Risk prevention method based on source-load balance

Abstract

The invention discloses a risk prevention method based on source-load balance, and relates to the technical field of power systems. In the prior art, in the risk prevention of source-load balance, when a plurality of faults emerge are not considered, the chain influence of the whole system possibly occurs. The method comprises the following steps: collecting multi-source data of a power system, and constructing a fault assessment model, the construction mode being: preprocessing the multi-source data, and constructing a fault database; constructing a fault tree and an event tree based on the fault database, and calculating the probability of occurrence of top events in the fault tree; calculating occurrence probabilities and influence degrees of all influence consequences in the event tree; and fusing the fault tree and the event tree to form a fault evaluation model, and obtaining a fault combination and an influence degree and influence loss thereof. According to the method, mutual influence and chain reaction between faults are analyzed in a mode of combining fault tree analysis and event tree analysis; potential influences of complex faults are identified and evaluated, and system crash caused by chain faults can be prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a risk prevention method based on source-load balance. Background Art

[0002] Source-load balancing refers to the coordination between power sources (generation facilities) and loads (power users) to ensure the stable, secure, and economical operation of the power system. In the context of the current global energy transition, the power system faces increasingly complex challenges. With the rapid development of renewable energy and the continuous growth of electricity demand, the importance of source-load balancing has become increasingly prominent.

[0003] A search revealed Chinese patent number CN202310205364.2, which discloses a method for assessing source-load imbalance in a regional power grid. This method is applied to a multi-region interconnected power grid, where individual regional power grids exchange power via tie lines. The multi-region interconnected power grid includes power generation equipment, transmission line equipment, and power consumption equipment. The power generation equipment includes conventional generators, wind turbines, and photovoltaic power generation equipment. This method, by accounting for source-grid-load uncertainty, assesses the source-load imbalance of individual regions within a multi-region interconnected power grid, providing a basis for grid operation and planning.

[0004] However, existing risk mitigation approaches for source-load balancing lack a comprehensive design of combined failure scenarios, potentially overlooking certain important failure modes and leading to underestimation of risk. In particular, most assessment models typically rely on linear causal relationships between failure events, failing to deeply analyze the interplay and cascading effects that can result from the emergence of multiple failures. For example, a failure on one transmission line can cause a sharp increase in the load on adjacent lines, potentially triggering their own failures. If the overload on other lines is not promptly assessed and adjusted, it could ultimately lead to the collapse of the entire system. Therefore, a risk mitigation approach based on source-load balancing is proposed to address these issues. Summary of the Invention

[0005] Technical problems solved In response to the above-mentioned shortcomings of the existing technology, the present invention provides a risk prevention method based on source-load balance, which can effectively solve the problem that the existing technology fails to consider the chain effects that may occur in the overall system when multiple faults emerge in the risk prevention of source-load balance.

[0006] Technical Solution To achieve the above objectives, the present invention is implemented through the following technical solutions: The present invention provides a risk prevention method based on source-load balance. The technical solution adopted by the present invention is as follows: multi-source data of the power system is collected and a fault assessment model is constructed. The construction method is as follows: Preprocess multi-source data and construct a fault database; construct a fault tree and event tree based on the fault database, and calculate the probability of occurrence of the top event in the fault tree; calculate the probability of occurrence and impact of all impact consequences in the event tree; fuse the fault tree and event tree to form a fault assessment model to obtain the fault combination, its impact level and impact loss; Calculate the risk score of the fault combination, sort the risk scores of different fault combinations in descending order to form a priority list; compare all the risk scores with the pre-set risk threshold range to determine the processing level of the fault combination; construct the processed fault combinations into a maintenance data set, and calculate the comprehensive processing score of each fault combination in the maintenance data set; update the risk score of the fault combination based on the comprehensive processing score to obtain the priority correction score; sort the priority correction scores in descending order to obtain a priority correction list, and merge it into multi-source data.

[0007] Among them, the multi-source data includes historical fault data, equipment operating parameters, environmental parameters and prevention feedback data; historical fault data includes fault type, occurrence frequency and impact range; prevention feedback data includes average response time, average recovery time and user complaint rate of fault handling.

[0008] The method of constructing the fault tree is as follows: Extract the top event from the fault database and place it at the top level of the fault tree. Add bottom events below the top event and add logic gates between the top and bottom events. Then add secondary bottom events below each bottom event and repeat the above steps until the fault tree reaches the preset number of levels. The calculation formula for the probability of occurrence of the top event is: Where, Top event The probability of occurrence; n is the number of bottom events; is the i-th bottom event probability of occurrence; is the time weight of the i-th bottom event; is the correlation factor of the ith bottom event; is the intensity factor of the ith bottom event; All top events are combined to form a risk accident set.

[0009] The event tree is constructed as follows: Select the top event from the risk accident set as the initial event; extract the impact consequences caused by the initial event from the fault database; place the initial event in the first layer and add the impact consequences of the initial event below; add the secondary events triggered below the impact consequences; repeat the above steps until the event tree reaches the preset number of layers, completing the event tree construction; The calculation formula for the probability of occurrence of impact consequences is: Where, is the total probability of occurrence of all impact consequences in the event tree; is the probability of occurrence of the i-th impact consequence; The calculation formula for the impact degree of the consequences is: Where, The degree of influence of all the consequences on the initial event; is the influencing factor of the i-th impact consequence; is the adjustment factor; represents the dependency coefficient between the i-th impact consequence and the j-th impact consequence; is the probability of occurrence of the jth impact consequence; m is the total number of impact consequences.

[0010] The calculation formula for the impact degree of the fault combination is: ; Where EFF represents the impact of all fault combinations in the power system; F is the total number of faults; k is the number of faults in the fault combination; Indicates the total number of consequences of the fault number k in the current fault combination; is the impact factor of the jth impact consequence; j is the index of the impact consequence; is the kth fault in the fault combination; The calculation formula for the impact loss of the fault combination is: ; Where LOSS is the quantitative value of the loss suffered by the power system under all fault combinations; is the i-th fault combination probability of occurrence; is the quantified value of the loss caused by the i-th fault combination, N is the total number of simulations, is the loss quantification value obtained from the a-th simulation.

[0011] The calculation formula of the correlation factor is: ; Where, Indicates the jth bottom event probability of occurrence; Represents the correlation coefficient between the i-th bottom event and the j-th bottom event; the calculation formula is: Where, represents the covariance between the i-th bottom event and the j-th bottom event; and represent the standard deviation of the i-th bottom event and the j-th bottom event respectively.

[0012] The dynamic update method of the correlation coefficient is: The correlation matrix R is constructed based on the correlation coefficient between the bottom events, and the attenuation factor is preset. Calculate the timeliness weight using the following formula: Where, is the timeliness weight of the ith bottom event; t is the current time, is the time when the i-th bottom event occurs; is the sensor data impact factor of the i-th bottom event; The correlation matrix is ​​updated using the timeliness weight. The update formula is: Where, is the updated correlation coefficient between the i-th bottom event and the j-th bottom event; is the timeliness weight of the i-th bottom event; is the timeliness weight of the jth bottom event.

[0013] The sensor data influencing factor is obtained as follows: The anomaly monitoring model is constructed using multi-source data in the following way: Merge different features in the equipment operating parameters according to timestamps to form data points with timestamps, and merge all data points within a fixed historical time period into a feature data set; preprocess the feature data set; Based on the preset K value, determine the number of nearest neighbor points for each data point in the feature data set and generate the K nearest neighbor set for each data point , P is the data point currently being processed; Calculate the reachable density of each data point using the following formula: Where, is the reachable density of data point P; is the Euclidean distance between the data point P and the neighbor point Q; Then, based on the reachable density of data point P, the degree of abnormality is calculated. The calculation formula is: Where, is the local outlier factor of data point P; is the reachability density of neighbor point Q; Represents the number of neighbor points in the K nearest neighbor set; Based on the local outlier factor value of the data point P, the abnormality is judged. The data point P is marked as an outlier and its corresponding anomaly score is calculated using the following formula: Where, is the anomaly score of data point P; is the median of the local outlier factors of all data points in the feature dataset; is the standard deviation of the local outlier factors of all data points in the feature dataset; Mapping the anomaly score to the sensor data impact factor is calculated as follows: .

[0014] The calculation formula for the risk score of the fault combination is: Where, Score the risk of a combination of failures; is the weight coefficient of the impact degree of the fault combination; is the probability of occurrence of fault combination C The weight coefficient of The weight coefficient for quantifying the loss of the fault combination; The recovery time of the fault combination The weight coefficient of .

[0015] The comprehensive processing score of the fault combination is calculated as follows: Where, Score the comprehensive handling of the fault combination; is the average response time of the fault combination; is the average processing time of the fault combination; The rate of user complaints handled for the fault combination; Assign weights to average response time; Assign weights to the average processing time; Assign weights to complaint rates.

[0016] The risk score update formula is: Where, Correction points for the priority of the fault combination; is the value of the i-th evaluation indicator in the risk score; Control coefficient for risk scoring; is the value of the i-th evaluation index in the comprehensive processing score.

[0017] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. In the present invention, by collecting historical fault data, equipment operating parameters, environmental parameters and prevention feedback data of the power system, a comprehensive data foundation is established, which can realize dynamic assessment of the risk level of the power system, timely identify potential risks, and help formulate targeted response measures, thereby reducing the risk of power supply interruption.

[0018] 2. In the present invention, a combination of fault tree analysis and event tree analysis is adopted to deeply analyze the mutual influence and chain reaction between faults; the potential impact of complex faults is more comprehensively identified and evaluated, and the accuracy and effectiveness of fault management are improved. In particular, in the case of multiple faults, system crashes caused by chain faults can be prevented.

[0019] 3. In the present invention, based on the impact degree, occurrence probability and loss quantification value of the fault combination, the risk score of the fault combination is calculated and sorted to achieve priority management of high-risk faults; the risk response speed is accelerated, ensuring that resources can be quickly deployed when a fault occurs, reducing the impact on users, and enhancing the emergency management capabilities of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The figure is a schematic diagram of the risk prevention method process of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0022] Example Reference Figure 1 , this case proposes a risk prevention method based on source-load balance, which includes the following steps: Step 1: Collect multi-source data from the power system, including historical fault data, equipment operating parameters, environmental parameters, and preventive feedback data. Historical fault data includes fault type, frequency, and impact range. This historical fault data can be used to identify the most common fault modes in the power system, providing a critical data foundation for subsequent maintenance and preventive measures. For example, high-frequency faults can be prioritized for maintenance to ensure stable system operation. Collecting equipment operating parameters and environmental factors (such as temperature and humidity) can better understand the operating characteristics of power system hardware under various environmental conditions, providing effective data for subsequent power system compliance forecasting and enabling more efficient load distribution. Preventive feedback data includes average response time, average recovery time, and customer complaint rate for fault handling. This preventive feedback data is incorporated into the prioritization of fault combinations, enabling more effective, accurate, and practical sorting to ensure the continued safe operation and rapid response capabilities of the power system.

[0023] Multi-source data can be used to dynamically assess the current risk level of the power system, helping to identify potential risks in a timely manner, take appropriate countermeasures, and reduce the risk of power supply interruptions.

[0024] Step 2: Use multi-source data to build a fault assessment model to obtain the fault combination, its impact degree and impact loss. The method of building a fault assessment model is: S201: Preprocess the multi-source data, including data cleaning and standardization, and construct the processed multi-source data into a fault database.

[0025] S202: Use FTA to process the fault database, analyze the single faults in the historical fault data and the fault combinations caused by them, and construct a fault tree to clarify the root cause of the fault.

[0026] S203: Use ETA to process the risk accident set, visualize the possible consequences of each top event, and build an event tree to evaluate the chain reactions that may be triggered by the top event, thereby gaining a deep understanding of the potential risks of different top events and providing data support for the formulation of corresponding response strategies.

[0027] S204: Fault trees and event trees are integrated to form a fault assessment model. The fault tree provides a basic framework for identifying various faults within the power system, clarifying how each fault leads to a top event. The ETA provides the fault assessment model with the various consequences of these faults, reflecting the interactions between different faults. By inputting multi-source data into the fault assessment model, all possible fault combinations in the power system can be determined, along with their corresponding impact levels and losses.

[0028] The calculation formula for the impact degree and impact loss of the fault combination is: ; Where EFF represents the impact of all fault combinations in the power system, which is a cumulative value that evaluates the comprehensive impact of each fault combination on the power system; n is the total number of faults; k is the number of faults in the fault combination; Indicates the total number of consequences of the fault number k in the current fault combination; is the influencing factor of the jth influencing consequence; j is the index of the influencing consequence; It represents the probability of a combination of faults occurring simultaneously; is the first fault in the fault combination, The second fault in the fault combination, and so on, until the kth fault .

[0029] ; Where LOSS is the quantitative value of the loss suffered by the power system under all fault combinations (such as total economic loss or other losses); F is the total number of fault combinations; is the i-th fault combination probability of occurrence; is the quantified value of the loss caused by the i-th fault combination, estimated using the Monte Carlo simulation method; where N is the total number of simulations; is the loss quantification value obtained from the a-th simulation.

[0030] The fault assessment model can be used to identify the root causes of different faults in the power system and evaluate their potential impact, effectively supporting emergency response preparation; and the combination of fault trees and event trees can accurately reflect the interactions between faults and provide data support for risk management.

[0031] Step 3: Based on the fault combination, its impact level and impact loss, calculate the risk score of the fault combination. The calculation formula is: Where, Score the risk of a combination of failures; is the weight coefficient of the impact degree of the fault combination; is the probability of occurrence of fault combination C The weight coefficient of The weight coefficient for quantifying the loss of the fault combination; The recovery time of the fault combination Weight coefficient, recovery time Obtained based on statistical analysis of historical fault data from multiple sources. The risk scores of different fault combinations are sorted from large to small to form a priority list. The priority list is then compared with the pre-set risk threshold range, and fault combinations with risk scores greater than the risk threshold range are set as high-priority faults; fault combinations with risk scores within the risk threshold range are set as medium-priority faults; and fault combinations with risk scores less than the risk threshold range are set as low-priority faults. Clarifying responsibility allocation and resource deployment for fault combinations at different handling levels can significantly shorten response time when a fault occurs, ensuring rapid action, effectively improving the emergency response capabilities of the power system, reducing the impact of risks on users and companies, and ensuring efficient and stable operations.

[0032] Step 4: The processed fault combinations are integrated into a maintenance dataset, and the comprehensive processing score of each fault combination in the maintenance dataset is calculated using the following formula: Where, Score the comprehensive handling of the fault combination; is the average response time of the fault combination; is the average processing time of the fault combination; The rate of user complaints handled for the fault combination; Assign weights to average response time; Assign weights to the average processing time; Assign weights to complaint rates.

[0033] Through the comprehensive processing score of fault combinations, a quantitative evaluation standard is provided for the processing of each fault combination to measure the impact and response effect of different fault combinations; thus, a decision support tool based on actual data is provided for the power system, enabling it to make quick and wise choices in a dynamically changing environment.

[0034] Step 5: Update the risk score based on the comprehensive processing score of the fault combination to obtain the priority correction score. The update formula is: Where, Correction points for the priority of the fault combination; is the value of the i-th evaluation indicator in the risk score; It represents the maximum value of all evaluation indicators in the risk score, and represents the best value of the current power system in different evaluation indicators in risk assessment (e.g. It can be the impact of the fault combination, the loss quantification value, the recovery time or the probability of occurrence), emphasizing the importance of high performance, and is suitable for situations where good performance needs to be given priority; is the control coefficient of the risk score, with a value range of [0, 1], which is used to control the relative influence of the maximum and minimum values ​​in the risk score; is the value of the i-th evaluation index in the comprehensive processing score; The lowest value of different evaluation indicators (e.g. This could be average response time, average handling time, or complaint rate), used to highlight bottlenecks or shortcomings in power system fault handling, especially under uncertain conditions, to ensure a minimum level of control; To control the supplementary part of the coefficient, ensure that the weighted sum of the priority correction points is 1.

[0035] The priority correction points of all fault combinations are sorted from large to small and merged to form a priority correction list; the priority correction list is integrated into multi-source data to provide basic data for subsequent risk assessment and model updates, thereby improving existing emergency plans and operating procedures and forming the latest risk prevention plan, covering risk identification, response, monitoring and continuous improvement.

[0036] In step 2, the fault tree is constructed as follows: The top event in the historical fault data (i.e., the fault event that the fault assessment model needs to predict and prevent) is extracted as the starting point of the fault tree, representing the main fault or failure phenomenon.

[0037] Draw a fault tree structure and break down the various bottom events that lead to the top event. Specifically, place the top event at the top level of the fault tree. Then, build the tree layer by layer, adding bottom events that trigger the top event (the cause of the top event, i.e., the fault; there is a causal relationship between the top and bottom events. For example, if the top event is a transmission line fault, bottom events include conductor breakage, overload, equipment failure, and severe weather). Add logic gates between the top event and the bottom events below it to represent the relationship. Logic gates include AND and OR gates. An AND gate requires all input bottom events to occur for a top event to occur; an OR gate requires only any one input bottom event to occur for a top event to occur. Further analyze each bottom event to identify the secondary bottom events that cause it. Repeat this process (adding logic gates between each bottom event and its secondary bottom event) until the fault tree reaches the preset number of levels.

[0038] Calculate the probability of occurrence of the top event in the fault tree to quantitatively assess the risk level caused by the top event. The calculation formula for the probability of occurrence of the top event is: Where, Top event The probability of occurrence is the failure event that the fault assessment model needs to predict and prevent; n is the number of bottom events; Represents the product symbol, which is used to calculate the combined probability of multiple bottom events not occurring; is the i-th bottom event The probability of occurrence is the frequency of occurrence of the i-th bottom event in a fixed time period in the fault database; is the time weight of the i-th bottom event; is the correlation factor of the i-th bottom event, which reflects the correlation effect between bottom events. For example, whether the occurrence of the i-th bottom event affects the probability of occurrence of other bottom events. By introducing the correlation factor, the probability can be adjusted. When there are multiple correlated events, the result is more accurate. is the intensity factor of the ith bottom event, indicating the degree of impact of the ith bottom event on the top event. It enables a more accurate assessment of the risk caused by the top event when a bottom event occurs, and is set based on historical data.

[0039] All possible top events are combined to form a risk accident set.

[0040] The event tree is constructed as follows: Select any one or more top events from the risk accident set as the initial event; extract all possible impact consequences of the initial event from the fault database (for example, when the initial event is a transmission line failure, mark it as "The transmission line failure will cause a large-scale power outage in the region."). Place the initial event in the first layer and build an event tree from top to bottom, adding the impact consequences of the initial event. For each impact consequence, continue to build branches downward, adding secondary events that may be triggered by the impact consequence. Then, add the corresponding impact consequence to each secondary event, and repeat the above steps until the event tree reaches the preset number of layers, completing the event tree construction.

[0041] Calculate the probability of occurrence of each impact consequence and its impact extent using the following formula: Where, is the total probability of occurrence of all impact consequences in the event tree, reflecting the risk level of all possible consequences of the initial event and used to assess the overall risk of the initial event; m is the total number of impact consequences; is the probability of occurrence of the ith impact consequence, obtained based on statistical analysis of experimental data, and its value range is [0, 1]; When Indicates that the consequences of the impact are unlikely to occur; is the impact coefficient of the ith impact consequence, reflecting the relative importance of the ith impact consequence to the overall risk of the initial event.

[0042] Where, The degree of impact of all impact consequences on the initial event, used to assess the potential losses and risks of the initial event; is the impact factor of the ith impact consequence, which indicates the possible impact of the ith impact consequence on the initial event, which can be the economic loss or power outage time caused; is the adjustment coefficient, which is used to control the intensity of the interaction between the impact consequences. Its value range is [0, 1]. The larger the value, the more obvious the impact of the interaction. represents the dependency coefficient between the ith impact consequence and the jth impact consequence, indicating the degree of influence of the jth impact consequence on the ith impact consequence (e.g., if a power outage may cause equipment damage, then the probability of equipment damage will depend in some way on the probability of power outage); is the probability of occurrence of the jth impact consequence.

[0043] The calculation formula for the correlation factor of the i-th bottom event is: Where, Indicates the jth bottom event probability of occurrence; It represents the correlation coefficient between the ith bottom event and the jth bottom event, reflecting the degree of correlation between the ith bottom event and the jth bottom event.

[0044] The formula for calculating the correlation coefficient is: Where, represents the covariance between the i-th bottom event and the j-th bottom event; and represent the standard deviation of the i-th bottom event and the j-th bottom event respectively.

[0045] The operation of the power system is affected by many factors, such as equipment status, environmental conditions, and real-time load. Over time, the practical significance of certain data (such as historical fault records or equipment status) will weaken and cannot accurately reflect the current system's true condition. Therefore, relying solely on static historical data will not be sufficient to make an accurate risk assessment. As a result, when calculating the degree of correlation between different bottom-level events, the timeliness of the data cannot be guaranteed, and thus the accuracy of the impact of the fault combination and the impact loss assessment cannot be guaranteed. By introducing timeliness weights and dynamically updating the correlation coefficients between different bottom-level events, the effectiveness and accuracy of the impact and impact loss assessment can be effectively improved; the dynamic update method is: The correlation matrix R is constructed based on the correlation coefficients between the i-th bottom event and all other bottom events; the expression of the correlation matrix is: . Set a decay factor with a value less than 1 , which is used to control the impact of historical fault data, so that the impact and loss assessment using historical fault data are timely.

[0046] The timeliness weight is calculated based on the attenuation factor. The calculation formula is: Where, is the timeliness weight of the ith bottom event. This value represents the importance of the ith bottom event at the current time t. The higher the value, the greater the impact of the event on the final output. t is the current time, is the time when the i-th bottom event occurs; Indicates the time between the current event t and the i-th bottom event The time difference between the two events indicates the newness of the i-th event. If the time difference is large, it means that the i-th event occurred a long time ago, and the timeliness weight will be significantly reduced. is the sensor data impact factor of the i-th bottom event, indicating the influence of the current device status on the timeliness weight.

[0047] The correlation matrix is ​​updated using the timeliness weight. The update formula is: Where, is the updated correlation coefficient between the i-th bottom event and the j-th bottom event; is the timeliness weight of the i-th bottom event; is the timeliness weight of the jth bottom event. By updating the correlation matrix, the relationship between different bottom events can be more accurately identified, ensuring that the latest bottom event can affect the risk assessment, and then timely identifying potential high-risk factors and making adjustments to reduce the possibility of failure. Based on the latest data and the dynamically updated correlation matrix, resources can be allocated more effectively (for example, when the increased possibility of failure of a certain equipment is identified, maintenance or backup can be prioritized, which can significantly reduce potential losses). This enables the power system to better adapt to the rapidly changing dynamic environment, improves the overall safety and reliability of the power system, reduces failures and downtime caused by emergencies, and thus better serves users.

[0048] However, the timeliness weights calculated based on temporary data cannot reflect the actual status of the equipment in a timely and accurate manner. Due to the lag in understanding the current status of the equipment, this information delay may result in the equipment not being discovered in time when problems occur, thus delaying maintenance and repairs, and leading to incorrect judgments and response strategies. By building an anomaly monitoring model based on the equipment operating parameters in multi-source data, the anomaly score of the equipment in the power system is obtained, which is mapped to the influencing factors of the sensor data. This can reflect the abnormal conditions of the equipment in a more timely manner, ensure the ability to respond quickly to environmental changes, and significantly improve the accuracy of the timeliness weight calculation. The calculation formula is: .

[0049] The way to build an anomaly monitoring model is: The different features of the equipment operating parameters (equipment temperature, voltage, current, etc.) are merged according to the collection timestamp to form data points with timestamps. All data points within a fixed historical time period are merged into a feature data set; the feature data set is preprocessed, including data cleaning and data standardization.

[0050] Based on the preset K value, determine the number of nearest neighbor points of each data point in the feature data set; then use the Euclidean distance to determine the similarity between each data point and all other neighbor points, and generate the K nearest neighbor set of each data point , P is the data point currently being processed.

[0051] Calculate the reachable density of each data point to measure the density of the data point in the K nearest neighbor set. The calculation formula is: Where, is the reachable density of data point P; is the Euclidean distance between the data point P and its neighbor point Q.

[0052] Then, based on the reachable density of data point P, the degree of abnormality is calculated. The calculation formula is: Where, is the local outlier factor of the data point P, which is used to reflect the abnormality of the data point P; is the reachability density of neighbor point Q; Indicates the number of neighbor points in the K nearest neighbor set.

[0053] The abnormality is judged based on the local outlier factor value of the data point P. If , indicating that the neighboring points around the data point P are denser than it, suggesting that it may be an outlier in the local environment, and thus marking the data point P as an abnormal point.

[0054] The corresponding anomaly score is calculated based on the local outlier factor of the outlier point. The calculation formula is: Where, is the anomaly score of data point P; is the median of the local outlier factors of all data points in the feature dataset; is the standard deviation of the local outlier factors of all data points in the feature dataset.

[0055] By combining the median and standard deviation analysis of local outlier factors, the influence of noise can be effectively suppressed, thereby improving the reliability of anomaly scores, enabling sensor data influencing factors to more accurately identify potential anomalies, thereby reducing false positives and negatives and improving the accuracy of data processing.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A risk prevention method based on source-load balance, characterized in that: The following steps are involved: Collect multi-source data of the power system and build a fault assessment model by: Preprocess multi-source data and build a fault database; construct fault trees and event trees based on the fault database, and calculate the probability of occurrence of the top event in the fault tree; Calculate the probability of occurrence and impact of all impact consequences in the event tree; Fault tree and event tree are integrated to form a fault assessment model to obtain the fault combination, its impact degree and impact loss; Calculate the risk score of the fault combination and sort the risk scores of different fault combinations in descending order to form a priority list; compare all risk scores with the pre-set risk threshold range to determine the handling level of the fault combination; The processed fault combinations are constructed into a maintenance data set, and the comprehensive processing score of each fault combination in the maintenance data set is calculated; Update the risk score based on the comprehensive processing score of the fault combination to obtain the priority correction score; Arrange the priority revisions in descending order to obtain a priority revision list, and merge it into the multi-source data.

2. The risk prevention method based on source-load balance according to claim 1, characterized in that: The multi-source data includes historical fault data, equipment operating parameters, environmental parameters and prevention feedback data; historical fault data includes fault type, occurrence frequency and impact range; prevention feedback data includes average response time, average recovery time and user complaint rate of fault handling.

3. The risk prevention method based on source-load balance according to claim 2, characterized in that: The method of constructing the fault tree is: Extract the top event from the fault database and place it at the top of the fault tree; add bottom events below the top event and add logic gates between the top and bottom events; Then add a secondary bottom event under each bottom event and repeat the above operation until the fault tree reaches the preset number of layers; The calculation formula for the probability of occurrence of the top event is: Where, Top event The probability of occurrence; n is the number of bottom events; is the i-th bottom event probability of occurrence; is the time weight of the i-th bottom event; is the correlation factor of the ith bottom event; is the intensity factor of the ith bottom event; All top events are combined to form a risk accident set.

4. The risk prevention method based on source-load balance according to claim 3, characterized in that: The method of constructing the event tree is: Select the top event from the risk accident set as the initial event; Extract the impact consequences caused by the initial event from the fault database; Place the initial event on the first layer and add the consequences of the initial event below. Add the secondary events triggered below the impact consequences; repeat the above steps until the event tree reaches the preset number of layers, completing the construction of the event tree; The calculation formula for the probability of occurrence of impact consequences is: Where, is the total probability of occurrence of all impact consequences in the event tree; is the probability of occurrence of the i-th impact consequence; The calculation formula for the impact degree of the consequences is: Where, The degree of influence of all the consequences on the initial event; is the influencing factor of the i-th impact consequence; is the adjustment factor; represents the dependency coefficient between the i-th impact consequence and the j-th impact consequence; is the probability of occurrence of the jth impact consequence; m is the total number of impact consequences.

5. The risk prevention method based on source-load balance according to claim 3, characterized in that: The calculation formula for the impact degree of the fault combination is: ； Where EFF represents the impact of all fault combinations in the power system; F is the total number of faults; k is the number of faults in the fault combination; Indicates the total number of consequences of the fault number k in the current fault combination; is the impact factor of the jth impact consequence; j is the index of the impact consequence; is the kth fault in the fault combination; The calculation formula for the impact loss of the fault combination is: ； Where LOSS is the quantitative value of the loss suffered by the power system under all fault combinations; is the i-th fault combination probability of occurrence; is the quantified value of the loss caused by the i-th fault combination, N is the total number of simulations, is the loss quantification value obtained from the a-th simulation.

6. The risk prevention method based on source-load balance according to claim 3, characterized in that: The calculation formula of the correlation factor is: ； Where, Indicates the jth bottom event probability of occurrence; Represents the correlation coefficient between the i-th bottom event and the j-th bottom event; the calculation formula is: Where, represents the covariance between the i-th bottom event and the j-th bottom event; and denote the standard deviation of the i-th bottom event and the j-th bottom event respectively; The dynamic update method of the correlation coefficient is: The correlation matrix R is constructed based on the correlation coefficient between the bottom events, and the attenuation factor is preset. Calculate the timeliness weight using the following formula: Where, is the timeliness weight of the i-th bottom event; t is the current time, is the time when the i-th bottom event occurs; is the sensor data impact factor of the i-th bottom event; The correlation matrix is ​​updated using the timeliness weight. The update formula is: Where, is the updated correlation coefficient between the i-th bottom event and the j-th bottom event; is the timeliness weight of the i-th bottom event; is the timeliness weight of the jth bottom event.

7. The risk prevention method based on source-load balance according to claim 6, characterized in that: The sensor data influencing factor is obtained as follows: The anomaly monitoring model is constructed using multi-source data in the following way: Merge different features in the equipment operating parameters according to timestamps to form data points with timestamps, and merge all data points within a fixed historical time period into a feature data set; Preprocess the feature dataset; Based on the preset K value, determine the number of nearest neighbor points for each data point in the feature data set and generate the K nearest neighbor set for each data point , P is the data point currently being processed; Calculate the reachable density of each data point using the following formula: Where, is the reachable density of data point P; is the Euclidean distance between the data point P and the neighbor point Q; Then, based on the reachable density of data point P, the degree of abnormality is calculated. The calculation formula is: Where, is the local outlier factor of data point P; is the reachability density of neighbor point Q; Represents the number of neighbor points in the K nearest neighbor set; Based on the local outlier factor value of the data point P, the abnormality is judged. The data point P is marked as an outlier and its corresponding anomaly score is calculated using the following formula: Where, is the anomaly score of data point P; is the median of the local outlier factors of all data points in the feature dataset; is the standard deviation of the local outlier factors of all data points in the feature dataset; Mapping the anomaly score to the sensor data impact factor is calculated as follows: 。 8. The risk prevention method based on source-load balance according to claim 5, characterized in that: The calculation formula for the risk score of the fault combination is: Where, Score the risk of a combination of failures; is the weight coefficient of the impact degree of the fault combination; is the probability of occurrence of fault combination C The weight coefficient of The weight coefficient for quantifying the loss of the fault combination; The recovery time of the fault combination The weight coefficient of .

9. The risk prevention method based on source-load balance according to claim 8, characterized in that: The comprehensive processing score of the fault combination is calculated as follows: Where, Score the comprehensive handling of the fault combination; is the average response time of the fault combination; is the average processing time of the fault combination; The rate of user complaints handled for the fault combination; Assign weights to average response time; Assign weights to the average processing time; Assign weights to complaint rates.

10. The risk prevention method based on source-load balance according to claim 9, characterized in that: The update formula of the risk score is: Where, Correction points for the priority of the fault combination; is the value of the i-th evaluation indicator in the risk score; Control coefficient for risk scoring; is the value of the i-th evaluation index in the comprehensive processing score.

Images