Bayesian neural network based substation seismic resilience rapid assessment method

Abstract

The application discloses a kind of based on bayesian neural network's substation seismic toughness fast evaluation method, the network model of substation system is built by directed acyclic graph, in the case where the reliability of each type of equipment is determined, the failure probability of system is quickly evaluated based on each path passing probability, the repair sequence of equipment after earthquake is analyzed by setting function index, function recovery function is determined and seismic toughness index is obtained, to realize the fast evaluation of substation system seismic toughness.The application has the advantages that complex network model is disassembled into the probability of each subnode and logical connection relationship, the data input and processing of network model are greatly reduced, the target probability and seismic level of each subnode are quickly output;The analysis and calculation of each subnode probability set in network model, the probability level of measured node is accurately and efficiently obtained, and the problem of poor convergence of sampling result is avoided.

Description

Technical Field

[0001] This invention relates to the field of evaluating the functional status, seismic resistance, and seismic toughness of substation systems, and particularly to a rapid evaluation method for the seismic toughness of substations based on Bayesian neural networks. Background Technology

[0002] Substations are power system networks composed of various electrical equipment connected by busbars. They possess high redundancy and interconnectivity, ensuring continued operation even after an earthquake. However, substations have suffered extensive damage in numerous major earthquakes both domestically and internationally. The resulting power outages caused by their malfunctions have led to significant direct and indirect economic losses. In the 1994 Beiling earthquake, widespread substation damage caused power outages, directly impacting the power supply to 1.1 million people and resulting in severe economic losses. In the 2008 Wenchuan earthquake, the Sichuan power grid was severely damaged, with over 170 substations of 35kV and above damaged, and it took several months for them to return to normal operation. In the 2011 Great East Japan Earthquake, 67 substations were severely damaged, and it took 10 days for 95% of the power system to return to normal. This demonstrates that for sudden and destructive natural disasters like earthquakes, it is extremely difficult to accurately predict and effectively defend against them before they occur. Extensive research has been conducted on the seismic performance of substation systems and their internal equipment. While these studies can reflect the seismic resistance of various equipment from a probabilistic perspective and demonstrate the seismic resistance level of electrical equipment and even the system level, the occurrence of earthquakes and other disasters is not instantaneous. There are many factors involved, such as the duration of the earthquake, the post-earthquake repair time, the allocation of post-earthquake resources, and economic losses. Therefore, a comprehensive analysis of the seismic resistance of substations throughout the entire process from the occurrence of an earthquake to complete repair is a pressing issue that needs to be addressed.

[0003] In 1973, the concept of resilience first appeared in ecology. It refers to the ability of a system to resist and recover to its original normal functional level after being subjected to external disturbances. Since then, with in-depth research, the scope of resilience research has expanded to fields such as medicine, transportation, and urban pipeline networks. In the field of power systems, seismic resilience is theoretically summarized into five key characteristics: reliability, robustness, redundancy, resourcefulness, and speed. Reliability refers to the probability that a system will maintain its function reliably under earthquake loading; it is one of the key characteristics for measuring the seismic resistance of substations. Robustness refers to the system's ability to resist damage and maintain its functional level under earthquake loading. Redundancy refers to the excess capacity of substation system structural connections or equipment functional levels, ensuring that the system can still maintain normal functional status even when a few components are damaged. Resourcefulness refers to the coordination and linkage capabilities of various resources in the system. Speed ​​refers to the substation's ability to quickly recover from functional losses after an earthquake. These five characteristics well explain the important role of seismic resilience in practical engineering.

[0004] Building upon this foundation, research on the quantification of seismic toughness has gradually unfolded. Currently, the main methods for establishing substation functional models are fault tree and event tree techniques, or a combination of graph theory and success path analysis. These methods utilize state trees to construct a greatly simplified system model, calculating the failure probability of the entire system and explicitly considering the correlations between components, thus effectively establishing the structural characteristics of the substation. The sampling methods used in assessing the seismic toughness of substation systems are primarily Monte Carlo sampling and Latin hypercube sampling. These methods mainly obtain initial data through extensive random sampling, then filter and analyze the data according to requirements, thereby achieving the goal of quantifying the functional level and seismic toughness of the substation system. However, these mainstream sampling methods have significant drawbacks. Firstly, the large amount of initial data leads to a significant time expenditure on sampling and data processing. Reducing the number of sampling attempts results in insufficient random sampling data, poor convergence, and inaccurate results. Furthermore, substation systems have high equipment and line redundancy, and power users have varying electricity demands. Therefore, how to more comprehensively meet the functional requirements of substations and achieve more accurate and rapid seismic vulnerability assessment is a pressing issue that needs to be addressed.

[0005] Chinese patent CN1123293376B discloses a quantitative assessment method for the seismic toughness of substation systems based on Monte Carlo simulation. However, this method has significant drawbacks. First, the Monte Carlo method suffers from large data sampling and long processing time. Conversely, small sampling leads to poor convergence, meaning that extensive sampling simulations are required for accurate results, resulting in long processing cycles. Furthermore, seismic toughness assessment focuses on reflecting the system's ability to withstand and recover after an earthquake. However, this quantitative assessment method only represents the post-earthquake repair process with a single curve, ignoring repair efficiency and sequence. The post-earthquake repair method has a significant impact on seismic toughness; high repair efficiency leads to a significant decrease in the seismic toughness index. Therefore, this method cannot obtain an accurate seismic toughness level for substations. These two issues are crucial for research on the seismic toughness of substation systems and are areas this application aims to address. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a rapid assessment method for the seismic toughness of substations based on Bayesian neural networks. This method decomposes the complex network model into the probabilities and logical connections of each sub-node, which greatly reduces the amount of data input and processing of the network model and quickly outputs the target probability and seismic resistance level of each sub-node.

[0007] To address the above technical problems, this invention provides a rapid assessment method for the seismic toughness of substations based on Bayesian neural networks, comprising the following steps:

[0008] S1: Determine the seismic intensity of the area where the substation is located, and clarify the type, number, and system connection method of the substation equipment;

[0009] The seismic intensity of the area where the substation is located is represented by the peak ground acceleration (PGA) and used as an indicator to measure earthquake strength.

[0010] S2: Establish a Bayesian neural network for the substation system, build a directed acyclic graph (DAG) between equipment and busbars in the substation, and obtain the failure probability of the tested sub-nodes based on the reliability analysis of various types of equipment.

[0011] In the directed acyclic graph of the substation, each node represents an electrical device, and each line represents a busbar connecting the devices.

[0012] The reliability of the various types of equipment is expressed in the form of seismic vulnerability curves, which follow a log-normal cumulative distribution with a median of μ and a log-standard deviation of β, as shown in the following formula:

[0013] .

[0014] S3: Assign a criticality coefficient Q to power users of different importance levels, and establish a functional evaluation index K for the substation system based on the failure probability of outgoing line units supplying various types of power users, as shown in the following formula:

[0015] ;

[0016] in: The important user coefficients for the qth outgoing line unit that can operate normally after the earthquake; The probability that i outgoing line units of the substation will operate normally after the earthquake; is the important user coefficient of the wth outgoing unit of the substation; n is the total number of outgoing units of the substation system; the larger the system function index K is, the smaller the impact of earthquake on the substation's function and the better the system function.

[0017] The electricity users are classified into three levels: special grade, first grade, and second grade important electricity users. The coefficient for special grade important users is Q1=2, the coefficient for first grade important users is Q2=1.5, and the coefficient for second grade important users is Q3=1.

[0018] S4: Calculate the post-earthquake functional recovery function Q(t) of the substation based on the substation functional evaluation index to characterize the change of the substation functional recovery level over time;

[0019] The substation repairs are carried out in descending order of equipment repair efficiency, with the equipment repair efficiency index P being... i As shown in the following formula:

[0020] ;

[0021] Where: P i K represents the repair efficiency index for the i-th type of equipment. n The post-earthquake functional level of the substation; K bi The functional level of the substation after the repair of equipment of type i; T i Let be the repair time for the i-th type of device.

[0022] S5: Based on the substation functional recovery function Q(t), a quantitative standard for the seismic toughness index R of the substation is proposed, using the area enclosed by function and time as the standard. This index reflects the substation's ability to resist damage and recover quickly under seismic loads, as shown in the following formula:

[0023] ;

[0024] in: The time of the earthquake damage. The time required for the substation system to fully recover its functions;

[0025] The magnitude of the seismic toughness index R reflects the seismic toughness level of the substation system. The larger the R, the lower the seismic toughness level of the substation, resulting in greater resource consumption and economic losses.

[0026] The superior effects of this invention are as follows:

[0027] 1) This invention decomposes the complex network model into the probability and logical connection relationship of each sub-node, which greatly reduces the amount of data input and processing of the network model, thereby quickly outputting the target probability and seismic resistance level of each sub-node;

[0028] 2) This invention obtains the probability level of the tested node accurately and effectively by analyzing and calculating the probability set of each child node in the network model, thus avoiding the problem of poor convergence of sampling results.

[0029] 3) This invention applies conditional constraints to any node in the network model, while satisfying the multi-state probability design of each node. It is applicable to substations of different regions and types, and satisfies the multi-state output of substations under different conditional constraints, effectively meeting the actual needs in engineering. Attached Figure Description

[0030] The accompanying drawings, which form part of this application, 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:

[0031] Figure 1 This is a flowchart illustrating the process of a rapid assessment method for the seismic toughness of substations according to a specific embodiment of the present invention.

[0032] Figure 2a This is a general plan view of a 220kV substation according to a specific embodiment of the present invention;

[0033] Figure 2b yes Figure 2a Side view of the high-voltage side equipment in section 1-1;

[0034] Figure 2c yes Figure 2a Side view of the low-voltage side equipment in section 2-2;

[0035] Figure 3 This is a directed acyclic graph of a 220kV substation, a specific embodiment of the present invention.

[0036] Figure 4 This is the functional recovery function Q(t) of the 220kV substation in a specific embodiment of the present invention;

[0037] Figure 5 This refers to the seismic toughness index R of the 220kV substation in a specific embodiment of the present invention. Detailed Implementation

[0038] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0039] Figure 1 A flowchart illustrating the workflow of a rapid assessment method for the seismic toughness of substations according to a specific embodiment of the present invention is shown. Figure 1 As shown, this invention provides a rapid assessment method for the seismic toughness of substations based on Bayesian neural networks, comprising the following steps:

[0040] S1: Determine the seismic intensity of the area where the substation is located and use the peak ground acceleration (PGA) as an indicator to measure the seismic intensity; clarify the type, number, and system connection method of the substation equipment to complete the initial condition setting.

[0041] S2: Establish a Bayesian neural network for the substation system, and construct a directed acyclic graph (DAG) between the equipment and the busbars of the substation. Each node in the DAG represents an electrical device, and each line represents a busbar connecting the devices.

[0042] The reliability of the various types of equipment is expressed in the form of seismic vulnerability curves, which follow a log-normal cumulative distribution with a median of μ and a log-standard deviation of β, as shown in the following formula:

[0043] ;

[0044] The failure probability of the tested sub-node is obtained based on the reliability analysis of various types of equipment.

[0045] S3: Power users are classified into three levels of importance: Special Grade, Level 1, and Level 2. A user coefficient Q is assigned to each level of importance; specifically: Special Grade important user coefficient Q1 = 2, Level 1 important user coefficient Q2 = 1.5, and Level 2 important user coefficient Q3 = 1. A functional evaluation index K for the substation system is established based on the failure probability of the outgoing line units supplying various types of power users, as shown in the following formula:

[0046] ;

[0047] in: The important user coefficients for the qth outgoing line unit that can operate normally after the earthquake; The probability that i outgoing line units of the substation will operate normally after the earthquake; is the important user coefficient of the w-th outgoing unit of the substation; n is the total number of outgoing units of the substation system; the larger the system function index K is, the smaller the impact of earthquake on the substation's function and the better the system function.

[0048] S4: The substation repairs are carried out in descending order of equipment repair efficiency, with the equipment repair efficiency index P being... i As shown in the following formula:

[0049] ;

[0050] Where: P i K represents the repair efficiency index for the i-th type of equipment. n The post-earthquake functional level of the substation; K bi The functional level of the substation after the repair of equipment of type i; T i The repair time for the i-th type of equipment;

[0051] The post-earthquake functional recovery function Q(t) is calculated based on the substation functional evaluation index to characterize the change of the substation functional recovery level over time.

[0052] S5: Based on the substation functional recovery function Q(t), a quantitative standard for the seismic toughness index R of the substation is proposed, using the area enclosed by function and time as the standard. This index reflects the substation's ability to resist damage and recover quickly under seismic loads, as shown in the following formula:

[0053] ;

[0054] in: The time of the earthquake damage. The time required for the substation system to fully recover its functions;

[0055] The magnitude of the seismic toughness index R reflects the seismic toughness level of the substation system. The larger the R, the lower the seismic toughness level of the substation, resulting in greater resource consumption and economic losses.

[0056] This invention is aimed at Figure 2a The specific embodiment shown is the general layout plan of a 220kV substation. Figure 2b A side view of the high-voltage side equipment is shown; Figure 2c A side view of the low-voltage side equipment is shown. A rapid seismic toughness assessment is conducted, and based on China's seismic classification, the structural fortification intensity for this region is determined to be 7 degrees (0.15g). According to the importance of the power system, the fortification intensity needs to be increased by one degree, with a design basic seismic acceleration of 0.3g. The substation is mainly divided into five parts: 6 incoming line units, 2 high-voltage side busbar units, 3 transformer units, 2 low-voltage side busbar units, and 12 outgoing line units. Electrical energy enters the substation system through the incoming line units, is transmitted to the transformer units via the high-voltage side busbar units, and then to the power users via the outgoing line units through the low-voltage side busbar units. The substation has six types of equipment: DS-H, DS-V, CT, CB, TF, and PI, representing horizontal telescopic disconnect switches, vertical telescopic disconnect switches, current transformers, circuit breakers, transformers, and post insulators, respectively. The seismic intensity, substation equipment types, numbers, and system connection methods are determined to complete the initial condition setting for step S1.

[0057] Based on this, step S2 is performed to establish a Bayesian neural network model of the 220kV substation system, resulting in a directed acyclic graph of the substation, such as... Figure 3 As shown in Table 1, I1-I6 represent six incoming line units, B1 and B2 represent two high-voltage side busbar units, T1-T3 represent three transformer units, B3 and B4 represent two low-voltage side busbar units, and O1-O12 represent twelve outgoing line units. The median and logarithmic standard deviation parameters for each type of equipment are shown in Table 1. Based on the reliability analysis of each type of equipment, the failure probabilities of different functional states of the substation system are shown in Table 2.

[0058] Table 1:

[0059] ;

[0060] Table 2:

[0061] ;

[0062] Based on this, step S3 is performed to calculate and analyze the functional evaluation index K of the substation system, as shown in the following formula:

[0063] ;

[0064] in: The important user coefficients for the qth outgoing line unit that can operate normally after the earthquake; The probability that i outgoing line units of the substation will operate normally after the earthquake; is the important user coefficient of the wth outgoing line unit of the substation; n is the total number of outgoing line units in the substation system.

[0065] The calculated functional evaluation index K of the substation system under an earthquake intensity of 0.3g is 0.64.

[0066] Based on this, step S4 is performed. Substation repairs are carried out in descending order of equipment repair efficiency according to equipment category. The calculated equipment repair efficiency indices are shown in Table 3. This yields the equipment repair order and the functional recovery function Q(t), as follows: Figure 4 As shown, the functions of each device are gradually restored over time until they return to their initial functional level:

[0067] Table 3:

[0068] .

[0069] Based on this, S5 is proposed. According to the substation functional recovery function Q(t), the area enclosed by function and time is proposed as the quantitative standard for the seismic toughness index R of the substation. This index reflects the substation's ability to resist damage and recover quickly under seismic action, as shown in the following formula:

[0070] ;

[0071] in: The time of the earthquake damage. The time required for the substation system to fully recover its functions;

[0072] The calculated seismic toughness index of the substation system is R=3.376. The seismic toughness index R, as... Figure 5 As shown.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A rapid assessment method for the seismic toughness of substations based on Bayesian neural networks, characterized in that: The steps include the following: S1: Determine the seismic intensity of the area where the substation is located, and clarify the type, number, and system connection method of the substation equipment; S2: Establish a Bayesian neural network for the substation system, build a directed acyclic graph (DAG) between equipment and busbars in the substation, and obtain the failure probability of the tested sub-nodes based on the reliability analysis of various types of equipment. S3: Assign a criticality coefficient Q to power users of different importance levels, and establish a functional evaluation index K for the substation system based on the failure probability of outgoing line units supplying various types of power users, as shown in the following formula: ； in: The important user coefficients for the qth outgoing line unit that can operate normally after the earthquake; The probability that i outgoing line units of the substation will operate normally after the earthquake; is the important user coefficient of the wth outgoing unit of the substation; n is the total number of outgoing units of the substation system; the larger the system function index K is, the smaller the impact of earthquake on the substation's function and the better the system function. S4: Calculate the post-earthquake functional recovery function Q(t) of the substation based on the substation functional evaluation index to characterize the change of the substation functional recovery level over time; S5: Based on the substation functional recovery function Q(t), a quantitative standard for the seismic toughness index R of the substation is proposed, using the area enclosed by function and time as the standard. This index reflects the substation's ability to resist damage and recover quickly under seismic loads, as shown in the following formula: ； in: The time of the earthquake damage. The time required for the substation system to fully recover its functions; The magnitude of the seismic toughness index R reflects the seismic toughness level of the substation system. The larger the R, the lower the seismic toughness level of the substation, resulting in greater resource consumption and economic losses.

2. The rapid assessment method for seismic toughness of substations based on Bayesian neural networks according to claim 1, characterized in that: In step S1, the seismic intensity of the area where the substation is located is represented by the peak ground acceleration (PGA) and used as an indicator to measure the intensity of the earthquake.

3. The rapid assessment method for seismic toughness of substations based on Bayesian neural networks according to claim 1, characterized in that: In step S2, each node in the directed acyclic graph of the substation represents an electrical device, and each line represents a busbar connecting the devices.

4. The rapid assessment method for seismic toughness of substations based on Bayesian neural networks according to claim 1, characterized in that: In step S2, the reliability of various types of equipment is expressed in the form of seismic vulnerability curves. The seismic vulnerability curves follow a log-normal cumulative distribution with a median of μ and a log-standard deviation of β, as shown in the following formula: 。 5. The rapid assessment method for seismic toughness of substations based on Bayesian neural networks according to claim 1, characterized in that: In step S3, the electricity users are classified into special-grade, first-grade, and second-grade important electricity users, with a coefficient Q1=2 for special-grade important users, a coefficient Q2=1.5 for first-grade important users, and a coefficient Q3=1 for second-grade important users.

6. The rapid assessment method for seismic toughness of substations based on Bayesian neural networks according to claim 1, characterized in that: In step S4, the substation repairs are performed in descending order of equipment repair efficiency according to equipment type, where the equipment repair efficiency index P... i As shown in the following formula: ； Where: P i K represents the repair efficiency index for the i-th type of equipment. n The post-earthquake functional level of the substation; K bi The functional level of the substation after the repair of equipment of type i; T i Let be the repair time for the i-th type of device.

Images