Power distribution network security evaluation and tracing method and system based on er and prospect theory

Abstract

The present application relates to a kind of based on ER and foreground theory distribution network security evaluation and tracing method and system, belong to electric communication technical field.The present application establishes equipment-feeder-network three-level index system, gives each index quantification calculation formula;Using improved ER rule is modified to index BPA double, multi-level evidence is aggregated from bottom to top by fusing structure, business importance;Markov chain is introduced to predict security posture and dynamically update evaluation reference point, according to index comprehensive confidence self-adapting adjustment foreground theory weight curvature, strengthen the risk perception of small probability high loss, calculate dynamic security prospect value;Risk contribution degree model is built, set two-level tracing rule, risk feeder and equipment are located reversely after early warning, form evaluation-tracing closed loop.Surrounding system is by multifunctional submodule Collaborative implementation above-mentioned algorithm.The present application can inhibit data noise, accurately perceive dynamic risk, consider whole network situation evaluation and bottom risk source positioning, applicable to new type distribution network security operation.

Description

Technical Field

[0001] This invention relates to a method and system for power distribution network security assessment and tracing based on ER and prospect theory, belonging to the field of power technology. Background Technology

[0002] With the large-scale integration of new entities such as distributed photovoltaic power, electric vehicles, and flexible loads, the bidirectional flow characteristics of power distribution networks are becoming increasingly apparent, and the random fluctuations on both the source and load sides are strengthening. The uncertainty, volatility, and coupling of the operating status are significantly enhanced. In complex scenarios with superimposed multi-source disturbances and accelerated evolution of local risks, traditional single-moment, single-level security assessment methods are no longer sufficient to meet the needs of security situation awareness and risk propagation identification under complex operating conditions.

[0003] The core of distribution network safety assessment lies in mapping multi-dimensional indicators characterizing operational features such as voltage levels, load states, and power supply reliability into safety status evidence based on a predetermined set of safety levels, and obtaining the safety level and risk level of the assessed object through a certain fusion mechanism. However, existing methods often use single objects or flat systems as analysis units, failing to fully explore the multi-level correlations and transmission relationships between equipment, feeders, and the network, and insufficiently considering the heterogeneity of different assessed objects in terms of structural load-bearing capacity and business sensitivity. Furthermore, existing indicator fusion weights often rely on subjective experience or static statistical characteristics, making it difficult to handle conflicting evidence, incomplete information, and noise disturbances in monitoring data under complex scenarios, thus reducing assessment accuracy. In addition, existing methods lack an evaluation mapping mechanism that couples data reliability with operational and maintenance decision-making risk preferences, resulting in insufficient perception of low-probability, high-loss events. More importantly, existing methods lack a closed-loop linkage mechanism between overall risk characterization and local risk tracing, making it difficult to decompose and map the overall network alarm results to specific risk carriers such as feeders and equipment, and failing to effectively transmit assessment results to source diagnosis. Therefore, there is an urgent need to construct a distribution network safety assessment method and system that can take into account both dynamic safety assessment and risk source tracing, so as to achieve accurate positioning from overall risk to local risk sources. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this invention is to provide a method and system for distribution network security assessment and tracing based on ER and prospect theory.

[0005] The present invention provides a distribution network security assessment and source tracing method based on ER and prospect theory, comprising the following steps:

[0006] S1: Construct a hierarchical safety assessment index system covering the equipment layer, feeder layer, and network layer; the equipment layer selects equipment load rate, frequency and duration of forward and reverse heavy overload of distribution transformers, and voltage qualification rate at grid connection point to characterize the operating pressure of a single device and the local voltage quality; the feeder layer selects feeder load rate, power backflow risk level, and feeder voltage deviation to characterize the power flow disturbance risk of multi-node coupling in the feeder; the network layer selects maximum power supply capacity, voltage over-limit frequency, and load balance degree to measure the overall power supply safety margin of the entire network. S2: Construct an improved two-stage ER assessment architecture. In the first stage, the basic probability allocation (BPA) of each indicator within a single object is generated. The indicator importance coefficient and evidence reliability coefficient are used for double correction. The improved ER rules are used to fuse and obtain the security level confidence distribution of a single object. In the second stage, the structural importance and business importance of equipment and feeders are calculated respectively. The comprehensive hierarchical weight is obtained by coupling the weight fusion coefficient. The confidence of the lower-level assessment is discounted by strength and used as sub-evidence. The ER is aggregated from bottom to top to obtain the overall security level confidence distribution of the feeder layer and network layer. S3: Construct a dynamic evaluation model that couples Markov prediction with adaptive prospect theory. Construct a Markov chain state space with four security levels. Predict the security level distribution at the next moment based on the state transition probability matrix and calculate the risk evolution trend factor. Dynamically update the evaluation reference point by combining the current static expected utility, the predicted expected utility, and the trend sensitivity coefficient. Aggregate the reliability of the underlying indicators to obtain the comprehensive confidence level. Adaptively adjust the curvature parameter of the probability weight function of the prospect theory. The lower the data reliability, the greater the weight of low-probability, high-loss events. Combine the value function to solve the comprehensive security prospect value at each level. S4: Establish a two-level closed-loop tracing mechanism based on prospect value decomposition. When the comprehensive security prospect value of the network layer is lower than the warning threshold, calculate the risk contribution of each lower-level unit based on the ideal security benchmark prospect value. First, perform network-feeder tracing to screen feeders with risks exceeding the threshold. Then, perform feeder-equipment tracing to locate risky equipment for each risky feeder. Complete the reverse risk location from the entire network to a single device, forming a closed loop of assessment, warning and risk tracing linkage.

[0007] Furthermore, in step S1, the equipment load rate is the ratio of the real-time load of the equipment to its rated capacity; the frequency of distribution transformer forward and reverse heavy overload is the number of transitions from power flow switching to heavy overload within the day; the duration of heavy overload is the cumulative duration of the load rate exceeding the threshold within the evaluation window; and the grid connection point voltage qualification rate is the percentage of the time the voltage falls within the allowable deviation range within the window.

[0008] Furthermore, in step S1, the risk level of power backfeeding at the feeder layer is constructed based on the utility function of the reverse active power at the feeder head end, and the risk weight of the high-power backfeeding condition is amplified by the risk preference coefficient; the degree of feeder voltage deviation is a quantitative index of the overall deviation of the voltage of all nodes of the feeder relative to the reference voltage.

[0009] Furthermore, in step S1, the maximum power supply capacity of the network layer is the sum of the maximum load that the main transformers of the entire network can bear under the N-1 safety constraint; the load balance is characterized by the root mean square difference of the load rate of each feeder to represent the degree of uneven load distribution in the entire network.

[0010] Furthermore, in step S2, the improved ER fusion rule is used to synthesize multiple sets of modified BPAs, discounting and weakening low-reliability conflict evidence, and cumulatively strengthening high-reliability consistent evidence. After fusion, the confidence distribution of each security level is output after normalization.

[0011] Furthermore, in step S2, the comprehensive hierarchical weight is obtained by weighted fusion of structural importance and business importance. Structural importance relies on the rated capacity of equipment and the peak load of feeders to characterize the degree of topological criticality, while business importance relies on the load sensitivity coefficient to characterize the reliability requirements of power supply services.

[0012] Furthermore, the update rule for the dynamic reference point in step S3 is as follows: if the risk evolution trend factor is less than 0, it means that the risk has deteriorated, and the dynamic reference point is raised to improve the assessment sensitivity; if the trend factor is greater than or equal to 0, it means that the risk is stable or improving, and the dynamic reference point gradually falls back to the preset safety benchmark.

[0013] Furthermore, the adaptive curvature adjustment logic in step S3 is as follows: when the overall confidence level decreases, the curvature of the probability weight function decreases, thereby enhancing the perception of extreme tail risks and realizing a defense assessment logic where the more fuzzy the monitoring data, the more conservative the assessment results.

[0014] Furthermore, step S4, risk contribution, is used to quantify the degree to which lower-level units cause a decrease in the security prospect value of the upper-level system. Feeders and devices with risk contribution values ​​higher than the corresponding discrimination coefficient threshold are identified as risk sources. The discrimination coefficient can be flexibly adjusted to improve the source tracing sensitivity according to the operation and maintenance scenario.

[0015] A distribution network security assessment and tracing system based on ER and prospect theory includes: The data acquisition submodule is used to collect telemetry and teleindication multi-source operational data from the equipment layer, feeder layer, and network layer, and to complete data preprocessing. The index calculation and BPA generation submodule is used to calculate the security assessment indexes at each level of the device layer, feeder layer, and network layer, and generate the basic probability allocation (BPA) corresponding to each security level through membership function mapping. The Evidence Correction and ER Fusion submodule is used to correct the BPA based on the indicator importance coefficient and the evidence reliability coefficient, and to obtain the safety level reliability distribution of the assessed object by fusion using improved ER rules. The multi-level aggregation submodule is used to construct a comprehensive hierarchical weight based on structural importance and business importance, and aggregate the lower-level evaluation results as sub-evidence from bottom to top to obtain the upper-level evaluation reliability distribution. The Markov prediction submodule is used to predict the safety level distribution at the next moment based on the safety level state transition probability matrix and to calculate the risk evolution trend factor. The dynamic reference point update submodule is used to dynamically update the dynamic reference points for safety assessment based on risk evolution trend factors. The adaptive prospect scoring submodule is used to adaptively adjust the curvature parameter of the prospect theoretical probability weight function based on the comprehensive confidence of the underlying indicators, and calculate the comprehensive safety prospect value of each level of the evaluation object by combining the value function. The risk tracing submodule is used to calculate the risk contribution of lower-level units based on prospective value decomposition, perform two-level tracing and discrimination between network and feeder and feeder and equipment, and locate key risk feeders and risk equipment. The results output submodule is used to output the comprehensive security prospect value, security level judgment result, and risk tracing and location result of each layer of the device layer, feeder layer, and network layer. The communication submodule is used to upload security assessment data, risk warning information, and source tracing results to the dispatch center.

[0016] By means of the above-described solution, the present invention has at least the following advantages: 1. Existing methods often use single objects or flat systems as analysis units, failing to fully explore the multi-level correlations and transmission relationships between equipment, feeders, and networks. Furthermore, they do not adequately consider the structural and service heterogeneity of the assessed objects, leading to insufficient robustness in security status assessments under complex and uncertain environments. Therefore, this invention constructs a hierarchical security assessment index system covering the equipment layer, feeder layer, and network layer, mapping operational observations to security level evidence. It further proposes a multi-level ER fusion framework that considers index importance, evidence reliability, and structural / service heterogeneity. In the index fusion stage, it performs dual corrections on the importance and reliability of multiple index evidence within the same object. In the hierarchical aggregation stage, it introduces structural importance and service importance to construct a comprehensive hierarchical weight, achieving robust quantitative assessment from the equipment layer to the feeder layer and then to the network layer from the bottom up.

[0017] 2. Existing methods rely heavily on subjective experience or static statistical characteristics for their fusion weights, making it difficult to handle conflicting evidence, incomplete information, and noise disturbances. Furthermore, they lack a mapping mechanism that couples data reliability with risk preference, providing only static assessment results and failing to adequately perceive risk deterioration trends and low-probability, high-loss events. Therefore, this invention constructs a dynamic quantification mechanism based on Markov evolution and confidence-adaptive prospect theory. On one hand, Markov chains are used to characterize the transition patterns of security levels over time and achieve short-term trend prediction, constructing a risk-driven dynamic reference point and transforming risk deterioration trends into a more stringent assessment benchmark. On the other hand, the comprehensive confidence level obtained from the fusion of multi-source evidence is mapped to the curvature parameter of the prospect theory probability weight function. When data quality declines, the decision weights for low-probability, high-risk events are automatically amplified, forming a defensive logic of "the more ambiguous the data, the more conservative the assessment," thereby enhancing the ability to perceive risk deterioration trends and low-probability, severe events.

[0018] 3. Existing methods lack a closed-loop linkage between overall risk characterization and local risk tracing, failing to decompose and map the overall network alarm results to specific feeders and equipment, making it difficult to achieve a seamless connection from macro-level situational awareness to micro-level fault location. Therefore, this invention proposes a two-level risk tracing method based on prospective value decomposition. It constructs a risk contribution index based on the decline in comprehensive security prospective value, establishes two-level tracing rules for network-feeder and feeder-equipment levels, achieving a closed-loop unification of bottom-up hierarchical assessment and top-down reverse diagnosis, providing methodological support for proactive defense and lean operation and maintenance of distribution networks.

[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show a certain embodiment of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a block diagram of the power distribution network safety assessment and risk tracing system that integrates multi-level ER and adaptive prospect theory according to the present invention; Figure 2 This is a flowchart of the dynamic security assessment and risk tracing method for distribution networks that integrates multi-level ER and adaptive prospect theory. Detailed Implementation

[0022] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0023] This invention proposes a dynamic security assessment and risk tracing system for distribution networks that integrates multi-level ER (Earnings Response) and adaptive prospect theory. Its architecture is as follows: Figure 1 As shown, it includes a data acquisition submodule, an indicator calculation and BPA generation submodule, an evidence correction and ER fusion submodule, a multi-level aggregation submodule, a Markov prediction submodule, a dynamic reference point update submodule, an adaptive prospect scoring submodule, a risk tracing submodule, a result output submodule, and a communication submodule.

[0024] The data acquisition submodule is responsible for collecting multi-source operational data, such as telemetry and telesignaling, from equipment, feeders, and network layers, and performing preprocessing. The indicator calculation and BPA generation submodule is responsible for calculating the security assessment indicators at each level and mapping them to the basic probability assignment (BPA) for each security level through the membership function. The Evidence Correction and ER Fusion submodule is responsible for correcting the BPA based on the indicator importance coefficient and the evidence reliability coefficient, and using improved ER rules to fuse the security level reliability distribution of the evaluated object. The multi-level aggregation submodule is responsible for constructing a comprehensive hierarchical weight based on structural importance and business importance, and aggregating the lower-level evaluation results as sub-evidence from bottom to top. The Markov prediction submodule is responsible for predicting the safety level distribution at the next moment based on the state transition probability matrix and calculating the risk evolution trend factor. The dynamic reference point update submodule is responsible for updating the dynamic reference point based on the risk evolution trend; The adaptive prospect scoring submodule is responsible for adaptively adjusting the curvature of the probability weight function based on the comprehensive confidence level, and calculating the comprehensive safe prospect value by combining the value function. The risk tracing submodule is responsible for calculating the risk contribution based on prospective value decomposition, performing two-level tracing and identification of network-feeder and feeder-equipment, and locating key risk feeders and risky equipment. The output submodule is responsible for outputting the security prospect value, security level and tracing results of each level; The communication submodule is responsible for sending security assessment results, early warning information, and source tracing and location results to the dispatch center.

[0025] This invention proposes a method for dynamic security assessment and risk tracing of distribution networks that integrates multi-level ER (Earnings Response) and adaptive prospect theory. The specific process is as follows: Figure 2 As shown.

[0026] S1: Construct a hierarchical safety assessment index system covering the equipment layer, feeder layer, and network layer. The equipment layer reflects the operating pressure of a single device and the voltage quality at the grid connection point; the feeder layer describes the load-bearing level, power flow direction changes, and voltage distribution disturbances under the coupled operation of multiple nodes within the same feeder; the network layer measures the overall safety margin from the perspective of the entire network's power supply capacity, over-limit events, and load distribution balance, providing a unified quantitative basis for safety assessment.

[0027] S2: Construct a two-stage evaluation architecture based on improved ER and multi-level aggregation. In the indicator fusion stage, the evidence of multiple indicators within the same evaluation object is modified for both importance and reliability, and the security level reliability distribution of the object is obtained by using improved ER rules. In the hierarchical aggregation stage, structural importance and business importance are introduced to construct a comprehensive hierarchical weight, and the evaluation results of lower-level objects are weighted and propagated to achieve bottom-up quantitative evaluation from the device layer to the feeder layer and then to the network layer.

[0028] S3: Construct a dynamic security assessment model based on Markov prediction and adaptive prospect theory. Markov chains are used to characterize the security level transition patterns and predict short-term trends, constructing risk-driven dynamic reference points. The comprehensive confidence level is mapped to an adaptive curvature parameter of the probability weight function, and the comprehensive security prospect value of each assessment object is calculated using a value function, enabling proactive perception of risk deterioration trends and low-probability, high-loss events.

[0029] S4: Construct a hierarchical risk tracing mechanism based on prospect value decomposition. When the comprehensive safety prospect value is lower than the warning threshold, calculate the risk contribution of each object, and sequentially execute the two-level tracing discrimination rules of network-feeder and feeder-equipment to identify the set of key risk feeders and the set of risky equipment, forming a dynamic closed loop of "assessment and warning - tracing and location".

[0030] In a further embodiment, step S1 includes: This invention establishes a multi-level security assessment index system comprising the equipment layer, feeder layer, and network layer. Assume the distribution network system is... , contains The feeder lines are represented as a set. For any feeder It contains a total of The set of devices is represented as Select within the evaluation period. The set of evaluation moments is represented as: This system can not only depict the overall security status of the power distribution network, but also retain the local risk characteristics of individual feeders and individual devices.

[0031] S1.1 Equipment Layer Security Assessment Indicators The equipment layer focuses on the operational pressure and local voltage quality constraints of a single device under the influence of the new main body access, and selects indicators such as equipment load rate, frequency and duration of forward and reverse heavy overload of distribution transformer, and voltage qualification rate at grid connection point.

[0032] 1) Equipment Load Rate: This indicator characterizes the operating pressure of equipment and its impact on system safety. When equipment operates under high load rates for extended periods, the increased thermal stress and insulation aging lead to a higher risk of failure, easily causing low voltage at the end of the line and in transformer substations. The rated capacity of the equipment is defined as... At any moment The actual load of the equipment is Then the equipment load rate Represented as (1) 2) Frequency and duration of heavy overload in both forward and reverse directions: This indicator describes the frequency and duration of heavy overload conditions experienced by the transformer under two power flow directions: power flowing from the power source side to the load side and power transmitted in reverse by distributed generation. The heavy overload criterion is defined as a load rate not less than a threshold value. Define the positive and negative overload state indicator variables as follows: (2) (3) In the formula, This is an indicator function; it takes the value 1 if the condition is true, and 0 otherwise. Indicating a positive trend, Indicates reverse current flow. Defined as... Starting point, duration is Evaluation window The duration of the forward and reverse overload within the evaluation window is denoted as . and , represented as (4) (5) The frequency of heavy overload in both directions of the distribution transformer is counted based on the initial transition from non-heavy overload to heavy overload within a day, and recorded as follows: and ,Right now (6) (7) In the formula, This is a collection of data for the entire day.

[0033] 3) Grid connection point voltage qualification rate: This indicator describes the proportion of time during the evaluation window that the voltage at the grid connection point falls within the allowable voltage deviation range, so as to reflect the impact of new entities connected to the grid on the local voltage level and voltage over-limit risk. The qualification criterion is based on the rated voltage as the reference, and the voltage amplitude is defined to be qualified when it is within the interval , then the grid connection point voltage qualification rate at the evaluation moment is expressed as (8) In the formula, and are the upper and lower limits of the voltage amplitude.

[0034] S1.2 Feeder layer operation state indicators The feeder layer reflects the coupled operation state of multiple devices and multiple nodes on the same feeder, as well as the changes in power flow direction and voltage disturbance risk. The selected indicators are feeder load rate, reverse power flow risk level, and feeder voltage deviation degree.

[0035] 1) Feeder load rate: This indicator characterizes the utilization degree of the feeder's carrying level relative to its allowable transmission capacity during operation. An excessively high load rate will lead to increased current and losses, and raise the risk of power flow over-limit and line faults. The feeder load rate indicator is defined as the maximum value of the ratio of the feeder current to its maximum allowable transmission current, denoted as , that is (9) In the formula, is the current of feeder , is the maximum allowable transmission current.

[0036] 2) Reverse power flow risk level: This indicator describes the severity of active power reverse flow occurring at the head end of the feeder. Based on the amplitude of the reverse power, this indicator assigns a higher weight to large reverse power conditions through a risk-preference utility function, highlighting the potential impact of reverse power flow on distribution network security. The reverse power flow risk level at moment is denoted as , that is (10) In the formula, is the power at the head end of feeder , is the allowable upper limit of reverse power, is the risk preference coefficient.

[0037] 3) Feeder Voltage Deviation: This indicator characterizes the overall voltage deviation level of each node in the feeder relative to the reference voltage, reflecting the overall disturbance intensity and spatial influence range of the voltage distribution after the new main body is connected. Assume the feeder... The set of nodes is The degree of feeder voltage deviation Represented as (11) In the formula, For nodes At any moment The voltage.

[0038] S1.3 Network Layer Operation Status Indicators The network layer characterizes the overall security margin and operational boundary constraint level of the entire network under the combined action of multiple new types of entities, and selects indicators such as maximum power supply capacity, voltage over-limit frequency, and load balance.

[0039] 1) Maximum Power Supply Capacity: This indicator reflects the maximum load level that a distribution network can withstand under given network topology, equipment capacity, and operational constraints while meeting the N-1 safety criterion. It is a key indicator for characterizing network security from the perspective of power supply capacity and structural safety margin. The maximum power supply capacity indicator is defined as the sum of the active power of each main transformer under the maximum power supply capacity condition under N-1 safety constraints, denoted as... ,Right now (12) In the formula, For the main transformer of the equipment layer The load capacity under maximum power supply conditions. The main transformer equipment set.

[0040] 2) Voltage Exceedance Frequency: This indicator is used to count the number of times the voltage of all network nodes exceeds the allowable deviation range within a day, reflecting the intensity of discrete events that violate voltage constraints and their impact on system operational safety. Let the set of network nodes be... Define the voltage over-limit indication function as follows: (13) In the formula, For nodes exist Voltage at any given moment; V max and V min These represent the upper and lower limits of the allowable deviation range for node voltages, respectively; the symbol "∨" is the logical "OR" operator, indicating that the indicator function will set to 1 when any voltage exceedance condition is met. The frequency of network layer voltage exceedances is expressed as... (14) 3) Load Balance: A high proportion of new main transformers connecting to the grid in certain areas may exacerbate load imbalance, leading to prolonged high-load operation of some equipment and increasing the risk of local grid operation problems. Load balance characterizes the degree of load distribution balance among different main transformers and feeders, and is defined as the root mean square difference of the feeder load rate, denoted as . ,Right now (15) In the formula, This represents the average feeder load rate.

[0041] In a further embodiment, step S2 includes: S2.1 Assessment Levels and Basic Probability Allocation Define the identification framework for distribution network security assessment as follows: These correspond to four levels: safe (S1), cautious (S2), abnormal (S3), and severe (S4). For the assessment object... (Equipment, feeder, or network), note its first... The normalized observations of each indicator are Evidence is constructed by mapping the fuzzy membership function to the support for each level. Confidence structure: (16) In the formula, For the first Each indicator supports the target at the level. The probability, This represents an uncertain probability.

[0042] S2.2 Indicator Fusion Based on Improved ER To reflect the differences in information contribution and data quality among different indicators, an importance coefficient is introduced. With reliability coefficient .in, Obtained by methods such as the coefficient of variation, it is used to allocate the influence intensity of different indicators on the evaluation during fusion; The reliability of evidence is estimated by historical accuracy or missing rate / noise level. The modified BPA is defined as follows: (17) In the formula, For the first The normalization factor for each piece of evidence. When When, satisfy ,in As evidence BPA was not rated at any probability level.

[0043] By utilizing the improved ER synthesis rule, the evaluation object is fused. All the indicators and evidence below. Assume there are two independent weighted pieces of evidence. and The joint reliability after fusion Represented as (18) (19) When the reliability of a piece of evidence When it decreases, the corresponding discount factor Increasing the value weakens the contribution of this evidence to the joint support; conversely, when... At higher levels, the fusion process relies more heavily on the commonality of evidence intersections. This enhances the accumulation of consistent information, suppresses the amplification of conflicting information, and achieves robust synthesis of multi-source uncertain information. (Assessment object) After normalizing all the indicator evidence, the probability distribution of the object at each security level is obtained. (20) S2.3 Multi-level aggregation considering structural and business heterogeneity Structural importance reflects the physical position of equipment or lines within the distribution network topology. (Define feeders) The following device set is ,equipment The rated capacity is The importance of the equipment structure is then expressed as (twenty one) The importance of a feeder structure is determined by its load-bearing capacity. (The definition of a feeder is missing from the original text.) The peak load is The feeder set is The importance of the feeder structure is expressed as: (twenty two) Different feeders or transformer substations have significantly different requirements for power quality and reliability, i.e., service heterogeneity. A user sensitivity coefficient is introduced. Indicating the service sensitivity of the feeder power supply object, then the feeder Business importance is expressed as (twenty three) The importance of the integrated structure and the importance of the business were evaluated. The overall hierarchical weight is (twenty four) In the formula, This is a weight fusion coefficient used to adjust the relative contribution of structural importance and business importance in the overall weight, reflecting the emphasis on network structure security or business sensitivity requirements under different operating scenarios.

[0044] (4) Sub-evidence propagation and multi-level fusion equipment Rank reliability This is considered as a sub-evidence of the feeder layer. Layer weights are taken into account. It can construct a strength discount for its sub-evidence. (25) In the formula, The comprehensive hierarchical weights calculated according to formula (24) are taken here as the equipment weights. Corresponding hierarchical weights; The device obtained from equation (20) At a safe level The reliability.

[0045] The evidence obtained by fusing the feeder's own indicators, together with the sub-evidence from all equipment inputs, is subjected to ER synthesis according to (17)–(20) to obtain the feeder's safety level confidence distribution. .

[0046] Similarly, the results of each feeder are weighted according to the overall level. The network layer sub-evidence is constructed and then synthesized with the inherent network layer indicator evidence to obtain the network security level reliability distribution. .

[0047] In a further embodiment, step S3 includes: S3.1 Risk Situation Prediction Based on Markov Chains Static assessments only provide the current state and are insufficient to describe whether the risk is worsening or when it might cross the danger threshold. To achieve proactive early warning, a Markov chain is introduced to characterize the transition pattern of the security level over time, and based on this, a predicted distribution for the next time step is obtained.

[0048] Set of security levels Viewed as a Markov chain state space. For the object... Construct the state transition probability matrix , represented as (26) In the formula, Indicates the object of evaluation From state Evolved to The probability. Matrix It can be obtained from historical running sequences or simulation sample sequences using maximum likelihood estimation. Based on Markov properties, the object... exist Predicted state distribution at time step Represented as (27) The prediction results reflect the potential trend of physical performance degradation or risk diffusion of an object without intervention.

[0049] In prospect theory, the reference point determines whether a decision-maker judges the outcome as a gain or a loss. Traditional fixed reference points are ill-suited to the dynamically changing risk environment of distribution networks. A standard utility score vector is defined. (Unified across all objects), at any time Expected utility during rest period and prediction time The expected utilities are respectively (28) Risk evolution trend factor is (29) when When this occurs, it indicates a decline in predictive utility and a tendency for the system to evolve towards a higher risk level. This indicates that the risk is stabilizing or improving. Let... As a safety reference point, then the dynamic reference point Updated to (30) In the formula, and This is the trend sensitivity coefficient. When... When risks worsen, the reference point needs to be raised to increase assessment sensitivity; when When the risk is stable, the trend will gradually revert to the benchmark after it improves.

[0050] S3.2 Dynamic Evaluation Model Based on Adaptive Foreground Theory Prospect theory categorizes security assessment results into gains or losses, with the division determined by a dynamic reference point. For the assessment object... The Security Level Its profit or loss relative to the reference point is (31) In the formula, For security level H n The corresponding standard utility score; The dynamic reference point is obtained by updating equation (30).

[0051] The value function measures the perceived gain or loss of the actual state relative to a reference point, transforming the level of security into a subjective gain or loss relative to a security benchmark. According to the classic Tversky-Kahneman form, the value function is expressed as... (32) In the formula, This represents the risk attitude coefficient, indicating diminishing marginal utility. The loss aversion coefficient reflects the characteristic that decision-makers are far more sensitive to security benefits (losses) than to security margins (gains).

[0052] The reliability of power distribution network monitoring data exhibits significant time-varying characteristics due to noise and missing data. According to the fuzzy aversion principle, when the reliability of assessment data is low, decision-makers tend to exhibit a stronger risk-averse tendency, resulting in a more drastic distortion in their probability perception; conversely, when the data is highly reliable, decisions revert to rational expectations. To ensure that decision weights reflect the quality of data support, the following definition is established. The overall confidence level is an aggregation of the reliability of its underlying indicators. (33) In the formula, For the evaluation object The complete set of underlying physical metrics included. For a moment index Reliability. Utilizing Dynamically adjust the curvature parameter of the decision weight function To achieve adaptive adjustment of curvature, i.e. (34) In the formula, This is the baseline curvature parameter in traditional prospect theory. When the reliability of key indicators decreases, the overall confidence level... Significantly reduced curvature The decrease in probability weighting enhances the nonlinear distortion of probability in the probability weighting function, manifesting as an upward adjustment of the relative weight of low-probability events, thus triggering a more conservative probability perception. Therefore, the probability weighting function is expressed as... (35) Furthermore, regarding The final decision weights are obtained through normalization. By combining the value function with dynamic decision weights, we obtain the time step... Evaluation object Overall security prospects (36) The larger the value, the more likely the object is to be larger. It is safer relative to the current benchmark. This is due to the introduction of a dynamic reference point. When an object shows a deteriorating trend, the reference point rises, leading to a change in the profit or loss value. Decrease or even turn into a negative value, thereby... Significant decrease, achieving early warning; adaptive curvature mechanism then enables They are more sensitive to tail risks, forming a security defense logic of "the more ambiguous the data, the more conservative the decision-making".

[0053] In a further embodiment, step S4 includes: S4.1 Risk Contribution Calculation Risk contribution is used to quantify the degree of impact of lower-level units on the decline of the overall prospect value of the upper-level system.

[0054] For the upper-level system Its lower-level unit set is defined as Then the first Risk contribution of each unit Represented as (37) In the formula, The foreground value represents the ideal safety condition, and is usually the foreground value corresponding to the highest assessment level. The larger the value, the more severe the deterioration of the unit's own condition and the more important its position in the system, making it a major factor contributing to the increased overall risk of the system.

[0055] S4.2 Two-level source tracing discrimination rules Based on risk contribution indicators, a two-level traceability process is established: network-feeder and feeder-equipment. Network-feeder tracing: When the network foreground value When the risk level is below the warning threshold, calculate the risk contribution of all feeders in the entire network. If it exists (38) Then determine the feeder These serve as risk feeders, forming a risk feeder set. .in, This is the feeder discrimination coefficient, used to control the sensitivity of the tracing results.

[0056] Feeder-Equipment Traceability: For each risky feeder Calculate the risk contribution of its subordinate equipment. If it exists (39) Then determine the device These are risky devices, constituting a set of faulty / vulnerable devices. .in, This is the equipment discrimination coefficient.

[0057] In summary, the two-level source tracing and discrimination rules start from the network layer critical alarm, first identifying the feeder that contributes the most, and then locating the equipment that contributes the most within the feeder, thus achieving a connection from macro-level situational awareness to micro-level fault location, supporting the overall assessment and reliable operation and maintenance of the distribution network.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for distribution network security assessment and tracing based on ER and prospect theory, characterized in that, Includes the following steps: S1: Construct a hierarchical safety assessment index system covering the equipment layer, feeder layer, and network layer; the equipment layer selects equipment load rate, frequency and duration of forward and reverse heavy overload of distribution transformers, and voltage qualification rate at grid connection point to characterize the operating pressure of a single device and the local voltage quality; the feeder layer selects feeder load rate, power backflow risk level, and feeder voltage deviation to characterize the power flow disturbance risk of multi-node coupling in the feeder; the network layer selects maximum power supply capacity, voltage over-limit frequency, and load balance degree to measure the overall power supply safety margin of the entire network. S2: Construct an improved two-stage ER assessment framework. In the first stage, the basic probability allocation (BPA) of each indicator generated within a single object is performed. The indicator importance coefficient and evidence reliability coefficient are used to complete the double correction. The improved ER rules are used to fuse and obtain the security level reliability distribution of a single object. In the second stage, the structural importance and service importance of equipment and feeders are calculated separately. The comprehensive hierarchical weight is obtained by coupling the weight fusion coefficients. The confidence of the lower layer assessment is discounted by strength and used as sub-evidence. The ER is aggregated from bottom to top to obtain the overall security level confidence distribution of the feeder layer and network layer. S3: Construct a dynamic evaluation model that couples Markov prediction with adaptive prospect theory. Construct a Markov chain state space with four security levels. Predict the security level distribution at the next moment based on the state transition probability matrix and calculate the risk evolution trend factor. Dynamically update the evaluation reference point by combining the current static expected utility, the predicted expected utility, and the trend sensitivity coefficient. Aggregate the reliability of the underlying indicators to obtain the comprehensive confidence level. Adaptively adjust the curvature parameter of the probability weight function of the prospect theory. The lower the data reliability, the greater the weight of low-probability, high-loss events. Combine the value function to solve the comprehensive security prospect value at each level. S4: Establish a two-level closed-loop tracing mechanism based on prospect value decomposition. When the comprehensive security prospect value of the network layer is lower than the warning threshold, the risk contribution of each lower-level unit is calculated based on the ideal security benchmark prospect value. First, perform network-feeder tracing to screen feeders exceeding the threshold for risk. Then, perform feeder-equipment tracing to locate the risky equipment for each risky feeder, completing the reverse risk location from the entire network to a single device, forming a closed loop of assessment, early warning, and risk tracing.

2. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that: In step S1, the equipment load rate is the ratio of the real-time load of the equipment to its rated capacity; the frequency of distribution transformer forward and reverse heavy overload is the number of transitions from power flow switching to heavy overload within the day; the duration of heavy overload is the cumulative duration of the load rate exceeding the threshold within the evaluation window; and the grid connection point voltage qualification rate is the percentage of the time the voltage falls within the allowable deviation range within the window.

3. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, Step S1: The risk level of power backfeeding in the feeder layer is constructed based on the utility function of the reverse active power at the feeder head end, and the risk weight of the high-power backfeeding condition is amplified by the risk preference coefficient. The feeder voltage deviation is a quantitative indicator of the overall deviation of the voltage at all nodes of the feeder relative to the reference voltage.

4. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, Step S1: The maximum power supply capacity of the network layer is the sum of the maximum load that the main transformers of the entire network can bear under the N-1 safety constraint. Load balance is characterized by the root mean square difference of the load rates of each feeder, which represents the degree of uneven load distribution across the entire network.

5. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, In step S2, the improved ER fusion rule is used to synthesize multiple sets of modified BPAs, discounting and weakening low-reliability conflict evidence, and cumulatively strengthening high-reliability consistent evidence. After fusion, the confidence distribution of each security level is output after normalization.

6. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, In step S2, the comprehensive hierarchical weight is obtained by weighted fusion of structural importance and business importance. Structural importance is based on the rated capacity of equipment and the peak load of feeders to characterize the degree of topological criticality, while business importance is based on the load sensitivity coefficient to characterize the reliability requirements of power supply services.

7. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, The rule for updating the dynamic reference point in step S3 is as follows: if the risk evolution trend factor is less than 0, it means that the risk has worsened, so the dynamic reference point is raised to improve the assessment sensitivity; if the trend factor is greater than or equal to 0, it means that the risk is stable or improving, so the dynamic reference point gradually falls back to the preset safety benchmark.

8. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, The adaptive curvature adjustment logic in step S3 is as follows: when the overall confidence level decreases, the curvature of the probability weight function decreases, which enhances the perception of extreme risks at the tail end and realizes the defense assessment logic that the more fuzzy the monitoring data, the more conservative the assessment result.

9. The distribution network security assessment and tracing method based on ER and prospect theory according to claim 1, characterized in that, Step S4, Risk Contribution, is used to quantify the degree to which lower-level units cause a decrease in the security prospect value of the upper-level system. Feeders and devices with a Risk Contribution higher than the corresponding discrimination coefficient threshold are identified as risk sources. The discrimination coefficient can be flexibly adjusted to improve the source tracing sensitivity according to the operation and maintenance scenario.

10. A distribution network security assessment and tracing system based on ER and prospect theory, characterized in that, include: The data acquisition submodule is used to collect telemetry and teleindication multi-source operational data from the equipment layer, feeder layer, and network layer, and to complete data preprocessing. The index calculation and BPA generation submodule is used to calculate the security assessment indexes at each level of the device layer, feeder layer, and network layer, and generate the basic probability allocation (BPA) corresponding to each security level through membership function mapping. The Evidence Correction and ER Fusion submodule is used to correct the BPA based on the indicator importance coefficient and the evidence reliability coefficient, and to obtain the safety level reliability distribution of the assessed object by fusion using improved ER rules. The multi-level aggregation submodule is used to construct a comprehensive hierarchical weight based on structural importance and business importance, and aggregate the lower-level evaluation results as sub-evidence from bottom to top to obtain the upper-level evaluation reliability distribution. The Markov prediction submodule is used to predict the safety level distribution at the next moment based on the safety level state transition probability matrix and to calculate the risk evolution trend factor. The dynamic reference point update submodule is used to dynamically update the dynamic reference points for safety assessment based on risk evolution trend factors. The adaptive prospect scoring submodule is used to adaptively adjust the curvature parameter of the prospect theoretical probability weight function based on the comprehensive confidence of the underlying indicators, and calculate the comprehensive safety prospect value of each level of the evaluation object by combining the value function. The risk tracing submodule is used to calculate the risk contribution of lower-level units based on prospective value decomposition, perform two-level tracing and discrimination between network and feeder and feeder and equipment, and locate key risk feeders and risk equipment. The results output submodule is used to output the comprehensive security prospect value, security level judgment result, and risk tracing and location result of each layer of the device layer, feeder layer, and network layer. The communication submodule is used to upload security assessment data, risk warning information, and source tracing results to the dispatch center.

Images