Distribution network power failure risk assessment and early warning method based on multi-source information fusion

Abstract

The invention discloses a distribution network power failure risk assessment and early warning method based on multi-source information fusion, and relates to the technical field of intelligent distribution network risk assessment, and the method comprises the steps: decoding a risk rule set into twinborn parameters through a semantic parser, and outputting a reliability index matrix based on a digital twinborn risk deduction engine; according to the reliability index matrix, a three-pool linkage optimization deduction engine is adopted, and a grading strategy instruction is generated through a particle swarm collaborative optimization mechanism; executing the grading strategy instruction, and generating a new energy fluctuation response quantitative index through a control disturbance-power supply response transfer method; and obtaining a risk assessment early warning result through the new energy fluctuation response quantitative index in combination with a rural distribution network partition risk threshold. According to the invention, accurate quantification and reliable early warning of the power failure risk of the rural power grid scene are achieved by generating the risk assessment early warning result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent distribution network risk assessment, and in particular to a distribution network power failure risk assessment and early warning method based on multi-source information fusion. BACKGROUND

[0002] The distribution network power failure risk assessment and early warning technology based on multi-source information fusion plays a crucial role in the field of intelligent distribution network risk assessment today. With high penetration of new energy into rural power grids and frequent extreme weather events, rural distribution networks face the dual challenges of distributed photovoltaic volatility and spatio-temporal coupling of disaster risks. International standards require the construction of a multi-dimensional dynamic early warning system that integrates weather trajectories, device topology, and photovoltaic state for rural areas. Industry practice shows that spatial correlation analysis of typhoon paths and rural network equipment, as well as quantitative assessment of photovoltaic fluctuations on power failure probability, have become core technical directions supporting planned power failure decisions in rural areas, and related methods have made substantial progress in improving the accuracy of rural power supply reliability index calculations.

[0003] In the field of distribution network power failure risk assessment and early warning based on multi-source information fusion, traditional technical solutions face the problem of static rule base not being able to adapt to real-time fluctuations in new energy output, and the fixed confidence weight generated by historical verification leading to delayed risk response. In addition, the lack of closed-loop verification mechanism for hierarchical strategy instructions and on-site execution results, as well as the cumulative error formed by the deviation between planning strategies and actual operation, restricts the power failure control accuracy in high photovoltaic penetration areas. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a distribution network power failure risk assessment and early warning method based on multi-source information fusion to solve the problems of dynamic response lag of rule base and lack of strategy closed-loop verification.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a distribution network power failure risk assessment and early warning method based on multi-source information fusion, which includes collecting multi-source heterogeneous spatio-temporal data, and obtaining a multi-dimensional data matrix through a spatio-temporal coordinate conversion engine; Based on the multi-dimensional data matrix, a spatio-temporal knowledge graph is constructed through a spatio-temporal double-flow graph neural network, and a risk rule set containing confidence weights is generated according to a rule engine analysis; Through a semantic parser, the risk rule set is decoded into twin parameters, and based on a digital twin risk deduction engine, a reliability index matrix is output; According to the reliability index matrix, a three-pool linkage optimization deduction engine is used to generate hierarchical strategy instructions through a particle swarm cooperative optimization mechanism; The hierarchical strategy instruction is executed to generate a new energy fluctuation response quantitative index through a control disturbance-power response transmission method. Through the new energy fluctuation response quantitative index, in combination with a rural distribution network partition risk threshold, a risk assessment early warning result is obtained.

[0007] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, the multi-source heterogeneous space-time data is collected, a multi-dimensional data matrix is obtained through a space-time coordinate conversion engine, and the steps are as follows, The multi-source heterogeneous space-time data is unified to a WGS84 geographic framework through the space-time coordinate conversion engine to generate standardized space-time data. According to the standardized space-time data, the multi-dimensional data matrix is integrated and reorganized through tensor reorganization operation.

[0008] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, the multi-dimensional data matrix is input into a space-time double-flow graph neural network to perform a space-time double-flow graph neural network graph convolution layer spatial feature aggregation operation and a time feature propagation operation to construct a space-time knowledge graph. According to the fuzzy analytic hierarchy process of the rule engine, the node attributes and edge relationships in the space-time knowledge graph are analyzed to generate an unweighted risk rule set. According to the fuzzy analytic hierarchy process of the rule engine, the node attributes and edge relationships in the space-time knowledge graph are analyzed to generate an unweighted risk rule set. Through historical verification analysis, the unweighted risk rule set is injected with a confidence weight according to the accuracy rate of the historical verification data to output a risk rule set containing a confidence weight.

[0009] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, the risk rule set is decoded into twin parameters through a semantic parser, and a reliability index matrix is output based on a digital twin risk deduction engine, and the steps are as follows, The risk rule set is converted into a parameterized instruction through the semantic parser, and a twin parameter is generated through a physical dimension binding and parameter structure packaging operation. According to the twin parameter, the environment configuration of the digital twin risk deduction engine is initialized, the state evolution process of the distribution network under different boundary conditions is simulated, and a risk deduction result data set is generated. Based on the risk deduction result data set, a multi-dimensional power failure influence weighting and abandoned light loss dynamic proportion quantification method is used to generate a reliability index matrix.

[0010] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, wherein: according to the reliability index matrix, the three-pool linkage optimization deduction engine is adopted, and the steps are as follows, According to the reliability index matrix, the three-pool linkage optimization deduction engine is adopted, and the steps are as follows, Through the particle swarm cooperative optimization mechanism, the three-pool cooperative optimization element data is processed, and the particle position update sequence is obtained.

[0011] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, wherein: through the particle swarm cooperative optimization mechanism, the hierarchical strategy instruction is generated, and the steps are as follows, The non-dominated sorting and congestion distance evaluation operation is performed on the particle position update sequence by adopting the Pareto frontier screening mechanism, and the particle position code with the optimal fitness is output; The particle position code with the optimal fitness is mapped to generate the hierarchical strategy instruction by the instruction conversion rule.

[0012] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, wherein: the hierarchical strategy instruction is executed, and the new energy fluctuation response quantitative index is generated by the control disturbance-power source response transmission method, and the steps are as follows, According to the hierarchical strategy instruction, the operation command is sent to the circuit breaker and the photovoltaic inverter by the rural power distribution network control terminal, and the control disturbance signal is generated by the rural power grid disturbance constraint rule; Based on the control disturbance signal, the signal characteristic is extracted by applying the frequency domain conversion algorithm and the transient response; According to the rule processing signal characteristic, the new energy fluctuation response quantitative index representing the new energy fluctuation characteristic is generated by the fluctuation amplitude calculation rule and the normalization processing method.

[0013] As a preferred scheme of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, wherein: the risk assessment and early warning result is obtained by combining the new energy fluctuation response quantitative index with the rural power distribution network partition risk threshold, and the steps are as follows, Based on the rural power distribution network regional characteristics and the historical early warning event safety boundary benchmark, the rural power distribution network partition risk threshold and the rural power grid risk characteristic are defined by the unit consistency verification mechanism; According to the new energy fluctuation response quantitative index and the rural power distribution network partition risk threshold, the relative deviation amplitude is calculated by the rural power grid dynamic correction factor; The relative deviation amplitude data is executed by the deviation amplitude-early warning level mapping rule, and the risk assessment and early warning result is generated by the early warning level matching and rural power grid risk characteristic fusion operation.

[0014] In a second aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion according to the first aspect of the present application.

[0015] In a third aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion according to the first aspect of the present application.

[0016] The present application has the following beneficial effects: through the three-pool linkage optimization deduction engine, the synergistic optimization elements of the power failure demand pool, the power failure plan pool and the full-amount power failure information pool are fused, the multi-target dynamic optimization and zero-conflict control of the planned power failure demand are realized; at the same time, through the rural power grid risk characteristic fusion operation, the regional vulnerability and exposure degree and other characteristics are quantified into fusion coefficients, the risk assessment and early warning results are generated, the accurate quantification and reliable pre-control of the rural power grid scene power failure risk are achieved, and finally the power failure impact "first calculation and then use" mechanism is formed, which supports the power supply enterprise to optimize the power failure arrangement and improves the user satisfaction. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0018] Fig. 1 The flowchart of the power distribution network power failure risk assessment and early warning method based on multi-source information fusion.

[0019] Fig. 2 The flowchart of the multi-dimensional data matrix generation.

[0020] Fig. 3 The flowchart of the spatio-temporal knowledge graph and risk rule set generation.

[0021] Fig. 4 The flowchart of the twin parameters and reliability index matrix generation. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0023] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0024] Secondly, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. Each of the various embodiments presented in this specification are not necessarily mutually exclusive, but can be combined with each other in various ways. Moreover, each of the embodiments presented in this specification can be combined with each other in various ways.

[0025] Reference Signs List Figs. 1-4 For one embodiment of the present application, the embodiment provides a power distribution network outage risk assessment and early warning method based on multi-source information fusion, comprising the following steps: S1, collect multi-source heterogeneous spatio-temporal data, and obtain a multi-dimensional data matrix through a spatio-temporal coordinate conversion engine.

[0026] Through the spatio-temporal coordinate conversion engine, the multi-source heterogeneous spatio-temporal data is unified to the WGS84 geographic framework to generate standardized spatio-temporal data; It should be noted that the multi-source heterogeneous spatio-temporal data includes typhoon path point cloud of meteorological bureau, power grid GIS device coordinates, photovoltaic inverter operating state, and tree barrier laser radar data. Further, through the spatio-temporal coordinate conversion engine, coordinate system conversion, timestamp alignment, and physical quantity dimension normalization operations are performed to uniformly convert the multi-source heterogeneous spatio-temporal data to standardized spatio-temporal data under the WGS84 geographic framework, generating standardized spatio-temporal data with unified longitude, latitude, elevation reference, and timestamp.

[0027] According to the standardized spatio-temporal data, integration and reorganization are performed through tensor reorganization operation to form a multi-dimensional data matrix.

[0028] Further, according to the device state quantity, photovoltaic operation parameter, tree barrier distance and typhoon threat data, based on the principles of thermodynamics and electromagnetism, the device operation state is quantified; for the photovoltaic operation parameters (such as component voltage, current), the fault arc characteristics are captured in combination with the photoelectric conversion mechanism; for the tree barrier distance data, the wind deflection dynamics model is used to simulate the dynamic safety margin of the conductor and the tree; for the typhoon threat data (such as wind speed, air pressure, moving path), based on fluid mechanics, the path dynamic equation set is established, the typhoon influence intensity is quantified, the time sequence, space and coupling parameters of the data output are mapped to the grid unit according to the IEC 61968 standard, each unit is six-dimensionally structured and coded and a tree barrier-wind speed risk label matrix is generated, through multi-physical field coupling analysis (interactive influence of device-photovoltaic-tree barrier-typhoon), the device temperature rise rate, photovoltaic current 3rd harmonic component, tree barrier safety margin and typhoon velocity divergence parameters are extracted, after range standardization and normalization processing, a quantifiable feature vector and feature dimension which can be directly input into a machine learning model are formed, and based on the standardized space-time data, through tensor reorganization operation, spatial grid discretization, time slot division, multi-source data channel allocation and feature dimension alignment operations are performed, a four-dimensional tensor structure containing time dimension, space dimension, data source channel dimension and feature dimension is constructed, and a multi-dimensional data matrix integrating rural distribution network device state, photovoltaic operation parameter, tree barrier distance and typhoon threat is formed.

[0029] S2, based on the multi-dimensional data matrix, a spatio-temporal knowledge graph is constructed through a space-time double-flow graph neural network, and a risk rule set containing confidence weight is generated according to a rule engine analysis.

[0030] By inputting the multi-dimensional data matrix into the space-time double-flow graph neural network, the spatial feature aggregation operation and the time feature propagation operation of the space-time double-flow graph neural network graph convolution layer are performed, and the spatio-temporal knowledge graph is constructed; Further, based on the historical disaster event device damage, line trip label data, the parameters are optimized through multi-task loss (including classification regression loss, physical constraint regularization term and adjacency topology consistency loss), through the design of time and space double-flow branch architecture, the physical constraint and data-driven are fused to iteratively optimize the parameters to convergence through reverse propagation, the construction of space-time double-flow graph neural network is carried out, through the dynamic construction of adjacency topology by the configurable space relation engine (the geographical radius adjacency is adopted in the typhoon scene, and the vegetation density weighted distance is adopted in the forest fire scene), the spatial neighborhood correlation is obtained based on the physical topology priori (such as the connection relationship of power grid equipment) and data-driven similarity (such as the correlation degree of equipment state in historical disasters); Then the device cluster features (such as the ice thickness of the conductor in the ice disaster, the thermal radiation intensity of the device in the forest fire) are aggregated by the space-time double-flow graph neural network graph convolution layer, and the time dimension data (such as the device temperature rise rate collected every 15 minutes, the photovoltaic current harmonic component, the tree barrier safety margin, the typhoon wind speed observation value) in the multi-dimensional data matrix is processed. The space-time double-flow graph neural network constructs continuous time sequence fragments through time window division, and generates time sequence dynamic features based on time sequence feature engineering (abnormal point detection, trend slope calculation, periodic fluctuation extraction); Through time gating (such as GRU) processing of time sequence of node state (such as change gradient of temperature rise rate within 1 hour, time sequence fluctuation of thermal radiation intensity), the disaster dynamic evolution law such as forest fire spread rate and typhoon wind speed cumulative effect is captured, the time sequence change law depth is obtained based on time sequence dynamic features and time sequence dependent memory (historical state weighted retention), and the spatial neighborhood correlation and time sequence change law depth are fused into unified space-time feature vector through feature splicing or cross attention mechanism. After feature fusion, the unified space-time feature vector is mapped to meteorological events, power equipment, new energy clusters and user load nodes by the universal graph construction engine, and the causal relationship structure between nodes is determined according to the physical mechanism constraint derived based on the basic physical law (such as Kirchhoff's current law constraint and Newton's heat balance equation), and the weighted causal edge is constructed by combining the data-driven risk value (such as the probability of line trip caused by typhoon) learned by the space-time double-flow graph neural network. Through the graph generation algorithm based on rule and neural decoder, the dynamic weighted space-time knowledge graph is output.

[0031] According to the fuzzy analytic hierarchy process of the rule engine, the node attributes and edge relationships in the space-time double-flow space-time knowledge graph are analyzed to generate an unweighted risk rule set. It should be noted that the node attribute refers to the static state characteristics of the meteorological event, power equipment, new energy cluster and user load node in the space-time double-flow knowledge graph, including typhoon center pressure value and transformer aging index, and the edge relationship represents the causal logic chain of dynamic risk transmission between nodes.

[0032] Further, by extracting the node attributes in the spatio-temporal knowledge graph and calling the DL / T 1860 device vulnerability library, the node attributes are fuzzed into membership functions, the edge relationships are parsed, and the FAHP pair comparison matrix is constructed according to the disaster-device association priority criteria, the weights of disaster intensity, device vulnerability and environmental exposure are quantified, and the performance risk weighting coefficient is generated; then the fuzzy node attributes, weighted edge relationships and performance risk weighting coefficient are fused by the rule engine, the weighted rule matching is performed according to the fuzzy logic reasoning engine, and the unweighted risk rule set is generated.

[0033] Through historical verification analysis, the accuracy of the unweighted risk rule set in the historical verification data is injected with confidence weight, and the risk rule set containing confidence weight is output.

[0034] Further, by calling the historical verification data of the State Grid defect event library, the accuracy of the unweighted risk rule set in the spatio-temporal aligned disaster scene is injected with confidence weight, and the rule level weight is dynamically assigned based on historical evidence. Typical disaster event data (such as 200 times from 2019 to 2023, including 120 times of typhoon, 50 times of ice disaster and 30 times of forest fire) are extracted from the State Grid defect event library, each event record contains timestamp, geographic coordinates, device information and disaster parameters, and is associated with device state results (such as line trip, insulator flashover and device overload), and is packaged as an unweighted risk rule set. For the unweighted risk rule set, the historical trigger times and success times are counted. Taking the "typhoon scene line trip rule" as an example, among the 120 typhoon events, the trigger condition of "wind speed > 30 m / s and tree barrier safety margin < 1.2" is met 58 times, of which 49 times actually occur line trip, and the historical accuracy is derived as 84.5%. The risk rule set weight is dynamically assigned based on the historical accuracy, and the risk rule set containing confidence weight with traceability path is obtained.

[0035] S3, through the semantic parser, the risk rule set is decoded into twin parameters, and based on the digital twin risk deduction engine, a reliability index matrix is output.

[0036] Through the semantic parser, the risk rule set is converted into parameterized instructions, and through the physical dimension binding and parameter structure packaging operation, the twin parameters are generated; It should be noted that the twin parameters are structured data interfaces in the digital twin system, which contain physical quantity types, numerical thresholds, unit dimensions, weight coefficients and rule bases, realizing the computable conversion of risk rules to digital twin engine.

[0037] Furthermore, the risk rule set is extracted using natural language processing (NLP) technology in the semantic parser to identify physical quantity types (such as wind speed and tree obstacle safety margin), numerical thresholds, logical relationships, and scenario limitations. Then, the parsed physical quantities are standardized, calibrated, and bound using a physical quantity binding engine combined with international standard measurement systems (such as the SI unit system). The bound physical quantity types, numerical thresholds, logical relationships, and scenario limitations are encapsulated into machine-executable parameterized instructions, which are then converted into twin parameters through the interface adaptation module of the digital twin engine.

[0038] The environment configuration of the digital twin risk simulation engine is initialized based on the twin parameters. By simulating the evolution of the distribution network state under different boundary conditions, a risk simulation result dataset is generated. It should be noted that environmental configuration refers to the dynamic initialization process of the digital twin risk simulation engine for real-time monitoring of data streams, equipment limit parameters, and disaster boundary constraints.

[0039] Furthermore, by leveraging the power grid enterprise's Geographic Information System (GIS) and distribution network engineering archives, equipment ledger systems (such as PMS2.0) and equipment factory inspection reports, data interfaces with the National Meteorological Administration and regional meteorological and hydrological monitoring stations, and power grid operation management systems (such as SCADA) and historical fault databases, the rural distribution network topology, equipment physical parameters (such as transformer capacity and conductor impedance), real-time meteorological and hydrological data, and historical operating baselines are obtained. Combined with twin parameters, the initialization environment configuration for the digital twin risk simulation engine is performed. Through real-time model fusion and physical constraint verification mechanisms, a physically accurate virtual simulation scenario is constructed, and boundary value constraint simulation is used. The evolution process under disaster boundary conditions is iteratively calculated to solve the chain reaction of equipment failure, dynamically generating multi-dimensional risk quantification results in spatiotemporal space to obtain risk values. Based on the risk threshold and warning threshold of the DL / T power grid safety early warning guidelines, risk level coloring is implemented. When the risk value is ≥ the risk threshold, a red warning is triggered (requiring emergency response, such as cutting off power supply to high-risk tree-blocked sections). When the risk value is between the risk threshold and the warning, a yellow warning is initiated (implementing preventive control, such as switching reactive power devices). When the risk value is < the warning, a green safe state is marked (monitoring only, no intervention). Based on the spatiotemporal event coordinate binding and risk propagation path marking, a risk inference result dataset with multi-dimensional labels is generated.

[0040] Based on the risk simulation results dataset, a reliability index matrix is ​​generated by using a multi-dimensional weighted average of power outage impacts and a dynamic proportional quantification method for curtailment losses.

[0041] It should be noted that the multi-dimensional power outage impact weighting refers to a standardized rule that weights and aggregates the number of affected users in the projection results based on the social value weight of user type and the economic loss of power outage.

[0042] Further, based on the risk deduction result data set, first, the multi-dimensional power failure influence weighted method is used to aggregate user influence range data to generate a power failure comprehensive influence index; simultaneously, the dynamic proportion quantification method of light abandonment loss is used to output a two-dimensional decision table containing time and space risk linkage parameters, with the row dimension being the disaster evolution time step and the column dimension being the reliability index matrix of the regional / equipment type combination.

[0043] S4, according to the reliability index matrix, a three-pool linkage optimization deduction engine is used to generate a hierarchical strategy instruction through a particle swarm cooperative optimization mechanism.

[0044] According to the reliability index matrix, the three-pool cooperative optimization element data in the power failure demand pool, the power failure plan pool and the full-amount power failure information pool are loaded by the three-pool linkage optimization deduction engine; Further, according to the reliability index matrix, the predicted load gap distribution data in the power failure demand pool (such as the irrigation peak period and the township health clinic weight factor) are pre-fetched by the three-pool linkage optimization deduction engine; the active operation and maintenance strategy of the power failure plan pool is aligned; and the rural power grid fault mode library of the full-amount power failure information pool is associated, and the three-pool linkage optimization deduction engine performs cross-pool time and space matching and cost optimization, dynamically and iteratively calculates the optimal power failure scheduling plan (objective function: minimization of weighted power failure comprehensive influence index + light abandonment loss penalty term) with the rural reliability index matrix as the constraint condition, and finally generates a three-dimensional optimization scheme integrating the predicted demand, active intervention and historical experience, and generates three-pool cooperative optimization element data.

[0045] Through the particle swarm cooperative optimization mechanism, the three-pool cooperative optimization element data is processed to obtain a particle position update sequence; Further, the predicted load gap distribution of the power failure demand pool, the active operation and maintenance strategy parameters of the power failure plan pool and the historical fault mode of the full-amount power failure information pool are mapped into a particle position vector, and the reliability index matrix is taken as the core basis for judging the pros and cons of the particle position; according to the predicted load gap distribution and the historical fault mode, based on historical statistics and expert experience, the weight coefficients between 0 and 1 are allocated to time, region and user type, and the weighted sum formula is used to calculate the weighted power failure index which comprehensively reflects the power failure influence, combined with the light abandonment penalty term obtained by quantifying the economic loss of insufficient photovoltaic consumption by the active operation and maintenance strategy parameters, to generate a target value. In each iteration, the group optimal position and individual historical optimal position are obtained by comparing the target value, the particle position is updated by the standard particle swarm velocity-position update method, and a position update sequence containing time period-region-decision variable combination is generated, which records the optimization trajectory and converges to the optimal scheduling strategy.

[0046] The particle position update sequence is executed by the Pareto front screening mechanism to perform non-dominated sorting and congestion distance evaluation operations, and the particle position code with the optimal fitness is output; Further, the particle position update sequence is processed by using a Pareto frontier screening mechanism, non-dominated sorting is used to identify a non-inferior particle set (i.e. any particle cannot be comprehensively surpassed by other particles in the three-dimensional target of power loss, light penalty and treatment cost) that meets the multi-objective equilibrium solution, and then high-quality particles with sparse distribution are screened according to the congestion distance evaluation, and the particle position coding with the optimal fitness is obtained according to the minimum Manhattan distance-entropy weight comprehensive decision method.

[0047] The particle position coding with the optimal fitness is mapped by using an instruction conversion rule to generate hierarchical strategy instructions.

[0048] It should be noted that the instruction conversion rule refers to a hierarchical triggering mechanism based on the risk threshold and the early warning threshold, and the particle position coding value is dynamically bound to the operation mapping protocol of the distribution network operation specification; the hierarchical strategy instruction refers to a structured operation command divided into three risk levels of red, yellow and green, and the content is generated by the particle position parameter combined with the risk threshold and the early warning threshold determination result, and has four-tuple attributes of response level, execution subject, time limit requirement and standard basis.

[0049] Further, the particle position coding with the optimal fitness is mapped to hierarchical strategy instructions by using the instruction conversion rule, and three-level instruction differentiation is realized according to the risk threshold and the early warning threshold, when the particle parameter triggers the risk threshold, the red instruction (such as red, immediately remove the high-risk branch line power supply, and simultaneously start the standby power supply to protect the irrigation) is generated by calling the DL / T599 emergency treatment clause, when the particle parameter is in the early warning interval, the yellow instruction (such as yellow, transfer part of the load to the substation B within 2 hours) is generated according to the particle position coding value matching the power failure plan pool plan, and when the parameter is lower than the early warning threshold, the green monitoring instruction (such as green, real-time monitoring of photovoltaic voltage) is output, and finally the hierarchical strategy instruction is formed by integrating the risk level, the particle optimization parameter and the distribution network operation specification.

[0050] S5, execute the hierarchical strategy instruction, generate a new energy fluctuation response quantitative index by using a control disturbance-power source response transmission method.

[0051] According to the hierarchical strategy instruction, the control terminal of the rural distribution network drives the operation command to be sent to the circuit breaker and the photovoltaic inverter, and generates a control disturbance signal by using the rural network disturbance constraint rule; It should be noted that the rural network disturbance constraint rule refers to a voltage, frequency and harmonic instantaneous disturbance quantitative boundary set established based on the DL / T regulation and the historical disturbance library, the disturbance constraint threshold is derived by statistical analysis of typical disturbance events of the rural network, and the compliance label rule library is generated based on the disturbance constraint threshold to ensure that the control operation does not cause secondary risk.

[0052] Furthermore, based on the hierarchical strategy instructions, the rural power distribution network control terminal executes closed-loop operations. A red instruction triggers the circuit breaker to trip within milliseconds and forces the photovoltaic inverter to switch off-grid mode; a yellow instruction executes load transfer operations and adjusts the inverter to reduce output; a green instruction maintains the circuit breaker in its normally closed state and the inverter automatically stabilizes its voltage. Simultaneously, following rural power grid disturbance constraint rules, control disturbance signals are dynamically generated by real-time monitoring of parameters such as voltage change rate and harmonic distortion rate.

[0053] Based on the control disturbance signal, frequency domain transformation algorithm and transient response are applied to extract the signal features for rule processing; Furthermore, feature extraction is performed based on control disturbance signals (such as voltage surges caused by circuit breaker tripping and frequency fluctuations caused by photovoltaic grid disconnection). Real-time measured parameters are collected by multi-source monitoring equipment in rural power distribution networks. The Fast Fourier Transform (FFT) frequency domain conversion algorithm is used to analyze the signal spectrum distribution. Transient response analysis is performed simultaneously to quantify time-scale parameters. At the same time, high-frequency harmonic components and low-frequency power oscillation components are separated by multi-scale wavelet decomposition. Combined with rural power grid disturbance constraint rules, constraint threshold matching and compliance label binding are performed to generate signal features with rule processing labels.

[0054] Based on the signal characteristics processed by the rules, and through the fluctuation amplitude calculation rules and normalization method, a quantitative index of new energy fluctuation response that characterizes the fluctuation characteristics of new energy is generated.

[0055] Furthermore, based on the structural characteristics of rural power distribution networks (such as long lines with high impedance and distributed photovoltaic penetration) and fault statistics, feature weights for rural power distribution network scenarios are obtained. According to the rural power distribution network disturbance constraint rules, the measured parameters in the signal feature vector are dynamically matched with the power distribution network operation specifications. Based on the deviation direction and amplitude, compliance, critical and over-limit states are marked and bound to the corresponding industry standard clauses. At the same time, rural power distribution network scenario feature weights are superimposed to output a structured feature vector with rule processing labels. Through the new energy fluctuation characteristic adaptation algorithm, the fluctuation amplitude calculation rules are executed to analyze the feature value deviation. Simultaneously, the range normalization method is used to eliminate dimensional differences. The normalization processing results and fluctuation frequency parameters (over-limit fluctuation frequency parameters extracted from SCADA sampling sequences) are integrated through the new energy fluctuation aggregation formula. The weighted Mahalanobis distance-risk probability coupling algorithm is used to generate a quantitative index of new energy fluctuation response.

[0056] S6. By using quantitative indicators of new energy fluctuation response and combining them with risk thresholds for rural power distribution network zones, risk assessment and early warning results can be obtained.

[0057] Based on the regional characteristics of rural power distribution networks and the safety boundary benchmark of historical early warning events, a unit consistency verification mechanism is used to define the risk threshold of rural power distribution network zones and the risk characteristics of rural power distribution networks.

[0058] It should be noted that the risk characteristics of rural power grids refer to the systematic vulnerability caused by unique factors of rural power grids, which is characterized by low short-circuit capacity and high fault propagation rate.

[0059] Further, based on the regional characteristics of rural power grids and the historical early warning event library, first, the same regional and disaster type events (such as tree barrier pressure lines in typhoon scenarios and immersion failures during floods) are screened, and the safety boundary benchmark dataset is constructed through the key feature space dimension reduction and clustering alignment method. According to the ratio of photovoltaic installation capacity to maximum load of the area, combined with the frequency of photovoltaic off-grid in historical disasters, the risk coefficient is corrected to obtain the correction coefficient of the penetration rate of distributed photovoltaic in rural power grids. According to the unit consistency verification mechanism, the historical safety boundary extreme value (such as 95% quantile as the risk threshold benchmark) is calculated by quantile regression method, and the risk threshold is obtained after superimposing the penetration rate correction coefficient. At the same time, it is quantified as a risk adjustment parameter through statistical analysis (such as the probability distribution of voltage sag caused by photovoltaic off-grid) to generate the risk characteristics of rural power grids.

[0060] According to the new energy fluctuation response quantitative index and the rural distribution network partition risk threshold, the relative deviation amplitude is calculated through the dynamic correction factor of rural power grids. Further, through real-time monitoring devices (such as PMU), photovoltaic output rate, fluctuation amplitude, and multi-energy complementary system cooperative response time data are collected. Through the new energy fluctuation response quantitative index and the rural distribution network partition risk threshold, combined with the dynamic correction factor obtained by dynamically adjusting the regional difference factors, the relative deviation amplitude is calculated , the expression is: ; Among them, represents the relative deviation amplitude, represents the dynamic correction factor, represents the new energy fluctuation response quantitative index, which is the normalized measurement of the fluctuation intensity of new energy stations. The larger the value, the more intense the fluctuation, represents the rural distribution network partition risk threshold.

[0061] Through the deviation amplitude-early warning level mapping rule, the relative deviation amplitude is matched with the early warning level and the risk characteristics of rural power grids are fused to generate the risk assessment early warning result.

[0062] Further, according to the relative deviation amplitude (δ) value and the rural power grid early warning mapping rule (such as δ>20% for red, 10%<δ≤20% for yellow, and δ≤10% for green), the calculated value is matched to the red, yellow, and green early warning levels, and the three-dimensional risk assessment result is generated by fusing the unique risk characteristics of rural power grids (radiation network frame fault diffusion speed increases, tree barrier density increases breakdown probability, and distributed photovoltaic penetration rate causes voltage fluctuation): The red early warning triggers emergency response measures, generates a high-risk area power-off instruction in combination with the radial network framework characteristics of the rural power grid, forcibly cuts off high-risk branch lines at the second level, and starts the evacuation program; The yellow early warning performs preventive regulation, generates a load transfer + photovoltaic capacity reduction instruction in combination with the high photovoltaic penetration characteristics of the rural power grid, and requires part of the load to be transferred and the photovoltaic output to be reduced within 2 hours.

[0063] The green early warning starts an enhanced monitoring mode, generates a dynamic monitoring instruction in combination with the low short-circuit capacity characteristics of the rural power grid, and improves the monitoring frequency to 3 times that of the urban power grid (for example, data is collected every 5 minutes).

[0064] The embodiment also provides a computer device suitable for the power distribution network power failure risk assessment and early warning method based on multi-source information fusion, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the power distribution network power failure risk assessment and early warning method based on multi-source information fusion proposed in the above embodiment.

[0065] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can also be an external keyboard, touchpad or mouse, etc.

[0066] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for implementing power distribution outage risk assessment and early warning based on multi-source information fusion proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0067] To sum up, the application realizes multi-target dynamic optimization and zero conflict control of planned power outage demand by optimizing and deducing the synergistic optimization elements of the power outage demand pool, the power outage plan pool, and the full-amount power outage information pool through the three-pool linkage; meanwhile, the regional vulnerability and exposure degree and other characteristics are quantified into fusion coefficients through the operation of the rural power grid risk characteristic fusion, to generate a risk assessment and early warning result, achieve accurate quantification and reliable pre-control of the rural power grid scene power outage risk, and finally form a power outage impact "calculate first and then use" mechanism to support power supply enterprises to optimize power outage arrangement and improve user satisfaction.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application but not limit the application, and although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. A method for power outage risk assessment and early warning in distribution networks based on multi-source information fusion, characterized in that: include, Collect multi-source heterogeneous spatiotemporal data and obtain a multi-dimensional data matrix through a spatiotemporal coordinate transformation engine; Based on a multidimensional data matrix, a spatiotemporal knowledge graph is constructed using a spatiotemporal dual-flow graph neural network, and a risk rule set containing confidence weights is generated based on the parsing of the rule engine. The risk rule set is decoded into twin parameters by a semantic parser, and a reliability index matrix is ​​output based on the digital twin risk inference engine. Based on the reliability index matrix, a three-pool linkage optimization and deduction engine is adopted, and a hierarchical strategy instruction is generated through a particle swarm collaborative optimization mechanism. The hierarchical strategy instructions are executed, and quantitative indicators of new energy fluctuation response are generated through the control disturbance-power response transfer method. By using quantitative indicators of new energy fluctuation response and combining them with risk thresholds for rural power distribution network zones, risk assessment and early warning results are obtained.

2. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The process of collecting multi-source heterogeneous spatiotemporal data and obtaining a multidimensional data matrix through a spatiotemporal coordinate transformation engine is as follows: The spatiotemporal coordinate transformation engine unifies multi-source heterogeneous spatiotemporal data into the WGS84 geographic framework, generating standardized spatiotemporal data. Based on standardized spatiotemporal data, tensor recombination operations are used to integrate and recombine the data to form a multidimensional data matrix.

3. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The process involves constructing a spatiotemporal knowledge graph based on a multidimensional data matrix using a spatiotemporal dual-flow graph neural network, and generating a risk rule set containing confidence weights through rule engine parsing. The steps are as follows: By inputting a multidimensional data matrix into a spatiotemporal dual-flow graph neural network, spatial feature aggregation and temporal feature propagation operations are performed on the graph convolutional layers of the spatiotemporal dual-flow graph neural network to construct a spatiotemporal knowledge graph. Based on the fuzzy hierarchical analysis method of the rule engine, the node attributes and edge relationships in the spatiotemporal knowledge graph are analyzed to generate an unweighted risk rule set; Through historical validation analysis, based on the accuracy of the unweighted risk rule set in historical validation data, confidence weights are injected, and a risk rule set containing confidence weights is output.

4. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The steps are as follows: decoding the risk rule set into twin parameters using a semantic parser, and outputting a reliability index matrix based on a digital twin risk inference engine. The risk rule set is converted into parameterized instructions through a semantic parser, and twin parameters are generated through physical dimension binding and parameter structured encapsulation operations. The environment configuration of the digital twin risk simulation engine is initialized based on the twin parameters. By simulating the evolution of the distribution network state under different boundary conditions, a risk simulation result dataset is generated. Based on the risk simulation results dataset, a reliability index matrix is ​​generated by using a multi-dimensional weighted average of power outage impacts and a dynamic proportional quantification method for curtailment losses.

5. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The steps for using a three-pool linkage optimization and inference engine based on the reliability index matrix are as follows: Based on the reliability index matrix, the collaborative optimization element data of the three pools (power outage demand pool, power outage plan pool, and full power outage information pool) are loaded through the three-pool linkage optimization simulation engine. By using a particle swarm optimization mechanism, the data of the three pools for collaborative optimization are processed to obtain the particle position update sequence.

6. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 5, characterized in that: The generation of hierarchical strategy instructions through the particle swarm cooperative optimization mechanism is described in the following steps. The Pareto front screening mechanism is used to perform non-dominated sorting and crowding distance evaluation on the particle position update sequence, and output the particle position code with the best fitness. By using instruction conversion rules, the particle position encoding with optimal fitness is mapped to generate hierarchical strategy instructions.

7. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The execution of the hierarchical strategy instructions, through the control disturbance-power response transfer method, generates a quantitative index of new energy fluctuation response. The steps are as follows: According to the hierarchical strategy instructions, the rural power distribution network control terminal is driven to send operation commands to the circuit breaker and photovoltaic inverter, and control disturbance signals are generated through the rural power grid disturbance constraint rules. Based on the control disturbance signal, frequency domain transformation algorithm and transient response are applied to extract the signal features for rule processing; Based on the signal characteristics processed by the rules, and through the fluctuation amplitude calculation rules and normalization method, a quantitative index of new energy fluctuation response that characterizes the fluctuation characteristics of new energy is generated.

8. The distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in claim 1, characterized in that: The steps for obtaining risk assessment and early warning results by using quantitative indicators of new energy fluctuation response combined with risk thresholds for rural power distribution networks are as follows. Based on the regional characteristics of rural power distribution networks and the safety boundary benchmark of historical early warning events, a unit consistency verification mechanism is used to define the risk threshold of rural power distribution network zones and the risk characteristics of rural power distribution networks. Based on the quantitative indicators of new energy fluctuation response and the risk threshold of rural power grid zones, the relative deviation is calculated through the dynamic correction factor of the rural power grid. By using the deviation magnitude-warning level mapping rule, the relative deviation magnitude is matched with the warning level and integrated with the risk characteristics of the rural power grid to generate risk assessment and warning results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the distribution network power outage risk assessment and early warning method based on multi-source information fusion as described in any one of claims 1 to 8.

Images