Risk early warning method, system and equipment for power distribution network and medium

Abstract

The invention relates to a risk early warning method, system and device for a power distribution network and a medium, and the method comprises the steps: obtaining the historical operation data of the power distribution network, and constructing a multi-stage risk assessment matrix according to the historical operation data; obtaining multi-source operation data at the current moment, performing uncertainty modeling on the multi-source operation data based on the multi-level risk assessment matrix, and generating a multi-source evidence set; performing evidence fusion operation on the multi-source evidence set to obtain a first risk assessment result; under the condition that the first risk assessment result does not meet a preset stable condition, knowledge retrieval reasoning is executed based on the first risk assessment result and the multi-source operation data to obtain a knowledge reasoning result, and then the knowledge reasoning result and the multi-source evidence set are combined to obtain an enhanced evidence set; and performing secondary evidence fusion operation on the enhanced evidence set to obtain a final risk early warning result. The method has the effect of improving the risk assessment accuracy of the power distribution network.

Description

Technical Field

[0001] This invention belongs to the technical field of power system operation and safety control, and in particular relates to a risk early warning method, system, equipment and medium for power distribution networks. Background Technology

[0002] Currently, with the large-scale integration of distributed power sources, new energy sources, and electric loads, the operating conditions of distribution networks are becoming increasingly complex. Information such as equipment status, load levels, protection actions, and meteorological conditions is characterized by multi-source nature, time-varying changes, and increased uncertainty. Once a fault or anomaly occurs in the distribution network, it can easily lead to large-scale power outages and a decline in power quality. Therefore, it is necessary to identify and warn of potential risks in advance during operation to support refined operation and maintenance and fault prevention.

[0003] Existing risk assessment methods for power distribution networks often employ threshold determination and scoring based on a single or a few operational indicators, or introduce statistical or machine learning models to fit and predict historical fault data, or summarize multi-source data through simple weighting or rule-based logic to obtain risk levels. These methods rely on manual experience to set rules and thresholds, lack a unified characterization of different types of uncertainties such as equipment lifespan, event frequency, and real-time operating status, and handle conflicts and contradictions between multi-source data in a rather crude manner. This results in insufficient stability and accuracy of risk assessment results in complex operating scenarios, making it difficult to provide operation and maintenance personnel with targeted risk assessment basis and handling references.

[0004] The existing technical solutions mentioned above have the following drawbacks: the existing distribution network risk assessment methods lack accuracy and stability in the modeling and fusion processing of uncertainties in multi-source operation data, and therefore there is room for improvement. Summary of the Invention

[0005] The purpose of this invention is to provide a risk early warning method, system, device and medium for power distribution networks, so as to solve the technical problems of insufficient accuracy and stability of existing power distribution network risk assessment methods in terms of uncertainty modeling and fusion processing of multi-source operating data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a risk early warning method for a power distribution network, the method comprising: Obtain historical operating data of the power distribution network, and construct a multi-level risk assessment matrix based on the historical operating data; Acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-level risk assessment matrix based on the multi-source operational data, and generate a multi-source evidence set; Evidence fusion is performed on the multi-source evidence set to obtain the first risk assessment result; If the first risk assessment result does not meet the preset stability conditions, knowledge retrieval reasoning is performed based on the first risk assessment result and the multi-source operation data to obtain the knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain the enhanced evidence set. A second evidence fusion operation is performed on the enhanced evidence set to obtain the final risk warning result.

[0007] By adopting the above technical solution, and by acquiring historical operation data of the distribution network and constructing a multi-level risk assessment matrix, the relationship between risk factors and risk levels can be uniformly described at multiple levels, including equipment, lines, and regions, thus providing a structured and quantifiable basic model for subsequent risk assessment. By acquiring multi-source operation data at the current moment and generating a multi-source evidence set based on uncertainty modeling using the multi-level risk assessment matrix, operational information from different sources and of different types can be uniformly converted into evidence, thereby comprehensively characterizing the uncertainty of the current state of the distribution network. By performing evidence fusion calculations on the multi-source evidence set to obtain the first risk assessment result, the overall multi-source information can be used to assess the target... The system performs preliminary risk quantification and explicitly provides the uncertainty, thus avoiding the one-sidedness of assessment by a single indicator. When the first risk assessment result is unstable, it performs knowledge retrieval reasoning based on the first risk assessment result and multi-source operational data, and merges it with the multi-source evidence set to obtain an enhanced evidence set. This allows for the introduction of historical cases and knowledge graphs to enhance knowledge in high-uncertainty scenarios, thereby improving the reliability and interpretability of the risk assessment results. By performing secondary evidence fusion operations on the enhanced evidence set, the final risk warning result is obtained. This results in a more accurate and stable risk level output based on the fusion of original evidence and knowledge evidence, thus providing a highly credible warning basis for operation and maintenance decisions.

[0008] In one example, the present invention can be further configured as follows: the construction of a multi-level risk assessment matrix based on the historical operational data includes: Based on the distribution network ledger information and the historical operation data of the distribution network, the equipment, lines and areas in the distribution network are layered to determine the risk assessment objects of each layer; The characteristic data corresponding to the risk assessment object is obtained from the historical operation data of the distribution network. The characteristic data is used as the risk factor of the risk assessment object. Then, the risk factor is associated with the corresponding risk level proposition to obtain the matrix elements of the risk assessment object. The matrix elements are arranged in rows and columns according to the risk assessment object, and weight parameters and judgment thresholds are set for each matrix element. Regression analysis is performed on the weight parameters and judgment thresholds to calibrate them, thereby obtaining the multi-level risk assessment matrix.

[0009] By adopting the above technical solution, risk assessment objects can be determined by classifying equipment, lines and areas based on distribution network ledger information and historical operation data, risk factors can be extracted from historical operation data and established with risk level propositions, and regression analysis can be performed to calibrate the weight parameters and judgment thresholds of matrix elements. This enables the construction of a multi-level risk assessment matrix that matches the actual topology and fault characteristics of the distribution network, thereby improving the adaptability of the risk assessment model to objects at different levels and the accuracy of its characterization of the real risk distribution.

[0010] In one example, the present invention can be further configured as follows: the uncertainty modeling of the multi-source operational data based on the multi-level risk assessment matrix to generate a multi-source evidence set includes: The multi-source operational data is divided according to the risk assessment object to obtain several evidence bodies; Based on the multi-level risk assessment matrix, the operational indicators in the evidence body are matched with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition. The membership degree is normalized and mapped within a unified risk level proposition space to obtain the basic probability allocation of each piece of evidence on each risk level proposition, thereby generating the multi-source evidence set.

[0011] By adopting the above technical solution, the evidence body is obtained by dividing multi-source operational data according to the risk assessment object. The operational indicators in the evidence body are matched with risk factors based on the multi-level risk assessment matrix to obtain the membership degree on the risk level proposition. The membership degree is normalized and mapped to the basic probability allocation of each risk level proposition to generate a multi-source evidence set. This can systematically convert complex multi-source operational data into the BPA form that conforms to the evidence theory, thereby providing a unified data expression basis for subsequent evidence fusion and uncertainty quantification.

[0012] In one example, the present invention can be further configured as follows: based on the multi-level risk assessment matrix, matching the operational indicators in the evidence body with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition includes: Based on the operational indicators characterizing equipment lifespan and aging degree in the evidence body, the failure occurrence time characteristics are fitted by the Weibull distribution model to obtain the probability of occurrence of each risk level proposition in different time intervals, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition. Based on the operational indicators representing the frequency of fault events in the evidence body, the number of events arriving per unit time is fitted using a Poisson distribution model to obtain the probability of occurrence of each risk level proposition under different event frequencies, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition. Based on the continuous operational indicators representing the current operational status in the evidence body, the values ​​of the operational indicators are mapped to the membership degrees of each risk level proposition through a preset fuzzy membership function.

[0013] By adopting the above technical solutions, the following methods are used: Weibull distribution model is used to fit the fault occurrence time characteristics based on the operating indicators representing equipment lifespan and aging degree in the evidence body; Poisson distribution model is used to fit the event arrival process based on the operating indicators representing the frequency of fault events in the evidence body; and a preset fuzzy membership function is used to map the continuous operating indicators representing the current operating state in the evidence body to membership degrees. This allows for the precise characterization of the support for risk level propositions by selecting appropriate statistical and fuzzy modeling methods for lifespan-type, frequency-type, and continuous state-type indicators. This more realistically reflects the impact of different types of operating characteristics on risk and improves the accuracy of uncertainty modeling results.

[0014] In one example, the present invention can be further configured as follows: performing evidence fusion operation on the multi-source evidence set to obtain a first risk assessment result includes: The basic probability assignments are fused to obtain the fusion probability and the corresponding conflict coefficient; The fusion probability is subjected to trust and credibility calculations to obtain the trust and credibility of the target risk level proposition, and a confidence interval is constructed based on the trust and credibility. The first risk assessment result is generated based on the conflict coefficient and the confidence interval.

[0015] By employing the aforementioned technical solution, the fusion probability and corresponding conflict coefficient are obtained by fusing the basic probability allocations in a multi-source evidence set. This allows for the comprehensive integration of support information from different evidence bodies for the risk level proposition and the quantification of consistency between evidence, thereby identifying potential contradictions and conflicts while integrating multi-source information. By calculating the trust and credibility of the target risk level proposition based on the fusion probability and constructing a confidence interval, the credibility range of the risk assessment result can be represented in interval form, thus providing a basis for subsequent stability determination and threshold setting. By combining the conflict coefficient and the confidence interval to generate the first risk assessment result, a preliminary risk level can be output under the premise of considering the degree of evidence conflict and the magnitude of uncertainty, thereby avoiding blindly giving overly certain conclusions when the evidence quality is poor.

[0016] In one example, the present invention can be further configured as follows: when the first risk assessment result does not meet the preset stability conditions, knowledge retrieval reasoning is performed based on the first risk assessment result and the multi-source operational data to obtain a knowledge reasoning result, and then the knowledge reasoning result and the multi-source evidence set are merged to obtain an enhanced evidence set, including: The validity is checked based on the confidence interval and the conflict coefficient. When the interval width is greater than the preset stable interval threshold or the conflict coefficient is greater than the preset conflict threshold, the current evaluation scenario is determined to be an unstable scenario. Based on the unstable scenario, scenario feature information is extracted from the multi-source operational data and the first risk assessment result, and a query vector is constructed based on the scenario feature information. Based on the query vector, similar scenarios are retrieved in the historical event database, and related entities and relationships are retrieved in the power distribution network knowledge graph. The retrieval results are then input into a preset retrieval enhancement generation model to obtain knowledge reasoning results. The knowledge reasoning results are analyzed to obtain knowledge evidence. The basic probability distribution of the knowledge evidence is then combined with the basic probability distribution of each piece of evidence in the multi-source evidence set to obtain an enhanced evidence set.

[0017] By adopting the above technical solutions, and by using confidence interval width and conflict coefficient to verify the validity of unstable scenarios, it is possible to automatically identify objects with unreliable assessment results or excessive uncertainty, thereby triggering a more refined knowledge enhancement process in critical scenarios. In unstable scenarios, by extracting scenario feature information from multi-source operational data and the first risk assessment results to construct query vectors, and then retrieving similar knowledge from historical event databases and distribution network knowledge graphs, the results are input into the retrieval enhancement generation model to obtain knowledge reasoning results. This allows for supplementary reasoning of current risk scenarios based on rich historical experience and structured knowledge, thereby obtaining knowledge evidence regarding the causes and evolution trends of risks. By parsing the knowledge reasoning results to form knowledge evidence and merging it with the basic probability allocation in the multi-source evidence set to obtain an enhanced evidence set, the semantic-level knowledge reasoning results can be transformed into quantifiable evidence information that can participate in fusion, thereby improving the robustness and explanatory power of the overall risk assessment while ensuring formal consistency.

[0018] In one example, the present invention can be further configured such that the method further includes: Record the risk level corresponding to the final risk warning result, obtain the actual fault occurrence situation corresponding to the risk level during operation, and obtain operation feedback data by comparing the difference between the final risk warning result and the actual fault occurrence situation; Based on the operational feedback data, an evaluation index is constructed. The weight parameters and judgment thresholds in the multi-level risk assessment matrix are corrected according to the evaluation index. The weight parameters of each piece of evidence are adjusted according to the contribution of different pieces of evidence to the evaluation index, resulting in an updated multi-level risk assessment matrix and evidence weights.

[0019] By adopting the above technical solution, recording the risk level corresponding to the final risk warning result and obtaining the actual fault occurrence corresponding to that risk level during operation, and comparing the two to form operational feedback data, a historical sample set reflecting the warning hit, false alarm, and missed alarm situations can be constructed, thus providing an objective basis for subsequent model evaluation and parameter adjustment. By constructing evaluation indicators based on operational feedback data, and accordingly correcting the weight parameters and judgment thresholds in the multi-level risk assessment matrix and adjusting the weight parameters of different evidence bodies, online adaptive optimization of the risk assessment model and evidence weights can be achieved, thereby enabling the system to continuously conform to the actual risk level of the distribution network and improve the accuracy and reliability of the warning results during long-term operation.

[0020] In a second aspect, the present invention provides a risk early warning system for a power distribution network, the system comprising: The historical modeling module is used to acquire historical operating data of the distribution network and construct a multi-level risk assessment matrix based on the historical operating data. The evidence modeling module is used to acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-source operational data based on the multi-level risk assessment matrix, and generate a multi-source evidence set. The initial assessment fusion module is used to perform evidence fusion operations on the multi-source evidence set to obtain the first risk assessment result; The knowledge enhancement module is used to perform knowledge retrieval reasoning based on the first risk assessment result and the multi-source operational data when the first risk assessment result does not meet the preset stability conditions, to obtain a knowledge reasoning result, and then merge the knowledge reasoning result and the multi-source evidence set to obtain an enhanced evidence set. The final evaluation fusion module is used to perform secondary evidence fusion operations on the enhanced evidence set to obtain the final risk warning result.

[0021] By adopting the above technical solution, and by acquiring historical operation data of the distribution network and constructing a multi-level risk assessment matrix, the relationship between risk factors and risk levels can be uniformly described at multiple levels, including equipment, lines, and regions, thus providing a structured and quantifiable basic model for subsequent risk assessment. By acquiring multi-source operation data at the current moment and generating a multi-source evidence set based on uncertainty modeling using the multi-level risk assessment matrix, operational information from different sources and of different types can be uniformly converted into evidence, thereby comprehensively characterizing the uncertainty of the current state of the distribution network. By performing evidence fusion calculations on the multi-source evidence set to obtain the first risk assessment result, the overall multi-source information can be used to assess the target... The system performs preliminary risk quantification and explicitly provides the uncertainty, thus avoiding the one-sidedness of assessment by a single indicator. When the first risk assessment result is unstable, it performs knowledge retrieval reasoning based on the first risk assessment result and multi-source operational data, and merges it with the multi-source evidence set to obtain an enhanced evidence set. This allows for the introduction of historical cases and knowledge graphs to enhance knowledge in high-uncertainty scenarios, thereby improving the reliability and interpretability of the risk assessment results. By performing secondary evidence fusion operations on the enhanced evidence set, the final risk warning result is obtained. This results in a more accurate and stable risk level output based on the fusion of original evidence and knowledge evidence, thus providing a highly credible warning basis for operation and maintenance decisions.

[0022] In a third aspect, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the aforementioned risk warning method for a power distribution network.

[0023] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the aforementioned risk warning method for a power distribution network. Attached Figure Description

[0024] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of a risk warning method for a power distribution network in an embodiment of the present invention; Figure 2 This is a structural block diagram of a risk early warning system for a power distribution network according to an embodiment of the present invention; Figure 3 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0026] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0027] Example 1 like Figure 1 As shown, this invention discloses a risk early warning method for power distribution networks, specifically including the following steps: S10: Obtain historical operating data of the distribution network and construct a multi-level risk assessment matrix based on the historical operating data.

[0028] Specifically, historical operating data of the distribution network is acquired, including long-term equipment operation records, fault and maintenance archives, load and power quality statistics, and meteorological and environmental information. The historical operating data is organized and cleaned according to time, space, and equipment objects, and obviously abnormal data and severely missing data are removed. By statistically analyzing the operating behavior and fault performance of objects at different levels, and combining them with pre-set risk assessment dimensions, risk-related indicators are abstracted into several risk assessment factors. The risk assessment factors are then organized into a matrix according to the hierarchical relationship of equipment layer, line layer, and regional layer, thus forming a multi-level risk assessment matrix.

[0029] S20: Obtain multi-source operational data at the current moment, perform uncertainty modeling on the multi-source operational data based on the multi-level risk assessment matrix, and generate a multi-source evidence set.

[0030] Specifically, multi-source operational data at the current moment is obtained from the power distribution network dispatch and monitoring platform, protection system, online monitoring terminal and meteorological service interface, including equipment status, operating condition, protection and alarm information and external environmental data. The multi-source operational data is mapped to the risk assessment objects and risk factors defined in the multi-level risk assessment matrix. A unified uncertainty representation method is used to transform and standardize different types of data. Deterministic values, statistical values ​​and fuzzy values ​​are uniformly expressed as the support degree for propositions of each risk level, thereby forming an uncertainty description containing multiple evidence bodies, and the evidence bodies are summarized into a multi-source evidence set.

[0031] S30: Perform evidence fusion calculation on the multi-source evidence set to obtain the first risk assessment result.

[0032] Specifically, the support for risk level propositions corresponding to each evidence body in the multi-source evidence set is used as input. Support information from different evidence bodies is combined according to pre-set evidence fusion rules. The support for the same risk level proposition is superimposed and the support for mutually exclusive propositions is normalized to obtain the comprehensive measurement value of each risk level proposition and the overall assessment. The comprehensive measurement result is used as the first risk assessment result to characterize the preliminary risk level and corresponding credibility of the current distribution network object.

[0033] S40: If the first risk assessment result does not meet the preset stability conditions, knowledge retrieval reasoning is performed based on the first risk assessment result and multi-source operational data to obtain the knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain the enhanced evidence set.

[0034] Specifically, when an unstable or uncertain situation is determined based on the first risk assessment result, a scenario description for retrieval is constructed based on the risk level information contained in the first risk assessment result and the corresponding multi-source operational data. This scenario description is then submitted to a preset knowledge retrieval and reasoning module, which retrieves records similar to or related to the current scenario from historical events, operational experience, and procedural knowledge. Based on the retrieval results, reasoning is performed to generate knowledge reasoning results about the causes of risk and handling strategies. These knowledge reasoning results are then transformed into new evidence information and merged with the original multi-source evidence set to obtain an enhanced evidence set containing original evidence and knowledge evidence.

[0035] S50: Perform a secondary evidence fusion operation on the enhanced evidence set to obtain the final risk warning result.

[0036] Specifically, the enhanced evidence set is subjected to a secondary evidence fusion operation according to Dempster's combination rule to obtain the enhanced comprehensive basic probability allocation, i.e., m. final The combinatorial result on the risk event proposition F is denoted as m. final (F)=m total ⊕m RAG , where m total The overall basic probability assignment obtained after the first evidence fusion, m RAG To assign the basic probability of knowledge evidence obtained based on RAG knowledge reasoning to each proposition, and then according to The confidence level Bel'(F) and likelihood Pl'(F) of the enhanced risk event proposition F are calculated in a certain way, forming the enhanced confidence interval [Bel'(F),Pl'(F)], and the confidence level of the rule evidence output by the rule matching module is denoted as Conf. r The confidence level of case evidence obtained based on case retrieval and reasoning is denoted as Conf. cThe confidence level of real-time evidence formed by real-time monitoring and sensing information is denoted as Conf. s The three types of evidence are assigned weights w respectively. r w c w s And satisfy w r +w c +w s =1, using the weighted logodds method to calculate rule-based evidence, case evidence, and real-time evidence. Press again Calculate the overall logit value and obtain the final confidence level based on it. Finally, considering the position and width of the enhanced confidence interval [Bel'(F),Pl'(F)] and Conf... final The numerical value is used to set thresholds for multiple risk levels, and Conf final Objects that are relatively large and have a high overall [Bel'(F),Pl'(F)] are classified as high-risk objects, and Conf is classified as high-risk objects. final Objects that are relatively small and have a generally low risk level in their corresponding intervals are classified as low-risk objects. The final risk warning result is output, which includes the risk level, the enhanced confidence interval, the association rules, and the case explanation information.

[0037] In one embodiment, step S10, namely constructing a multi-level risk assessment matrix based on historical operational data, includes: S11: Based on the distribution network ledger information and historical operation data of the distribution network, the equipment, lines and areas in the distribution network are divided into layers to determine the risk assessment objects of each layer.

[0038] Specifically, by utilizing the distribution network ledger information, the topological and electrical connections of each feeder, transformer substation, switching equipment, and its associated users and areas in the distribution network are determined. The overall distribution network is divided into multiple levels, such as the equipment layer, line layer, and area layer. The equipment layer includes individual equipment objects such as transformers, circuit breakers, cables, and switchgear. The line layer includes power supply lines or substations composed of multiple devices connected in series or parallel. The area layer includes multiple lines or substations aggregated according to geographical regions or power supply ranges. The objects identified in each level are numbered and identified as risk assessment objects, resulting in a multi-layered risk assessment object set covering the main structure of the distribution network.

[0039] S12: Obtain the characteristic data corresponding to the risk assessment object from the historical operation data of the distribution network, use the characteristic data as the risk factors of the risk assessment object, and then establish a correspondence between the risk factors and the corresponding risk level propositions to obtain the matrix elements of the risk assessment object.

[0040] Specifically, for each risk assessment object, characteristic data related to the risk assessment object are extracted from the historical operation data of the distribution network, including the number of years of operation, cumulative number of faults, various types of faults, power outage duration, load factor, power quality indicators, operating environment indicators, and meteorological factors. The characteristic data are abstracted into risk factors describing different risk dimensions. A correspondence is established between each risk factor and a predefined risk level proposition. For example, a positive relationship is established between high load factor and high risk level proposition, and a positive relationship is established between fewer faults and low risk level proposition. Based on this, the contribution direction and initial strength of each risk factor on each risk level proposition are determined, and matrix elements are obtained.

[0041] S13: Arrange the matrix elements in rows and columns according to the risk assessment object, set weight parameters and judgment thresholds for each matrix element, and perform regression analysis to calibrate the weight parameters and judgment thresholds to obtain a multi-level risk assessment matrix.

[0042] Specifically, the matrix elements are arranged in rows and columns according to the hierarchy and numbering order of the risk assessment objects, with the risk assessment objects as rows and the risk factors as columns. This ensures that risk factors of the same level or category have consistent positions in the matrix. For each matrix element, a weight parameter reflecting its importance and a judgment threshold for distinguishing risk levels are set. Then, based on the actual failure occurrence and power outage impact of each risk assessment object in historical operating data, the risk level output by the matrix is ​​compared with the actual risk performance. Regression analysis or other statistical fitting methods are used to calibrate the weight parameters and judgment thresholds, so that the matrix is ​​more statistically consistent with the risk distribution pattern of historical data, thus obtaining a multi-level risk assessment matrix.

[0043] In one embodiment, step S20, namely, performing uncertainty modeling on multi-source operational data based on a multi-level risk assessment matrix to generate a multi-source evidence set, includes: S21: Divide the multi-source operational data according to the risk assessment object to obtain several evidence bodies.

[0044] Specifically, based on the risk assessment objects identified in the multi-level risk assessment matrix, the multi-source operation data at the current moment are classified by object. The operation monitoring quantities, protection alarm records, operation condition information and environmental measurement values ​​belonging to the same equipment, the same line or the same area are aggregated so that each type of aggregated data corresponds to an evidence body. Each evidence body contains a set of operation indicators and status information related to a certain risk assessment object, thereby obtaining several evidence bodies corresponding to different risk assessment objects or different information sources.

[0045] S22: Based on the multi-level risk assessment matrix, the operational indicators in the evidence body are matched with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition.

[0046] Specifically, based on the correspondence between risk factors and operational indicators in the multi-level risk assessment matrix, the operational indicator value at the current moment is read from each evidence body, and the operational indicators are associated with the risk factors of the same risk assessment object one by one. For indicators related to lifespan and aging, indicators related to event frequency, and indicators related to continuous operation status, appropriate statistical or fuzzy modeling methods are selected respectively to calculate the influence intensity of various operational indicators on different risk level propositions, and the influence intensity is expressed as the membership degree of the evidence body on the risk level proposition.

[0047] S23: Normalize the membership degree within a unified risk level proposition space to obtain the basic probability allocation of each piece of evidence on each risk level proposition, thereby generating a multi-source evidence set.

[0048] Specifically, firstly, within a unified risk level proposition space, various membership degrees are transformed into basic probability assignments (BPAs) for different types of evidence, including: for rule-based evidence, a matching score (S) is obtained based on the matching degree between the current event characteristics and the expert knowledge rule base. rule Define the rule matching function Set the reliability adjustment coefficient The basic probability distribution of the evidence for the calculation rule is as follows: , where θ represents an unknown or uncertain proposition without distinguishing risk levels; For evidence based on historical cases, a similarity score (sim) is calculated based on the similarity between the characteristics of the current event and previous failure cases. case ,definition Set the reliability coefficient of case evidence The basic probability distribution of the case evidence is calculated as follows: ; For real-time sensor-based evidence, taking temperature as an example, we obtain the membership degree of the risk level proposition related to temperature during operation, and use this membership degree as the basic probability assignment of the fault event proposition F, letting... ; For meteorological and environmental evidence, an environmental quantitative score S is calculated based on the correlation between meteorological data such as wind speed and rainfall intensity and the occurrence of the fault. weather Let environmental evidence weighting coefficients be set. The basic probability allocation of environmental evidence is defined as follows: ; The basic probability allocations of rule-based evidence, case evidence, real-time operational evidence, and environmental evidence on risk level propositions are obtained through the above methods. The allocation results of each evidence body are then organized and summarized within a unified risk level proposition space, thus obtaining the basic probability allocations of each evidence body on each risk level proposition and generating a multi-source evidence set.

[0049] In one embodiment, in step S22, based on the multi-level risk assessment matrix, the operational indicators in the evidence body are matched with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition, including: S221: Based on the operational indicators characterizing equipment lifespan and aging in the evidence body, the failure occurrence time characteristics are fitted using the Weibull distribution model to obtain the probability of occurrence of each risk level proposition in different time intervals, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition.

[0050] Specifically, based on operational indicators characterizing equipment lifespan and aging levels, such as years of operation, within the evidence body, and assuming the random variable of equipment lifespan is t, a Weibull distribution model is used to model the uncertainty of failure occurrence time, with its probability density function being: η is a shape parameter and a scale parameter. When β < 1, it corresponds to early failure, and the failure rate decreases with time. When β = 1, it corresponds to random failure, and the failure rate does not change much with time. When β > 1, it corresponds to wear failure, and the failure rate increases with time. The larger η is, the longer the overall lifespan of the equipment. The failure probability within the selected evaluation time interval is obtained by integrating or accumulating the lifespan index of the current equipment. The failure probability is divided into piecewise probabilities according to different risk level propositions and normalized and mapped to the membership degree of the corresponding risk level proposition.

[0051] S222: Based on the operational indicators representing the frequency of fault events in the evidence body, the number of event arrivals per unit time is fitted using a Poisson distribution model to obtain the probability of occurrence of each risk level proposition under different event frequencies, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition.

[0052] Specifically, based on operational indicators representing the frequency of fault events in the evidence body, such as the number of trips or over-limits per unit time, let the number of event arrivals within the time interval t be k. A Poisson distribution model is used to model the event arrival process, letting... λ is the average event arrival rate per unit time. The value of λ is estimated by historical event data. Then, the probability of event occurrence under different values ​​of k is calculated based on the number of events in the current observation interval. The probability quality corresponding to high event frequency is mapped to higher risk level propositions, and the probability quality corresponding to low event frequency is mapped to lower risk level propositions. The probabilities on each risk level proposition are normalized to obtain the corresponding membership degree.

[0053] S223: Based on the continuous operational indicators representing the current operational status in the evidence body, the values ​​of the operational indicators are mapped to the membership degrees of each risk level proposition through a preset fuzzy membership function.

[0054] Specifically, based on continuous operational indicators representing the current operating status in the evidence body, such as voltage deviation, load factor, and power quality indicators, a fuzzy membership function μ(x) or piecewise linear mapping function corresponding to each risk level proposition is preset for each type of continuous indicator. Taking temperature T as an example, the normal upper limit temperature can be set as T0, and the over-limit temperature difference can be set as ΔT. The membership function is defined. When T≤T0, μ(T)=0. When T increases with the increase of the over-limit, μ(T) varies between 0 and 0.9. μ(T) is used as the membership degree of the corresponding risk level proposition. The membership degree of other continuous operation indicators on each risk level proposition is calculated according to their respective preset fuzzy functions or interval mapping functions. The membership degrees of multiple continuous indicators can be weighted and combined to obtain the comprehensive membership degree of the evidence on each risk level proposition.

[0055] In one embodiment, step S30, which involves performing evidence fusion on the multi-source evidence set to obtain a first risk assessment result, includes: S31: The basic probability assignments are fused to obtain the fusion probability and the corresponding conflict coefficient.

[0056] Specifically, taking the basic probability allocation of each piece of evidence in the multi-source evidence set on the risk level proposition as input, let the initial comprehensive BPA be... The basic probability allocation corresponding to rule-based evidence, case evidence, sensor evidence, and environmental evidence. Sequentially combine the current composite BPA with the current BPA in Dempster's order. For any risk event proposition F, the combined composite BPA satisfies... Where K is the conflict coefficient, and A and B are the proposition sets of two pieces of evidence to be fused, respectively. This is achieved by iteratively updating m... total Until all evidence has been processed, the fusion probability m for each risk level proposition is obtained. total (F) and the corresponding conflict coefficient K, the conflict coefficient K ranges from [0,1], the larger K is, the worse the consistency between different evidence bodies.

[0057] S32: Calculate the confidence level and credibility of the fusion probability to obtain the confidence level and credibility of the target risk level proposition, and construct a confidence interval based on the confidence level and credibility.

[0058] Specifically, based on the fusion probability m total Using the identification framework consisting of each risk level proposition and its subset as the summation range, for each target risk level proposition F, calculate its confidence level Bel(F) and credibility level Pl(F). The confidence level Bel(F) represents the cumulative support of all propositions that are completely contained in F for F, and the credibility Pl(F) represents the maximum possible support of all propositions that have a non-empty intersection with F for F. Bel(F) and Pl(F) are used as the lower and upper bounds respectively to form the confidence interval [Bel(F),Pl(F)].

[0059] S33: Generate the first risk assessment result based on the conflict coefficient and confidence interval.

[0060] Specifically, the consistency among current multi-source evidence is evaluated based on the conflict coefficient K. K is compared with a preset conflict threshold interval to reflect the stability of the evidence fusion results. Risk level propositions are sorted according to Bel(F) or the median of the interval based on the confidence interval [Bel(F),Pl(F)]. Risk level propositions with larger Bel(F) or higher overall intervals are selected as the initial risk level of the current object. At the same time, to verify the reliability of the confidence interval, historical operating samples are used to divide the data into groups according to the Bel(F) interval, and the actual number of failures O in each group is counted. i And based on the total number of samples N and the expected failure probability p of that group i Calculate the expected number of failures = The chi-square test statistic was used. The reliability of the confidence interval is tested. When x² is less than the preset critical value, the current confidence interval is considered to be consistent with the actual fault distribution. When x² is greater than the critical value, the current assessment is considered to have a deviation. After combining the conflict coefficient K, the confidence interval [Bel(F),Pl(F)] and the chi-square test results, the selected risk level proposition, the corresponding interval form confidence index and the conflict coefficient are combined to generate the first risk assessment result.

[0061] In one embodiment, in step S40, when the first risk assessment result does not meet the preset stability condition, knowledge retrieval reasoning is performed based on the first risk assessment result and multi-source operational data to obtain a knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain an enhanced evidence set, including: S41: Perform a validity test based on the confidence interval and the conflict coefficient. If the interval width is greater than the preset stable interval threshold or the conflict coefficient is greater than the preset conflict threshold, the current evaluation scenario is determined to be an unstable scenario.

[0062] Specifically, the confidence intervals [Bel(F),Pl(F)] and corresponding conflict coefficients K of each target risk level proposition F in the first risk assessment results are read. The interval width w(F) = Pl(F) - Bel(F) of each risk level proposition is calculated, and the maximum width w is taken. maxAs a measure of overall uncertainty in the current evaluation scenario, w max Compared with the preset stable interval threshold, when w max The current evaluation result is considered stable when ε ≤ w and the conflict coefficient K is in the low conflict range; when ε < w max If w ≤ ε' or the conflict coefficient K is in the medium conflict range, such as 0.4 < K ≤ 0.7, the evaluation result is considered to have some instability. max If the conflict coefficient K is in the high conflict range (e.g., K > 0.7), the evaluation result is considered highly unstable. max If the current evaluation scenario of the corresponding object is marked as an unstable scenario when the conflict coefficient is greater than the stable interval threshold or the conflict coefficient is greater than the preset conflict threshold, the current evaluation scenario of the corresponding object will be marked as an unstable scenario to trigger subsequent knowledge retrieval and reasoning processing. Here, ε is the preset stable threshold for the confidence interval width and ε' is the preset upper limit threshold for the confidence interval width.

[0063] S42: Based on unstable scenarios, extract scenario feature information from multi-source operational data and the first risk assessment results, and construct a query vector based on the scenario feature information.

[0064] Specifically, for assessment objects marked as unstable scenarios, the equipment identifier, line and area information, measurement time and time window, current load status, protection and alarm records, and external meteorological environmental characteristics related to the object are extracted from multi-source operational data. Combined with the object's preliminary risk level, corresponding confidence interval width, and conflict coefficient in the first risk assessment results, these are used as scenario feature vectors. Equipment and topology information are encoded into equipment embedding and topology embedding, time and operating conditions are encoded into time features and operating condition features, and risk assessment results are encoded into risk label features. Through feature concatenation and linear mapping or embedding network transformation, the scenario feature vector is mapped to a unified high-dimensional vector space to obtain the query vector q for subsequent multi-channel retrieval.

[0065] S43: Based on the query vector, perform similar scene retrieval in the historical event database and related entity and relationship retrieval in the power distribution network knowledge graph. Input the retrieval results into the preset retrieval enhancement generation model to obtain knowledge reasoning results.

[0066] Specifically, using the query vector q as input, multi-channel retrieval is performed within a pre-constructed two-layer knowledge support system. The bottom layer stores structured information such as historical power outage events, equipment ledgers, and operation records in vectorized form. Candidate knowledge set K is obtained from the historical event database through vector similarity calculation. c1 The upper layer comprises a semantic knowledge graph consisting of expert experience, typical cases, and operation and maintenance procedures. Candidate knowledge set K is obtained through keyword-based lexical matching, graph path similarity analysis, and time-related retrieval. c2 , will K c1With K c2 Merged into a candidate knowledge set K c For each piece of knowledge k in the set, calculate the comprehensive relevance score r. k The scoring can comprehensively consider indicators such as vector similarity, lexical matching degree, graph path weight, and time decay factor, and can be achieved by learning ranking or weighted ranking from K... c The top-scoring knowledge items are selected to form the optimal knowledge subset K. The high-confidence risk event F corresponding to the unstable scenario is concatenated with the text or structured entries in the knowledge subset K to form a contextual input sequence. This sequence is then input into a pre-defined retrieval-enhanced generative model, which generates knowledge reasoning results that include risk cause analysis, failure evolution trends, and disposal strategy suggestions.

[0067] S44: Analyze the results of knowledge reasoning to obtain knowledge evidence, and merge the basic probability distribution of the knowledge evidence with the basic probability distribution of each evidence body in the multi-source evidence set to obtain an enhanced evidence set.

[0068] Specifically, the knowledge reasoning results are analyzed according to knowledge items. For each knowledge item k in the knowledge subset K, its support S for the risk event proposition F is extracted. k (F) Support can be quantified based on the confidence scores or keyword coverage of the descriptions of fault causes and risk levels in the reasoning results, and knowledge evidence weights r can be assigned to different knowledge items. k The weights can be set by comprehensively considering the reliability of the knowledge source, the similarity to the current scenario, and the historical application effect. Let the normalization coefficient λ be used to ensure the sum of probabilities. Then, the basic probability distribution of RAG-based knowledge evidence on the risk event proposition F can be expressed as: Let m RAG (θ)=1-m RAG (F), the obtained m RAG As a basic probability allocation representing knowledge evidence, it is merged with the basic probability allocation of each evidence body in the original multi-source evidence set under the same risk identification framework. Knowledge evidence is regarded as a new evidence body and added to the multi-source evidence set, thereby obtaining an enhanced evidence set that includes initial evidence and knowledge-enhancing evidence.

[0069] In one embodiment, the risk warning method for a power distribution network further includes: S60: Record the risk level corresponding to the final risk warning result, obtain the actual fault occurrence situation corresponding to the risk level during operation, and obtain operation feedback data by comparing the difference between the final risk warning result and the actual fault occurrence situation.

[0070] Specifically, for each final risk warning result, the identifier of the corresponding risk assessment object, the warning issuance time, and the warning duration window are recorded. Based on the risk level and final confidence level, this information is stored in the warning log in chronological order. During subsequent operation, operational data and fault event records related to the risk assessment object are continuously collected to determine whether a fault event matching the warning level occurs within the warning time window and the subsequent preset observation period, and the severity of the fault. Each warning sample is assigned an actual result label. When a fault occurs during the observation period that matches or is more severe than the warning level, it is recorded as... =1, recorded as when no related fault occurs or only an anomaly significantly less severe than the warning level occurs. =0, and the final confidence level corresponding to this sample is denoted as . By analyzing nearly N historical early warning samples, a pattern like this can be formed. The runtime feedback dataset, where h j This indicates the risk assessment object or risk assumption corresponding to the j-th early warning sample, which is used for subsequent weight adaptive optimization and risk assessment parameter correction.

[0071] S70: Based on operational feedback data, construct evaluation indicators, correct the weight parameters and judgment thresholds in the multi-level risk assessment matrix according to the evaluation indicators, and adjust the weight parameters of each piece of evidence according to the contribution of different pieces of evidence in the evaluation indicators, so as to obtain the updated multi-level risk assessment matrix and evidence weights.

[0072] Specifically, based on the runtime feedback dataset Construct an evaluation index based on the accuracy of early warning hits, and calculate the final confidence score for each sample j. Labels with actual results Substituting the log-likelihood loss function The weight vector W = (w_i) of rule-based evidence, case evidence, and real-time evidence. r ,w c ,w s Using W as the parameter to be optimized, the gradient of the loss function L with respect to W is calculated, and gradient descent or similar numerical optimization methods are used to iteratively update W, so that the actual hit samples are... Increased false positive samples The weighting of unreliable evidence sources is automatically reduced while the weighting of reliable evidence sources is increased. Based on this, the weight parameters and judgment thresholds with high correlation to prediction deviation in the multi-level risk assessment matrix are fine-tuned according to the error contribution of different risk assessment objects on each risk factor. This makes the risk level distribution output by the matrix gradually closer to the actual fault statistical distribution. At the same time, the weight coefficients and knowledge evidence adjustment coefficients of various knowledge channels in RAG retrieval are corrected by combining the deviation between the enhanced confidence interval [Bel'(F),Pl'(F)] and the true risk level. This achieves the coordinated adaptive optimization of the multi-level risk assessment matrix parameters, evidence body weights, and knowledge retrieval weights, resulting in an updated multi-level risk assessment matrix and evidence body weights for subsequent iterative risk warning.

[0073] Example 2 like Figure 2 As shown, based on the same inventive concept as the above embodiments, the present invention also provides a risk early warning system for power distribution networks, comprising: The historical modeling module is used to acquire historical operating data of the distribution network and construct a multi-level risk assessment matrix based on the historical operating data; The evidence modeling module is used to acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-source operational data based on a multi-level risk assessment matrix, and generate a multi-source evidence set. The initial assessment fusion module is used to perform evidence fusion operations on multi-source evidence sets to obtain the first risk assessment result; The knowledge enhancement module is used to perform knowledge retrieval reasoning based on the first risk assessment result and multi-source operational data when the first risk assessment result does not meet the preset stability conditions, to obtain the knowledge reasoning result, and then merge the knowledge reasoning result with the multi-source evidence set to obtain the enhanced evidence set. The final assessment fusion module is used to perform secondary evidence fusion operations on the enhanced evidence set to obtain the final risk warning result.

[0074] Optional, the historical modeling module includes: The hierarchical modeling submodule is used to hierarchically classify equipment, lines and areas in the distribution network based on distribution network ledger information and historical operation data of the distribution network, so as to determine the risk assessment objects of each layer. The factor extraction submodule is used to obtain the characteristic data corresponding to the risk assessment object from the historical operation data of the distribution network, use the characteristic data as the risk factors of the risk assessment object, and then establish a correspondence between the risk factors and the corresponding risk level propositions to obtain the matrix elements of the risk assessment object. The matrix calibration submodule is used to arrange matrix elements in rows and columns according to the risk assessment object, set weight parameters and judgment thresholds for each matrix element, and perform regression analysis to calibrate the weight parameters and judgment thresholds to obtain a multi-level risk assessment matrix.

[0075] Optionally, the evidence modeling module includes: The evidence segmentation submodule is used to segment multi-source operational data according to the risk assessment object to obtain several evidence bodies; The indicator matching submodule is used to match the operational indicators in the evidence body with the corresponding risk factors based on the multi-level risk assessment matrix, so as to obtain the membership degree of the evidence body in the risk level proposition. The BPA generation submodule is used to normalize the membership degree within a unified risk level proposition space, obtain the basic probability allocation of each piece of evidence on each risk level proposition, and then generate a multi-source evidence set.

[0076] Optionally, the indicator matching submodule includes: The life modeling unit is used to fit the failure occurrence time characteristics based on the operating indicators that characterize the equipment life and aging degree in the evidence body, and obtain the probability of occurrence of each risk level proposition in different time intervals by using the Weibull distribution model. The probability of occurrence is then converted into the membership degree of the corresponding risk level proposition. The frequency modeling unit is used to fit the number of events arriving per unit time to the operational indicators representing the frequency of fault events in the evidence body, and obtain the probability of occurrence of each risk level proposition under different event frequencies by using the Poisson distribution model. The probability of occurrence is then converted into the membership degree of the corresponding risk level proposition. The state fuzzy unit is used to map the values ​​of continuous operation indicators that represent the current operation status in the evidence body to the membership degree of each risk level proposition through a preset fuzzy membership function.

[0077] Optional, the initial evaluation integration module includes: The fusion calculation submodule is used to fuse the basic probability assignments to obtain the fusion probability and the corresponding conflict coefficient. The confidence interval submodule is used to calculate the confidence and credibility of the fusion probability to obtain the confidence and credibility of the target risk level proposition, and to construct a confidence interval based on the confidence and credibility. The initial assessment generation submodule is used to generate the first risk assessment result based on the conflict coefficient and confidence interval.

[0078] Optional knowledge enhancement modules include: The stability determination submodule is used to perform validity checks based on the confidence interval and the conflict coefficient. When the interval width is greater than the preset stable interval threshold or the conflict coefficient is greater than the preset conflict threshold, the current evaluation scenario is determined to be an unstable scenario. The scenario construction submodule is used to extract scenario feature information from multi-source operational data and the first risk assessment results based on unstable scenarios, and to construct query vectors based on the scenario feature information. The knowledge retrieval submodule is used to retrieve similar scenarios in the historical event database based on query vectors, and to retrieve relevant entities and relationships in the power distribution network knowledge graph. The retrieval results are then input into a preset retrieval enhancement generation model to obtain knowledge reasoning results. The knowledge evidence submodule is used to parse the knowledge reasoning results to obtain knowledge evidence. It then merges the basic probability distribution of the knowledge evidence with the basic probability distribution of each piece of evidence in the multi-source evidence set to obtain an enhanced evidence set.

[0079] Optionally, the risk warning system for the power distribution network may also include: The feedback acquisition module is used to record the risk level corresponding to the final risk warning result, obtain the actual fault occurrence corresponding to the risk level during operation, and obtain operation feedback data by comparing the difference between the final risk warning result and the actual fault occurrence. The adaptive optimization module is used to construct evaluation indicators based on operational feedback data, correct the weight parameters and judgment thresholds in the multi-level risk assessment matrix according to the evaluation indicators, and adjust the weight parameters of each piece of evidence according to the contribution of different pieces of evidence in the evaluation indicators, so as to obtain the updated multi-level risk assessment matrix and evidence weights.

[0080] Example 3 like Figure 3 As shown, the present invention also provides an electronic device 100 for implementing a risk warning method for a power distribution network; The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0081] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the risk warning method for power distribution network of Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0082] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0083] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0084] The memory 101 in the electronic device 100 stores multiple instructions to implement a risk warning method for a power distribution network, and the processor 102 can execute multiple instructions to achieve the following: Obtain historical operation data of the power distribution network and construct a multi-level risk assessment matrix based on the historical operation data; Acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-source operational data based on a multi-level risk assessment matrix, and generate a multi-source evidence set; Evidence fusion calculations are performed on the multi-source evidence set to obtain the first risk assessment result; If the first risk assessment result does not meet the preset stability conditions, knowledge retrieval reasoning is performed based on the first risk assessment result and multi-source operational data to obtain the knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain the enhanced evidence set. A second evidence fusion operation is performed on the enhanced evidence set to obtain the final risk warning result.

[0085] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0086] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0088] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0089] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A risk early warning method for power distribution networks, characterized in that, The method includes: Obtain historical operating data of the power distribution network, and construct a multi-level risk assessment matrix based on the historical operating data; Acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-level risk assessment matrix based on the multi-source operational data, and generate a multi-source evidence set; Evidence fusion is performed on the multi-source evidence set to obtain the first risk assessment result; If the first risk assessment result does not meet the preset stability conditions, knowledge retrieval reasoning is performed based on the first risk assessment result and the multi-source operation data to obtain the knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain the enhanced evidence set. A second evidence fusion operation is performed on the enhanced evidence set to obtain the final risk warning result.

2. The risk early warning method for power distribution networks according to claim 1, characterized in that, The construction of a multi-level risk assessment matrix based on the historical operational data includes: Based on the distribution network ledger information and the historical operation data of the distribution network, the equipment, lines and areas in the distribution network are layered to determine the risk assessment objects of each layer; The characteristic data corresponding to the risk assessment object is obtained from the historical operation data of the distribution network. The characteristic data is used as the risk factor of the risk assessment object. Then, the risk factor is associated with the corresponding risk level proposition to obtain the matrix elements of the risk assessment object. The matrix elements are arranged in rows and columns according to the risk assessment object, and weight parameters and judgment thresholds are set for each matrix element. Regression analysis is performed on the weight parameters and judgment thresholds to calibrate them, thereby obtaining the multi-level risk assessment matrix.

3. The risk early warning method for power distribution networks according to claim 2, characterized in that, The uncertainty modeling of the multi-source operational data based on the multi-level risk assessment matrix generates a multi-source evidence set, including: The multi-source operational data is divided according to the risk assessment object to obtain several evidence bodies; Based on the multi-level risk assessment matrix, the operational indicators in the evidence body are matched with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition. The membership degree is normalized and mapped within a unified risk level proposition space to obtain the basic probability allocation of each piece of evidence on each risk level proposition, thereby generating the multi-source evidence set.

4. The risk early warning method for power distribution networks according to claim 3, characterized in that, Based on the multi-level risk assessment matrix, the operational indicators in the evidence body are matched with the corresponding risk factors to obtain the membership degree of the evidence body in the risk level proposition, including: Based on the operational indicators characterizing equipment lifespan and aging degree in the evidence body, the failure occurrence time characteristics are fitted by the Weibull distribution model to obtain the probability of occurrence of each risk level proposition in different time intervals, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition. Based on the operational indicators representing the frequency of fault events in the evidence body, the number of events arriving per unit time is fitted using a Poisson distribution model to obtain the probability of occurrence of each risk level proposition under different event frequencies, and the probability of occurrence is converted into the membership degree of the corresponding risk level proposition. Based on the continuous operational indicators representing the current operational status in the evidence body, the values ​​of the operational indicators are mapped to the membership degrees of each risk level proposition through a preset fuzzy membership function.

5. The risk early warning method for power distribution networks according to claim 3, characterized in that, The step of performing evidence fusion operation on the multi-source evidence set to obtain the first risk assessment result includes: The basic probability assignments are fused to obtain the fusion probability and the corresponding conflict coefficient; The fusion probability is subjected to trust and credibility calculations to obtain the trust and credibility of the target risk level proposition, and a confidence interval is constructed based on the trust and credibility. The first risk assessment result is generated based on the conflict coefficient and the confidence interval.

6. The risk early warning method for power distribution networks according to claim 5, characterized in that, When the first risk assessment result does not meet the preset stability conditions, knowledge retrieval and reasoning are performed based on the first risk assessment result and the multi-source operational data to obtain a knowledge reasoning result. Then, the knowledge reasoning result and the multi-source evidence set are merged to obtain an enhanced evidence set, including: The validity is checked based on the confidence interval and the conflict coefficient. When the interval width is greater than the preset stable interval threshold or the conflict coefficient is greater than the preset conflict threshold, the current evaluation scenario is determined to be an unstable scenario. Based on the unstable scenario, scenario feature information is extracted from the multi-source operational data and the first risk assessment result, and a query vector is constructed based on the scenario feature information. Based on the query vector, similar scenarios are retrieved in the historical event database, and related entities and relationships are retrieved in the power distribution network knowledge graph. The retrieval results are then input into a preset retrieval enhancement generation model to obtain knowledge reasoning results. The knowledge reasoning results are analyzed to obtain knowledge evidence. The basic probability distribution of the knowledge evidence is then combined with the basic probability distribution of each piece of evidence in the multi-source evidence set to obtain an enhanced evidence set.

7. The risk early warning method for power distribution networks according to claim 1, characterized in that, The method further includes: Record the risk level corresponding to the final risk warning result, obtain the actual fault occurrence situation corresponding to the risk level during operation, and obtain operation feedback data by comparing the difference between the final risk warning result and the actual fault occurrence situation; Based on the operational feedback data, an evaluation index is constructed. The weight parameters and judgment thresholds in the multi-level risk assessment matrix are corrected according to the evaluation index. The weight parameters of each piece of evidence are adjusted according to the contribution of different pieces of evidence to the evaluation index, resulting in an updated multi-level risk assessment matrix and evidence weights.

8. A risk early warning system for power distribution networks, characterized in that, The system includes: The historical modeling module is used to acquire historical operating data of the distribution network and construct a multi-level risk assessment matrix based on the historical operating data. The evidence modeling module is used to acquire multi-source operational data at the current moment, perform uncertainty modeling on the multi-source operational data based on the multi-level risk assessment matrix, and generate a multi-source evidence set. The initial assessment fusion module is used to perform evidence fusion operations on the multi-source evidence set to obtain the first risk assessment result; The knowledge enhancement module is used to perform knowledge retrieval reasoning based on the first risk assessment result and the multi-source operational data when the first risk assessment result does not meet the preset stability conditions, to obtain a knowledge reasoning result, and then merge the knowledge reasoning result and the multi-source evidence set to obtain an enhanced evidence set. The final evaluation fusion module is used to perform secondary evidence fusion operations on the enhanced evidence set to obtain the final risk warning result.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the steps of the risk warning method for a power distribution network as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the steps of the risk warning method for a power distribution network as described in any one of claims 1 to 7.

Images