A multi-level fuzzy comprehensive evaluation method based on power distribution network security situation

Abstract

The application discloses a kind of based on power distribution network security situation Multistage fuzzy comprehensive evaluation method, it is related to power grid safety evaluation technical field, including acquisition power distribution network real-time operation data, environmental information and topological structure, combine history operation sequence to construct time series prediction model, output contains the dynamic safety margin sequence of key security index;Based on dynamic safety margin sequence, time attenuation cumulative operator is applied to the historical fuzzy membership degree of bottom layer security index, and the cumulative membership vector representing risk exposure intensity is generated;Call pre-built power distribution network knowledge graph, according to the semantic scene recognition of current operating state of power distribution network and the dynamic safety margin sequence, generate the multistage safety evaluation index subset adapted to current running context;Cumulative membership vector is mapped to the bottom layer security index of the multistage safety evaluation index subset, and the membership function with adjustable boundary parameter is constructed, and the fuzzy membership vector of bottom layer security index under preset comment set is output.

Description

Technical Field

[0001] This invention relates to the field of power grid security assessment technology, and in particular to a multi-level fuzzy comprehensive assessment method based on the security status of distribution networks. Background Technology

[0002] Power grid safety assessment technology refers to a technical system that uses real-time operating data, equipment status information, and external environmental parameters of the power system, combined with mathematical models, artificial intelligence algorithms, or expert rules, to quantitatively assess, identify risks, and classify the safety operation level of the power grid in the current and future periods.

[0003] Existing assessment methods generally use fixed thresholds as safety criteria, ignoring the real-time impact of factors such as fluctuations in new energy output, rapid load changes, and sudden changes in weather conditions on the actual safety margin of equipment. This leads to a large number of false alarms or missed alarms in high-penetration distributed power sources or under extreme weather conditions. Furthermore, traditional fuzzy comprehensive assessment is based solely on the membership degree calculated from the current instantaneous measurement, which cannot distinguish between short-term overruns and continuous high-voltage / high-temperature operation. It also lacks the ability to perceive hidden risks with time-cumulative characteristics, such as transformer thermal aging and cable insulation deterioration, resulting in a disconnect between safety situation assessment and physical reality. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-level fuzzy comprehensive evaluation method based on the safety status of distribution networks. This solves the problem that existing evaluation methods generally use fixed thresholds as safety criteria, ignoring the real-time impact of factors such as fluctuations in new energy output, rapid load changes, and sudden changes in weather conditions on the actual safety margin of equipment. This leads to a large number of false alarms or missed alarms in high-penetration distributed power sources or extreme weather conditions. Furthermore, traditional fuzzy comprehensive evaluation only calculates membership based on current instantaneous measurements, which cannot distinguish between short-term overruns and continuous high-voltage / high-temperature operation. It also lacks the ability to perceive hidden risks with time-cumulative characteristics, such as transformer thermal aging and cable insulation deterioration, resulting in a disconnect between safety status assessment and physical reality.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a multi-level fuzzy comprehensive evaluation method based on the security status of a power distribution network, comprising:

[0008] Collect real-time operation data, environmental information and topology of the power distribution network, combine them with historical operation sequences to build a time series prediction model, and output a dynamic safety margin sequence containing key safety indicators;

[0009] Based on the dynamic safety margin sequence, a time decay accumulation operator is applied to the historical fuzzy membership degree of the underlying safety indicators to generate a cumulative membership vector characterizing the intensity of risk exposure.

[0010] The pre-built distribution network knowledge graph is invoked, and semantic scene recognition is performed based on the current operating status of the distribution network and the dynamic safety margin sequence to generate a multi-level safety assessment index subset adapted to the current operating context.

[0011] The cumulative membership vector is mapped to the underlying security index of the multi-level security assessment index subset, a membership function with adjustable boundary parameters is constructed, and the fuzzy membership vector of the underlying security index under the preset comment set is output.

[0012] Information entropy is calculated based on fuzzy membership vectors, and dynamic combined weights are generated by integrating subjective weights determined by the analytic hierarchy process and topological sensitivity factors derived from power flow sensitivity.

[0013] Based on the physical coupling relationship between dynamic combined weights and underlying security indicators, a non-additive fuzzy measure is constructed. Choquet fuzzy integrals are used to aggregate the fuzzy membership vectors level by level to obtain the highest-level comprehensive fuzzy evaluation vector.

[0014] The comprehensive fuzzy evaluation vector is mapped to the ideal safety state, the distribution network safety status index is calculated, and a safety status assessment result containing risk level and weak links is generated. At the same time, the safety status assessment result is compared with the subsequent actual operation events to extract misjudged samples for updating the time series prediction model, membership function parameters, time decay accumulation operator and fuzzy measure, thus completing the evaluation model update.

[0015] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the step of collecting real-time operating data, environmental information, and topology of the distribution network, constructing a time-series prediction model by combining historical operating sequences, and outputting a dynamic security margin sequence of key security indicators includes:

[0016] The real-time measurement sequences of voltage amplitude, branch current, active power and reactive power of each node in the distribution network are obtained from the dispatch automation system.

[0017] Synchronously connect to the meteorological interface to obtain ambient temperature, humidity and wind speed;

[0018] Extract the current network connectivity matrix and device rated parameter vector using a geographic information system;

[0019] The electrical measurement sequences from the dispatch automation system, the environmental parameter sequences from the meteorological interface, and the static topology and equipment parameters provided by the geographic information system are aligned and fused according to a unified timestamp to form a multi-source time series;

[0020] After aligning multi-source time series data with a unified timestamp, the data is input into a long short-term memory neural network. Using observations from the past N time steps as input, the model is trained to predict the future. Safety margins of key safety indicators within the step;

[0021] Suppose a key runtime variable is at time 10:00. The predicted value is Its corresponding limit is The dynamic safety margin is defined as follows:

[0022] ;

[0023] in, Indicates time The safety margin, with a larger value indicating that it is further away from the boundary. These are the predicted values ​​of key running variables output by the time series forecasting model. The upper or lower limit of the operation of this variable is determined by the equipment nameplate parameters or scheduling procedures;

[0024] After the model converges, it outputs a dynamic safety margin sequence. .

[0025] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the step of applying a time-decaying cumulative operator to the historical fuzzy membership degree of the underlying security indicators based on the dynamic security margin sequence to generate a cumulative membership vector characterizing the intensity of risk exposure includes:

[0026] For the Historical membership sequence of each underlying security indicator under the security rating level An exponential decay weighting mechanism is introduced to calculate the cumulative membership degree. :

[0027] ;

[0028] Among them, the attenuation weight Defined as:

[0029] ;

[0030] in, Indicates the first The risk exposure intensity of each underlying security indicator For the first The degree to which this indicator belongs to the security level at a given historical moment. The length of the sliding time window. The attenuation coefficient;

[0031] All underlying security metrics Constructing cumulative membership vector , as an enhanced representation of the fuzzy input layer.

[0032] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the step of calling a pre-built distribution network knowledge graph, performing semantic scene recognition based on the current operating state of the distribution network and the dynamic security margin sequence, and generating a subset of multi-level security assessment indicators adapted to the current operating context includes:

[0033] Input the current topology connection matrix, distributed power grid connection flag, microgrid operation mode identifier, and load type code as query conditions into the knowledge graph reasoning engine;

[0034] The knowledge graph stores the relationships between pre-existing runtime scenario nodes and evaluation dimension nodes;

[0035] By using graph embedding similarity matching, the scene node with the highest cosine similarity to the current query vector is selected;

[0036] The set of primary indicators associated with the scene node, the subset of secondary indicators corresponding to each primary indicator, and the mapping relationship between the underlying observable variables and measurement points are activated to form a multi-level safety assessment indicator subset with a variable structure.

[0037] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the step of mapping the cumulative membership vector to the underlying security index of the subset of multi-level security evaluation indicators, constructing a membership function with adjustable boundary parameters, and outputting the fuzzy membership vector of the underlying security index under a preset evaluation set includes:

[0038] For the Each underlying security indicator is described by a piecewise exponential membership function, which represents its degree of membership to the security level.

[0039] ;

[0040] Among them, the lower limit and upper limit From dynamic safety margin sequence Real-time correction, the expression is:

[0041] ;

[0042] in, Indicates running value Membership degree to security level and These represent the left and right boundaries of the membership function, which dynamically shrink or expand with the safety margin. For the first The rated limit of each indicator, This is the margin scaling factor, used to adjust the boundary sensitivity. and These are shape adjustment parameters that control the steepness of the membership curve.

[0043] The cumulative membership vector Considering it as an equivalent safety membership degree, we inversely deduce its corresponding operating state, calculate its membership degree under warning and danger levels, and form a ternary fuzzy membership vector. .

[0044] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the step of calculating information entropy based on fuzzy membership vectors, fusing subjective weights determined by the analytic hierarchy process (AHP) with topology sensitivity factors derived from power flow sensitivity, and generating dynamic combined weights includes:

[0045] For the Fuzzy membership vectors of each indicator First, normalize to get:

[0046] ;

[0047] Calculate the information entropy again The expression is:

[0048] ;

[0049] This leads to the objective weight. The expression is:

[0050] ;

[0051] Meanwhile, subjective weights are obtained by constructing a judgment matrix by experts and then undergoing a consistency test. ;

[0052] Calculate the norm of the partial derivative of the power flow Jacobian matrix with respect to the node voltages, and use it as the topology sensitivity factor. ;

[0053] Introducing a Coordination Coefficient Generate dynamic combined weights, the expression is:

[0054] ;

[0055] in, Indicates the first The first indicator in the Normalized membership degree under each rating level Information entropy reflects the amount of effective information provided by the indicator. The lower the entropy, the higher the indicator's discriminative power. To provide objective weights based on data uncertainty, Subjective weighting based on expert experience, The topology sensitivity factor quantifies the structural importance of a device in the current network. This is a coordination coefficient used to balance the weighting of subjective and objective factors. These are the dynamic combination weights used for fuzzy synthesis.

[0056] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the method involves: constructing a non-additive fuzzy measure based on the physical coupling relationship between dynamic combined weights and underlying security indicators; and using Choquet fuzzy integrals to aggregate the fuzzy membership vectors level by level to obtain the highest-level comprehensive fuzzy evaluation vector, including:

[0057] For index pairs that are coupled, an interactive metric is defined based on their joint historical frequency of exceeding limits. ;

[0058] For each secondary indicator group Arrange their fuzzy membership vectors in descending order of membership degree:

[0059] ;

[0060] The corresponding indicator index set is denoted as ;

[0061] The second-order integrated membership degree is calculated using Choquet integral, and the expression is as follows:

[0062] ;

[0063] in, , Indicates the first The comprehensive fuzzy membership degree of each secondary indicator. For the sorted number Large membership values For the front A set consisting of high membership indices It is a fuzzy measure of the set, reflecting the overall risk level under the synergistic effect of indicators;

[0064] The repeated process aggregates upwards to the first-level indicator layer, resulting in a comprehensive fuzzy evaluation vector. .

[0065] As a preferred embodiment of the multi-level fuzzy comprehensive evaluation method based on the security status of distribution networks described in this invention, the following steps are included: mapping the comprehensive fuzzy evaluation vector to an ideal security state, calculating the distribution network security status index and generating a security status assessment result containing risk levels and weak links; simultaneously comparing the security status assessment result with subsequent actual operational events to identify and extract misjudged samples for updating the time series prediction model, membership function parameters, time decay accumulation operator, and fuzzy measure, thereby completing the evaluation model update, including:

[0066] Define the ideal safety state vector This indicates that the system is at a completely safe level.

[0067] Calculate the comprehensive fuzzy evaluation vector and The weighted Hamming distance is expressed as:

[0068] ;

[0069] in, The combined weights of the primary indicators;

[0070] The distribution network security status index is defined as follows:

[0071] ;

[0072] in, This indicates the degree of deviation between the current state and the ideal safe state. The final output is a safety score;

[0073] After the evaluation period ends, the comparison The warning level is related to whether a protective action, equipment tripping, or user complaint actually occurred.

[0074] like ≥80 but a fault occurs, mark as a missed alarm sample;

[0075] like ≤40 but no events are marked as false alarm samples;

[0076] Add the input sequence within the time window corresponding to the missed alarm sample to the training set to trigger incremental training of the LSTM model;

[0077] At the same time, adjust the attenuation coefficient of the corresponding indicator. Membership function shape parameters , And fuzzy measures enable the updating of parameters across the entire link.

[0078] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network as described in the first aspect of the present invention.

[0079] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network as described in the first aspect of the present invention.

[0080] The beneficial effects of this invention are as follows: by integrating time-series prediction-driven dynamic safety margin modeling, time-accumulated representation of risk exposure intensity, knowledge graph-guided adaptive reconstruction of the indicator system, boundary-adjustable membership function, dynamic combination weights that combine subjective and objective factors, non-additive Choquet fuzzy integral synthesis, and closed-loop self-evolution mechanism, it breaks through the inherent limitations of traditional multi-level fuzzy comprehensive evaluation methods in terms of static thresholds, fixed structures, linear weighting, and memoryless inputs. It effectively improves the dynamic adaptability, coupled risk perception capability, scenario generalization level, and long-term operational robustness of distribution network security situation assessment, and can accurately identify high-risk composite fault precursors. Attached Figure Description

[0081] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0082] Figure 1 This is a flowchart of the multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network in Example 1. Detailed Implementation

[0083] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0084] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0085] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0086] Example 1, referring to Figure 1 As one embodiment of the present invention, this embodiment provides a multi-level fuzzy comprehensive evaluation method based on the security status of a power distribution network, including:

[0087] S1. Collect real-time operation data, environmental information and topology of the power distribution network, combine with historical operation sequences to build a time-series prediction model, and output a dynamic safety margin sequence containing key safety indicators.

[0088] Furthermore, real-time measurement sequences of voltage amplitude, branch current, active power and reactive power at each node in the distribution network are obtained from the dispatch automation system;

[0089] Synchronously connect to the meteorological interface to obtain ambient temperature, humidity and wind speed;

[0090] Extract the current network connectivity matrix and device rated parameter vector using a geographic information system;

[0091] The electrical measurement sequences from the dispatch automation system, the environmental parameter sequences from the meteorological interface, and the static topology and equipment parameters provided by the geographic information system are aligned and fused according to a unified timestamp to form a multi-source time series;

[0092] After aligning multi-source time series data with a unified timestamp, the data is input into a long short-term memory neural network. Using observations from the past N time steps as input, the model is trained to predict the future. Safety margins of key safety indicators within the step;

[0093] Suppose a key runtime variable is at time 10:00. The predicted value is Its corresponding limit is The dynamic safety margin is defined as follows:

[0094] ;

[0095] in, Indicates time The safety margin, with a larger value indicating that it is further away from the boundary. These are the predicted values ​​of key running variables output by the time series forecasting model. The upper or lower limit of the operation of this variable is determined by the equipment nameplate parameters or scheduling procedures;

[0096] After the model converges, it outputs a dynamic safety margin sequence. .

[0097] It should be noted that by integrating multi-source real-time measurement, meteorological environment and topological structure data to construct a time series prediction model, and replacing the static threshold with a dynamic safety margin sequence, the safety boundary can adaptively evolve with the operating conditions. This effectively overcomes the problem of misjudgment or omission caused by fixed limits in traditional methods, and effectively improves the forward-looking perception capability of safety status under dynamic scenarios such as new energy fluctuations and load changes.

[0098] In particular, this step aligns three types of heterogeneous data—grid physical measurements, environmental disturbances, and network topology—into a deep time-series model in the time dimension. This allows the prediction of safety margins to not only rely on historical operating trends but also incorporate the mechanism by which the external environment affects equipment performance. As a result, a physically interpretable dynamic safety boundary generation mechanism is built at the source, providing a high-fidelity and forward-looking input foundation for subsequent fuzzy evaluation. This is something that traditional static threshold methods cannot achieve.

[0099] S2. Based on the dynamic safety margin sequence, a time decay accumulation operator is applied to the historical fuzzy membership degree of the underlying safety indicators to generate a cumulative membership vector representing the intensity of risk exposure.

[0100] Furthermore, regarding the first Historical membership sequence of each underlying security indicator under the security rating level An exponential decay weighting mechanism is introduced to calculate the cumulative membership degree. :

[0101] ;

[0102] Among them, the attenuation weight Defined as:

[0103] ;

[0104] in, Indicates the first The risk exposure intensity of each underlying security indicator For the first The degree to which this indicator belongs to the security level at a given historical moment. The length of the sliding time window. The attenuation coefficient;

[0105] All underlying security metrics Constructing cumulative membership vector , as an enhanced representation of the fuzzy input layer;

[0106] Among them, the underlying safety indicators refer to the basic safety variables that can be directly calculated or observed from real-time measurement data of the distribution network, equipment parameters or environmental information, and are used to characterize the local operating status of the system. These include: node voltage deviation rate, branch load rate, transformer load margin, distributed power output fluctuation rate, line N-1 check over-limit flag, and the correction coefficient of ambient temperature on the current carrying capacity of equipment.

[0107] It should be noted that by introducing a time decay accumulation operator to weight and aggregate historical membership, the evaluation input not only reflects the current instantaneous state, but also contains information on the persistence of recent risks and the frequency of impacts. This distinguishes between short-term disturbances and continuous pressure-type risks, enhances the sensitivity of the fuzzy input layer to hidden threats such as equipment aging and heat accumulation, and improves the physical consistency and temporal memory capability of security situation assessment.

[0108] In particular, the time decay accumulation mechanism used in this step is not a simple historical average, but simulates the cognitive habit of human schedulers to pay more attention to recent anomalies, giving higher weight to risk states closer to the current moment, while retaining a certain historical memory to identify slowly accumulating hidden dangers. This design upgrades fuzzy membership from an instantaneous snapshot to a risk exposure intensity representation with time perception capabilities, fundamentally solving the defect of traditional fuzzy input ignoring the characteristics of temporal evolution.

[0109] S3. Call the pre-built distribution network knowledge graph, perform semantic scene recognition based on the current operating status of the distribution network and the dynamic safety margin sequence, and generate a multi-level safety assessment index subset that is adapted to the current operating context.

[0110] Furthermore, the current topology connection matrix, distributed power grid connection flag, microgrid operation mode identifier, and load type code are used as query conditions input into the knowledge graph reasoning engine;

[0111] The knowledge graph stores the relationships between pre-existing runtime scenario nodes and evaluation dimension nodes;

[0112] By using graph embedding similarity matching, the scene node with the highest cosine similarity to the current query vector is selected;

[0113] Activate the set of primary indicators associated with the scene node, the subset of secondary indicators corresponding to each primary indicator, and the mapping relationship between the underlying observable variables and measurement points to form a multi-level safety assessment indicator subset with a variable structure.

[0114] Among them, the multi-level security assessment index subset adapted to the current operating context refers to a set of hierarchical assessment indicators that are most relevant to the current scenario, based on the current actual operating status of the distribution network (such as network topology, distributed power grid connection status, load type, meteorological environment and operating mode, etc.), and semantic scene recognition through a pre-built distribution network knowledge graph. This subset includes a first-level security dimension (voltage safety, equipment thermal stability, power supply reliability), corresponding second-level detailed indicators and their associated underlying observable security variables, which can accurately reflect the key risk points under the current operating conditions, thereby achieving contextual adaptation and focus of the assessment content.

[0115] It should be noted that by using the distribution network knowledge graph to achieve semantic understanding of the operation scenario and on-demand activation of the indicator system, the multi-level evaluation structure is transformed from a preset static tree into a context-driven dynamic subgraph, avoiding interference from irrelevant indicators on the evaluation results. This is especially suitable for complex operation modes such as high proportion of distributed power source access and flexible microgrid switching, and improves the generalization ability and engineering applicability of the method in heterogeneous distribution networks.

[0116] In particular, this step uses knowledge graphs to explicitly model the logical relationships between the distribution network's operating modes, equipment configurations, and evaluation indicators. This allows the system to automatically determine which indicators to look at based on the current scenario, much like an experienced dispatching expert. For example, it can activate frequency stability and energy storage SOC-related indicators when the microgrid is operating in islanded mode, and focus on feeder load balancing and protection coordination indicators when the main grid is experiencing a fault and power transfer. This semantic-driven indicator selection mechanism enhances the context sensitivity and task-specificity of the evaluation system.

[0117] S4. Map the cumulative membership vector to the underlying security index of the multi-level security assessment index subset, construct a membership function with adjustable boundary parameters, and output the fuzzy membership vector of the underlying security index under the preset comment set.

[0118] Furthermore, regarding the first Each underlying security indicator is described by a piecewise exponential membership function, which represents its degree of membership to the security level.

[0119] ;

[0120] Among them, the lower limit and upper limit From dynamic safety margin sequence Real-time correction, the expression is:

[0121] ;

[0122] in, Indicates running value Membership degree to security level and These represent the left and right boundaries of the membership function, which dynamically shrink or expand with the safety margin. For the first The rated limit of each indicator, This is the margin scaling factor, used to adjust the boundary sensitivity. and These are shape adjustment parameters that control the steepness of the membership curve.

[0123] The cumulative membership vector Considering it as an equivalent safety membership degree, we inversely deduce its corresponding operating state, calculate its membership degree under warning and danger levels, and form a ternary fuzzy membership vector. .

[0124] It should be noted that by linking the membership function boundary with the dynamic safety margin sequence, the fuzzy rules acquire environmental adaptive sensitivity—automatically tightening the evaluation criteria when the safety margin narrows to improve the timeliness of early warning, and appropriately relaxing them when the margin is loose to suppress false alarms. This achieves dynamic matching between the severity of the evaluation and the actual risk level of the system, and enhances the robustness of the method under special conditions such as extreme weather or maintenance transition periods.

[0125] In particular, the dynamic adjustment mechanism of the membership function boundary in this step essentially achieves online self-calibration of the evaluation rules—when the system predicts a narrowing of future safety margins, it automatically increases the evaluation sensitivity, enabling even minor exceedances to trigger warnings; conversely, when the system is in a state of ample safety margin, it moderately tolerates fluctuations to avoid frequent alarms interfering with operators' judgments. This adaptive fuzzy rule, with its balanced approach, ensures that the evaluation logic remains synchronized with the actual risk level of the power grid, enhancing the applicability and reliability of the method at different operational stages.

[0126] S5. Calculate the information entropy based on the fuzzy membership vector, and combine the subjective weights determined by the analytic hierarchy process with the topological sensitivity factor derived from the power flow sensitivity to generate dynamic combined weights.

[0127] Furthermore, regarding the first Fuzzy membership vectors of each indicator First, normalize to get:

[0128] ;

[0129] Calculate the information entropy again The expression is:

[0130] ;

[0131] This leads to the objective weight. The expression is:

[0132] ;

[0133] Meanwhile, subjective weights are obtained by constructing a judgment matrix by experts and then undergoing a consistency test. ;

[0134] Calculate the norm of the partial derivative of the power flow Jacobian matrix with respect to the node voltages, and use it as the topology sensitivity factor. ;

[0135] Introducing a Coordination Coefficient Generate dynamic combined weights, the expression is:

[0136] ;

[0137] in, Indicates the first The first indicator in the Normalized membership degree under each rating level Information entropy reflects the amount of effective information provided by the indicator. The lower the entropy, the higher the indicator's discriminative power. To provide objective weights based on data uncertainty, Subjective weighting based on expert experience, The topology sensitivity factor quantifies the structural importance of a device in the current network. This is a coordination coefficient used to balance the weighting of subjective and objective factors. These are the dynamic combination weights used for fuzzy synthesis.

[0138] It should be noted that by integrating expert experience, data uncertainty, and network physical structure to generate dynamic combined weights, the prior knowledge of scheduling procedures is preserved, while also responding to the information value of real-time operational data and the key position of equipment in the topology. This avoids the one-sidedness of a single weighting method and makes the weight allocation more in line with the safety logic of the distribution network structure-operation-cognition integration.

[0139] In particular, the three-source integrated weighting mechanism proposed in this step breaks through the limitations of traditional single weighting: subjective weights reflect procedural constraints and expert experience, objective weights reflect the distinguishing ability of the data itself, and topology sensitivity introduces key information about the physical structure of the power grid. The three work together to ensure that the weight allocation does not blindly rely on data noise or adhere to rigid rules, but forms a dynamic balance system that takes into account the multi-dimensional cognition of humans, machines, and the grid, ensuring that the evaluation results have both theoretical basis and engineering rationality.

[0140] S6. Based on the physical coupling relationship between dynamic combined weights and underlying security indicators, a non-additive fuzzy measure is constructed. The fuzzy membership vector is aggregated level by level using Choquet fuzzy integral to obtain the highest-level comprehensive fuzzy evaluation vector.

[0141] Furthermore, for index pairs that are coupled, an interactive metric is defined based on their joint historical frequency of exceeding limits. ;

[0142] For each secondary indicator group Arrange their fuzzy membership vectors in descending order of membership degree:

[0143] ;

[0144] The corresponding indicator index set is denoted as ;

[0145] The second-order integrated membership degree is calculated using Choquet integral, and the expression is as follows:

[0146] ;

[0147] in, , Indicates the first The comprehensive fuzzy membership degree of each secondary indicator. For the sorted number Large membership values For the front A set consisting of high membership indices It is a fuzzy measure of the set, reflecting the overall risk level under the synergistic effect of indicators;

[0148] The repeated process aggregates upwards to the first-level indicator layer, resulting in a comprehensive fuzzy evaluation vector. .

[0149] It should be noted that by using Choquet fuzzy integral as a non-additive synthesis operator, the synergistic deterioration effect between indicators is explicitly modeled, which breaks through the limitation of traditional linear weighting in underestimating composite risks. This significantly amplifies the comprehensive membership of high-risk scenarios, thereby more realistically reflecting the nonlinear and cascading characteristics in the evolution of distribution network faults and improving the early identification capability of potential cascading faults.

[0150] Specifically, the core idea behind this step, which employs a non-additive synthesis operator, is to acknowledge that the risks in the distribution network are not simply the sum of individual risk indicators, but rather exhibit a synergistic amplification effect where 1+1>2. For instance, low voltage may be manageable on its own, but if it is accompanied by heavy load, the coupling of insulation stress and thermal stress may lead to accelerated equipment degradation. The Choquet integral quantifies this interaction intensity through fuzzy measurement, enabling the comprehensive evaluation result to truly reflect the overall vulnerability of the system, rather than being a patchwork of isolated indicators. This is the key to improving the accuracy of high-risk scenario identification.

[0151] S7. Map the comprehensive fuzzy evaluation vector to the ideal safety state, calculate the distribution network safety status index and generate a safety status assessment result that includes risk level and weak links. At the same time, compare the safety status assessment result with the subsequent actual operation events to extract misjudged samples for updating the time series prediction model, membership function parameters, time decay accumulation operator and fuzzy measure, and complete the evaluation model update.

[0152] Furthermore, we define the ideal safety state vector. This indicates that the system is at a completely safe level.

[0153] Calculate the comprehensive fuzzy evaluation vector and The weighted Hamming distance is expressed as:

[0154] ;

[0155] in, The combined weights of the primary indicators;

[0156] The distribution network security status index is defined as follows:

[0157] ;

[0158] in, This indicates the degree of deviation between the current state and the ideal safe state. The final output is a safety score;

[0159] After the evaluation period ends, the comparison The warning level is related to whether a protective action, equipment tripping, or user complaint actually occurred.

[0160] like ≥80 but a fault occurs, mark as a missed alarm sample;

[0161] like ≤40 but no events are marked as false alarm samples;

[0162] Add the input sequence within the time window corresponding to the missed alarm sample to the training set to trigger incremental training of the LSTM model;

[0163] At the same time, adjust the attenuation coefficient of the corresponding indicator. Membership function shape parameters , And fuzzy measures enable the updating of parameters across the entire link.

[0164] It should be noted that by establishing a closed-loop feedback mechanism between the evaluation results and actual operational events, the missed alarm and false alarm samples are transformed into model optimization signals, driving the online evolution of core components such as time series prediction, membership function, cumulative operator and fuzzy measure. This enables the evaluation system to have continuous learning and self-correction capabilities, effectively cope with long-term evolutionary factors such as equipment aging, network structure reconstruction and control strategy updates, and ensure the stability and reliability of evaluation performance throughout the entire life cycle.

[0165] In particular, the closed-loop feedback mechanism constructed in this step enables the entire evaluation system to have a self-learning ability similar to an immune system: each misjudgment is regarded as an immune response, triggering the targeted adjustment of the model's internal parameters, so that the system can continuously accumulate experience and correct deviations in long-term operation. The continuous evolution characteristic ensures that the method will not quickly become ineffective due to changes in the power grid structure, equipment updates, or adjustments in operating strategies. It truly realizes the technological leap from one-time deployment to lifelong self-optimization, providing lasting and reliable security assessment support for smart distribution networks.

[0166] This embodiment also provides a computer device applicable to the multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network, including: 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 multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network as proposed in the above embodiment.

[0167] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0168] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0169] In summary, this invention overcomes the inherent limitations of traditional multi-level fuzzy comprehensive evaluation methods in terms of static thresholds, fixed structures, linear weighting, and memoryless input by integrating time-series prediction-driven dynamic safety margin modeling, time-accumulated representation of risk exposure intensity, knowledge graph-guided adaptive reconstruction of the indicator system, boundary-adjustable membership functions, dynamic combination weights that combine subjective and objective factors, non-additive Choquet fuzzy integral synthesis, and closed-loop self-evolution mechanism. It effectively improves the dynamic adaptability, coupled risk perception capability, scenario generalization level, and long-term operational robustness of distribution network security situation assessment, and can accurately identify precursors of high-risk composite faults.

[0170] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-level fuzzy comprehensive evaluation method based on the security status of a power distribution network, characterized in that: include: Collect real-time operation data, environmental information and topology of the power distribution network, combine them with historical operation sequences to build a time series prediction model, and output a dynamic safety margin sequence containing key safety indicators; Based on the dynamic safety margin sequence, a time decay accumulation operator is applied to the historical fuzzy membership degree of the underlying safety indicators to generate a cumulative membership vector characterizing the intensity of risk exposure. The pre-built distribution network knowledge graph is invoked, and semantic scene recognition is performed based on the current operating status of the distribution network and the dynamic safety margin sequence to generate a multi-level safety assessment index subset adapted to the current operating context. The cumulative membership vector is mapped to the underlying security index of the multi-level security assessment index subset, a membership function with adjustable boundary parameters is constructed, and the fuzzy membership vector of the underlying security index under the preset comment set is output. Information entropy is calculated based on fuzzy membership vectors, and dynamic combined weights are generated by integrating subjective weights determined by the analytic hierarchy process and topological sensitivity factors derived from power flow sensitivity. Based on the physical coupling relationship between dynamic combined weights and underlying security indicators, a non-additive fuzzy measure is constructed. Choquet fuzzy integrals are used to aggregate the fuzzy membership vectors level by level to obtain the highest-level comprehensive fuzzy evaluation vector. The comprehensive fuzzy evaluation vector is mapped to the ideal safety state, the distribution network safety status index is calculated, and a safety status assessment result containing risk level and weak links is generated. At the same time, the safety status assessment result is compared with the subsequent actual operation events to extract misjudged samples for updating the time series prediction model, membership function parameters, time decay accumulation operator and fuzzy measure, thus completing the evaluation model update.

2. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 1, characterized in that: The system collects real-time operating data, environmental information, and topology of the power distribution network, combines this data with historical operating sequences to construct a time-series prediction model, and outputs a dynamic safety margin sequence containing key safety indicators, including: The real-time measurement sequences of voltage amplitude, branch current, active power and reactive power of each node in the distribution network are obtained from the dispatch automation system. Synchronously connect to the meteorological interface to obtain ambient temperature, humidity and wind speed; Extract the current network connectivity matrix and device rated parameter vector using a geographic information system; The electrical measurement sequences from the dispatch automation system, the environmental parameter sequences from the meteorological interface, and the static topology and equipment parameters provided by the geographic information system are aligned and fused according to a unified timestamp to form a multi-source time series; After aligning multi-source time series data with a unified timestamp, the data is input into a long short-term memory neural network. Using observations from the past N time steps as input, the model is trained to predict the future. Safety margins of key safety indicators within the step; Suppose a key runtime variable is at time 10:

00. The predicted value is Its corresponding limit is The dynamic safety margin is defined as follows: ； in, Indicates time The safety margin, with a larger value indicating that it is further away from the boundary. These are the predicted values ​​of key running variables output by the time series forecasting model. The upper or lower limit of the operation of this variable is determined by the equipment nameplate parameters or scheduling procedures; After the model converges, it outputs a dynamic safety margin sequence. .

3. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 2, characterized in that: The method, based on the dynamic safety margin sequence, applies a time-decaying cumulative operator to the historical fuzzy membership degree of the underlying safety indicators to generate a cumulative membership vector characterizing the intensity of risk exposure, including: For the Historical membership sequence of each underlying security indicator under the security rating level An exponential decay weighting mechanism is introduced to calculate the cumulative membership degree. : ; Among them, the attenuation weight Defined as: ; in, Indicates the first The risk exposure intensity of each underlying security indicator For the first The degree to which this indicator belongs to the security level at a given historical moment. The length of the sliding time window. The attenuation coefficient; All underlying security metrics Constructing cumulative membership vector , as an enhanced representation of the fuzzy input layer.

4. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 3, characterized in that: The process involves calling a pre-built distribution network knowledge graph, performing semantic scene recognition based on the current operating state of the distribution network and the dynamic safety margin sequence, and generating a multi-level safety assessment index subset adapted to the current operating context, including: Input the current topology connection matrix, distributed power grid connection flag, microgrid operation mode identifier, and load type code as query conditions into the knowledge graph reasoning engine; The knowledge graph stores the relationships between pre-existing runtime scenario nodes and evaluation dimension nodes; By using graph embedding similarity matching, the scene node with the highest cosine similarity to the current query vector is selected; The set of primary indicators associated with the scene node, the subset of secondary indicators corresponding to each primary indicator, and the mapping relationship between the underlying observable variables and measurement points are activated to form a multi-level safety assessment indicator subset with a variable structure.

5. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 4, characterized in that: The process of mapping the cumulative membership vector to the underlying security index of the multi-level security assessment index subset, constructing a membership function with adjustable boundary parameters, and outputting the fuzzy membership vector of the underlying security index under a preset comment set includes: For the Each underlying security indicator is described by a piecewise exponential membership function, which represents its degree of membership to the security level. ; Among them, the lower limit and upper limit From dynamic safety margin sequence Real-time correction, the expression is: ; in, Indicates running value Membership degree to security level and These represent the left and right boundaries of the membership function, which dynamically shrink or expand with the safety margin. For the first The rated limit of each indicator, This is the margin scaling factor, used to adjust the boundary sensitivity. and These are shape adjustment parameters that control the steepness of the membership curve. The cumulative membership vector Considering it as an equivalent safety membership degree, we inversely deduce its corresponding operating state, calculate its membership degree under warning and danger levels, and form a ternary fuzzy membership vector. .

6. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 5, characterized in that: The process of calculating information entropy based on fuzzy membership vectors, integrating subjective weights determined by the analytic hierarchy process (AHP) with topological sensitivity factors derived from power flow sensitivity, and generating dynamic combined weights includes: For the Fuzzy membership vectors of each indicator First, normalize to get: ; Calculate the information entropy again The expression is: ; This leads to the objective weight. The expression is: ; Meanwhile, subjective weights are obtained by constructing a judgment matrix by experts and then undergoing a consistency test. ; Calculate the norm of the partial derivative of the power flow Jacobian matrix with respect to the node voltages, and use it as the topology sensitivity factor. ; Introducing a Coordination Coefficient Generate dynamic combined weights, the expression is: ; in, Indicates the first The first indicator in the Normalized membership degree under each rating level Information entropy reflects the amount of effective information provided by the indicator. The lower the entropy, the higher the indicator's discriminative power. To provide objective weights based on data uncertainty, Subjective weighting based on expert experience, The topology sensitivity factor quantifies the structural importance of a device in the current network. This is a coordination coefficient used to balance the weighting of subjective and objective factors. These are the dynamic combination weights used for fuzzy synthesis.

7. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 6, characterized in that: Based on the physical coupling relationship between dynamic combined weights and underlying security indicators, a non-additive fuzzy measure is constructed. Choquet fuzzy integrals are used to aggregate the fuzzy membership vectors level by level to obtain the highest-level comprehensive fuzzy evaluation vector, including: For index pairs that are coupled, an interactive metric is defined based on their joint historical frequency of exceeding limits. ; For each secondary indicator group Arrange their fuzzy membership vectors in descending order of membership degree: ; The corresponding indicator index set is denoted as ; The second-order integrated membership degree is calculated using Choquet integral, and the expression is as follows: ; in, , Indicates the first The comprehensive fuzzy membership degree of each secondary indicator. For the sorted number Large membership values For the front A set consisting of high membership indices It is a fuzzy measure of the set, reflecting the overall risk level under the synergistic effect of indicators; The repeated process aggregates upwards to the first-level indicator layer, resulting in a comprehensive fuzzy evaluation vector. .

8. The multi-level fuzzy comprehensive evaluation method based on the security status of the distribution network as described in claim 7, characterized in that: The process involves mapping the comprehensive fuzzy evaluation vector to an ideal safety state, calculating the distribution network safety situation index, and generating a safety situation assessment result that includes risk levels and weak points. Simultaneously, the safety situation assessment result is compared with subsequent actual operational events to identify discrepancies. Misjudged samples are extracted to update the time-series prediction model, membership function parameters, time decay accumulation operator, and fuzzy measure, thus completing the evaluation model update. This includes: Define the ideal safety state vector This indicates that the system is at a completely safe level. Calculate the comprehensive fuzzy evaluation vector and The weighted Hamming distance is expressed as: ; in, The combined weights of the primary indicators; The distribution network security status index is defined as follows: ; in, This indicates the degree of deviation between the current state and the ideal safe state. The final output is a safety score; After the evaluation period ends, the comparison The warning level is related to whether a protective action, equipment tripping, or user complaint actually occurred. like ≥80 but a fault occurs, mark as a missed alarm sample; like ≤40 but no events are marked as false alarm samples; Add the input sequence within the time window corresponding to the missed alarm sample to the training set to trigger incremental training of the LSTM model; At the same time, adjust the attenuation coefficient of the corresponding indicator. Membership function shape parameters , And fuzzy measures enable the updating of parameters across the entire link.

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 multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network 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 multi-level fuzzy comprehensive evaluation method based on the security status of the power distribution network as described in any one of claims 1 to 8.

Images