-source information fusion of the present application belongs to the technical field of intelligent power distribution system, power distribution switch control equipment manufacturing and the like.

[0175] It should be noted that, in the present application, the embodiments implemented on the side of the power distribution network fault data preprocessing system based on multi-source information fusion can be mutually referenced with the embodiments implemented on the side of the power distribution network fault data preprocessing method based on multi-source information fusion, and the present application will not be described one by one.

[0176] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A power distribution network fault data preprocessing method based on multi-source information fusion, characterized in that, The method comprises the following steps: constructing a dynamic topological manifold based on the physical connection relationship of the power distribution network equipment and multi-source data, wherein the multi-source data comprises electrical quantity data, meteorological data and equipment state data; in response to the manifold deformation caused by line switching or load mutation, identifying abnormal data in combination with the manifold curvature and the equipment connection path distance; encoding the abnormal data, the meteorological data and the equipment state data into quantum states, and generating an entangled feature vector through quantum gate operation; extracting basic features from the local structure of the dynamic topological manifold, generating a modulated feature after chaos sequence modulation of the basic features; and performing quantum fusion transformation on the modulated feature and the entangled feature vector to generate an enhanced feature; calculating the topological space correlation, chaotic behavior similarity and quantum correlation based on the enhanced feature, and constructing a multi-dimensional fusion weight; performing weighted fusion on the enhanced feature according to the multi-dimensional fusion weight, and outputting a preprocessed fault feature vector.

2. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 1, characterized in that, The method of constructing a dynamic topological manifold based on the physical connection relationship of the power distribution network equipment and multi-source data comprises the following steps: obtaining the physical connection relationship between the switch nodes of the power distribution network to form an adjacency matrix for representing the equipment connection topological structure; analyzing the current phase information in the electrical quantity data to calculate the dynamic impedance parameters between nodes; combining the temperature and humidity distribution in the meteorological data to establish a manifold curvature field model; determining the manifold metric tensor according to the insulation aging degree in the equipment state data; fusing the adjacency matrix, the dynamic impedance parameters, the manifold curvature field model and the manifold metric tensor to construct a basic topological manifold through a Riemann manifold embedding algorithm; configuring a dynamic response mechanism for the basic topological manifold to make the basic topological manifold have the characteristics of adaptive deformation with the running state; outputting a dynamic topological manifold that fuses the equipment connection relationship, electrical characteristics and environmental state based on the basic topological manifold.

3. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 2, characterized in that, The method of responding to the manifold deformation caused by line switching or load mutation, in combination with the manifold curvature and the equipment connection path distance to identify abnormal data, comprises the following steps: monitoring the local curvature change rate of the dynamic topological manifold during deformation; calculating the shortest topological path distance between the switch nodes of the power distribution network based on the physical connection relationship; based on the electrical quantity data, generating a characteristic fluctuation sequence reflecting the data fluctuation characteristics using a preset chaos mapping rule, and calculating the Lyapunov exponent of the characteristic fluctuation sequence; calculating the geodesic distance representing the actual connection relationship between the electrical quantity data points in the dynamic topological manifold; calculating the deviation rate of the geodesic distance and the shortest topological path distance between the corresponding nodes; when the local curvature change rate exceeds a first dynamic threshold value and the Lyapunov exponent exceeds a second dynamic threshold value, marking the corresponding area as a candidate abnormal area; in the candidate abnormal area, if the deviation rate continuously exceeds a third dynamic threshold value, the corresponding electrical quantity data point is determined as abnormal data.

4. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 2, characterized in that, The method of encoding the abnormal data, the meteorological data and the equipment state data into quantum states, and generating an entangled feature vector through quantum gate operation, comprises the following steps: Encode the abnormal data, the weather data and the equipment state data into independent quantum states respectively to obtain abnormal data quantum state, weather data quantum state and equipment state quantum state; Perform a double-bit entanglement gate operation on the abnormal data quantum state and the equipment state quantum state to generate an abnormal and equipment entangled state; Take the weather data quantum state as a control bit, and perform a controlled quantum gate operation on the abnormal and equipment entangled state to generate a cross-modal entangled state; Perform a quantum feature transformation operation on the cross-modal entangled state to output an entangled feature vector representing quantum correlation of multi-source data.

5. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 4, characterized in that, The method further comprises: Based on the local geometric properties of the dynamic topological manifold, extract the basic features representing the device connection topology structure, the basic features including local Gaussian curvature, geodesic line gradient and connection coefficient; Use the local Gaussian curvature to dynamically adjust the control parameters of the chaotic system to generate a chaotic modulation sequence associated with the local structure of the dynamic topological manifold; Fuse the basic features and the chaotic modulation sequence to generate a primary modulation feature.

6. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 5, characterized in that, The method further comprises: Based on the value of the local Gaussian curvature, perform weighted normalization processing on the primary modulation feature to output a modulation feature with topological constraints and chaotic characteristics.

7. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 6, characterized in that, The method further comprises: Extract the local Gaussian curvature component from the enhanced feature to calculate the curvature covariance matrix between the switch nodes of the power distribution network; Based on the chaotic modulation sequence, reconstruct the phase space and calculate the maximum Lyapunov exponent of the current chaotic sequence; Compare the maximum Lyapunov exponent with the fractal dimension difference of the historical fault state fractal dimension reference value; Based on the entangled feature vector and the typical fault mode quantum state, calculate the quantum correlation strength parameter; Generate a device topology correlation weight according to the curvature covariance matrix; Calculate the chaotic behavior similarity coefficient based on the fractal dimension difference; Fuse the device topology correlation weight, the chaotic behavior similarity coefficient and the quantum correlation strength parameter to construct a multi-dimensional fusion weight.

8. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 7, characterized in that, The method further comprises: Map the enhanced feature to the tangent space of the dynamic topological manifold; In the device connection topology dimension, apply the device topology correlation weight to perform weighted aggregation of adjacent nodes to generate a topological weighted feature; In the data fluctuation characteristic dimension, use the chaotic behavior similarity coefficient to enhance the amplitude of the abnormal fluctuation feature to generate a chaotic enhanced feature; In the cross-modal correlation dimension, implement feature selection weighted by the quantum correlation strength parameter to generate a quantum screened feature.

9. The power distribution network fault data preprocessing method based on multi-source information fusion according to claim 8, characterized in that, The method further comprises: Fuse the topological weighted feature, the chaotic enhanced feature and the quantum screened feature to form a three-dimensional feature tensor; Based on the manifold metric tensor, calculate the Riemann distance between the three-dimensional feature tensor and the historical fault template feature; map the three-dimensional feature tensor to a fault feature space when the Riemann distance is less than an adaptive threshold value; output the mapped three-dimensional feature vector as the preprocessed fault feature vector.

10. A power distribution network fault data preprocessing system based on multi-source information fusion, characterized in that, comprise: a dynamic topological manifold construction module, configured to construct a dynamic topological manifold based on physical connection relationships of power distribution network equipment and multi-source data, the multi-source data comprising electrical quantity data, meteorological data, and equipment state data; an abnormal data identification module, configured to identify abnormal data in combination with manifold curvature and equipment connection path distance in response to manifold deformation caused by line switching or load mutation; an entangled feature vector generation module, configured to encode the abnormal data, the meteorological data, and the equipment state data into quantum states, and generate an entangled feature vector through quantum gate operation; an enhanced feature generation module, configured to extract a basic feature from a local structure of the dynamic topological manifold, generate a modulated feature after chaotic sequence modulation of the basic feature, and perform quantum fusion transformation on the modulated feature and the entangled feature vector to generate an enhanced feature; a multi-dimensional fusion weight construction module, configured to calculate topological space correlation, chaotic behavior similarity, and quantum correlation based on the enhanced feature, and construct a multi-dimensional fusion weight; a fault feature vector output module, configured to perform weighted fusion on the enhanced feature according to the multi-dimensional fusion weight, and output a preprocessed fault feature vector.

Images