Distribution network power failure risk assessment method considering emergency resource dynamic scheduling in strong windy and rainy weather

Abstract

The invention provides a distribution network power failure risk assessment method considering emergency resource dynamic scheduling in strong wind and rain weather. The method comprises the following steps: in extreme weather of strong wind and rain, firstly, predicting tower and lead element fault probabilities of a distribution line according to forecast data of wind speed and rainfall intensity, and combining to obtain the fault probability of the whole line; generating an original fault scene set of the power distribution network by adopting Latin hypercube sampling, clustering original scenes by adopting a k-medoids method, and constructing a representative fault scene set by using a final clustering center to realize scene screening; further, aiming at each scene, constructing a power distribution network recovery optimization model considering mobile energy storage and maintenance resource dynamic collaborative scheduling, solving to obtain an optimal recovery decision, and calculating a power failure risk index of each node; and after all scenes are traversed, performing statistics to obtain a final power failure risk index of the power distribution network. According to the method, the technical problem that the real risk level of the power distribution network is difficult to accurately and objectively reflect in the prior art is solved.

Description

Technical Field

[0001] The invention relates to the field of electric power meteorological risk assessment, and in particular to a distribution network power outage risk assessment method considering emergency resource scheduling under strong wind and rain weather. Background Art

[0002] As global climate change intensifies, the frequency and intensity of extreme weather events have increased significantly, posing severe challenges to the safe and stable operation of power equipment and systems. Distribution equipment is widely distributed and has poor tolerance to extreme environments. Extreme weather such as severe convection and typhoons often lead to distribution line failures, which in turn cause regional power outages, causing significant impacts on urban operations, social economy, and people's lives. Therefore, how to effectively assess the risk of power outages in distribution networks under extreme weather such as strong winds and rains, and formulate scientific prevention and reinforcement strategies based on this, has become an important issue that needs to be urgently addressed in the power industry.

[0003] At present, the following methods are mainly used for the risk assessment of power outages in distribution networks: (1) Methods based on statistical analysis of historical data, which collect and analyze historical failure data, statistically calculate the long-term average failure rate of equipment, and conduct quantitative analysis of system risks on this basis. However, this method is overly dependent on the integrity and accuracy of historical data, and it is difficult to reflect the impact of the high-impact, low-frequency characteristics of extreme weather events on the failure rate; (2) Methods based on physical models, which establish complex equipment stress analysis models and calculate the failure probability of equipment under different weather conditions. However, this method is complex to calculate and difficult to be widely used in actual engineering. Therefore, traditional methods have certain limitations when facing special situations such as high-impact, low-frequency extreme weather, including: extreme weather such as strong wind and rain often significantly increase the failure rate of distribution equipment, and the traditional method based on the long-term average failure rate is difficult to reflect the dynamic impact of such extreme weather on the failure rate; especially in actual operation, facing extreme weather and the sudden failure events caused by it, power companies usually dispatch emergency power supplies, maintenance resources and other resources to respond quickly, but the existing risk assessment methods do not include these emergency measures in the assessment system, which not only leads to an underestimation of the real risk level of the distribution network, but also makes it difficult to reflect the effect of reasonable dispatch of emergency resources on reducing risks, and cannot provide effective guidance for the emergency process. In other words, the existing technology has the technical problem of being difficult to accurately and objectively reflect the real risk level of the distribution network. Summary of the invention

[0004] In view of this, the present invention provides a distribution network outage risk assessment method considering dynamic scheduling of emergency resources under strong wind and rain weather, which can solve the technical problem that it is difficult to accurately and objectively reflect the real risk level of the distribution network in the prior art.

[0005] The present invention discloses a distribution network outage risk assessment method considering dynamic scheduling of emergency resources under strong wind and rain weather, comprising:

[0006] Calculation of distribution line failure probability. Through the power meteorological services provided by professional meteorological departments, obtain weather forecast data including wind speed and rainfall intensity in the distribution network area, and calculate the failure probability of the distribution line in a specific period of time in the future.

[0007] Generation and reduction of fault scenarios. Based on the distribution line fault probability, Latin hypercube sampling is used to generate the original fault scenario set of the distribution network. The K-medoids method is further used to cluster and reduce the scenario set, and the cluster center is used as the representative fault scenario set.

[0008] Power outage risk assessment considering emergency resource scheduling. For each fault scenario s, a distribution network restoration optimization model considering the dynamic coordinated scheduling of mobile energy storage and maintenance resources is established to solve the optimal restoration decision and calculate the corresponding power outage risk index Index based on the optimization results. s Traverse all fault scenarios and calculate the final power outage risk index Index.

[0009] The specific steps of calculating the probability of failure of the power distribution line are as follows:

[0010] 1) Calculate the failure probability of tower components. Consider the vulnerability of the tower as a log-normal distribution. 10,k , Forecast rainfall intensity R k Under this condition, the failure probability of a certain tower k can be calculated by the following formula:

[0011]

[0012] Where P tower,k (Collapse|V 10,k ,R k ) represents the tower k at wind speed V 10,k , rainfall intensity R k The failure probability under V 10,k is the 10-minute average wind speed (m / s) at the tower k at a height of 10 m, R k is the predicted rainfall intensity at tower k (mm / h). In order to consider the equivalent wind speed after the superposition of wind and rain loads, the forecast wind speed can be corrected according to the rainfall intensity.

[0013] Among them, μ k , σ k are the mean and variance of the logarithmic distribution (i.e. normal distribution) of the tower vulnerability. Assume that when the wind speed is lower than the tower design wind speed V dsn,k When the wind speed exceeds 2 times the design wind speed, the failure probability of the tower is 0; when the wind speed exceeds 2 times the design wind speed, the failure probability is 1. According to the "3σ" principle, it can be calculated by the following formula:

[0014]

[0015] 2) Calculate the failure probability of conductor elements. First, calculate the failure rate of conductor k between adjacent towers:

[0016]

[0017] In the formula, λ cond,k is the failure rate of conductor k (times / km·h); L k is the length of the conductor k (km); V dsn,k , R dsn,k are the design wind speed and rainfall intensity of conductor k respectively; α k , β k , γ k is a constant. Further, based on the failure rate, the failure probability of conductor k in the future T period can be calculated as:

[0018]

[0019] Where P cond,k (Collapse|V 10,k ,R k ) represents the wire k at wind speed V 10,k , rainfall intensity R k The probability of failure.

[0020] 3) Calculate the failure probability of the entire line. The entire distribution line can be regarded as a series system consisting of poles, towers and conductors. If any component fails, the entire system will fail. Based on the reliability principle of the series system, the failure probability of the entire distribution line can be comprehensively calculated by the failure probability of the poles and conductors, as shown below:

[0021]

[0022] Where P entire,i is the failure probability of line i; sets T(i) and C(i) represent the tower and conductor sets contained in line i respectively.

[0023] The generation and screening of the fault scenarios are specifically performed as follows:

[0024] 1) Generation of the original fault scenario set based on Latin hypercube sampling. According to the failure probability P of each line in the distribution network entire,i , Latin hypercube sampling is used to generate the original fault scenario set of the distribution network. In order to reduce the dimension of the scenario, the failure rate (P entire,i >ε, where ε is the set threshold) to sample the key lines, thereby reducing the number of redundant scenarios. Each scenario is represented by a vector consisting of the state values ​​(0 or 1) of each line.

[0025] 2) Fault scenario screening based on K-medoids clustering. After generating a large number of original fault scenarios of the distribution network, in order to further reduce the number of scenarios and improve the evaluation efficiency, and considering the characteristics of the scenarios as binary (0-1) data samples, the K-medoids clustering method based on Hamming distance is used to cluster the original fault scenario set, and the final cluster center is used as the representative fault scenario set for the subsequent risk assessment process. The specific process of this method is as follows:

[0026] ① Initialization. Determine the number of clusters K2, which represents the number of final scenarios that you want to obtain through reduction. In the data samples corresponding to the original fault scenario set of the distribution network, randomly select K2 data points (i.e., the state value vectors corresponding to the scenarios) as the initial cluster centers.

[0027] ② Data point allocation. Calculate the distance between each data point (i.e. each fault scenario) and all cluster centers. The distance is calculated using the Hamming distance as follows:

[0028]

[0029] In the formula, x and y are two data points, corresponding to two system failure scenarios. According to the Hamming distance, each data point is assigned to the cluster to which the nearest cluster center belongs.

[0030] ③ Cluster center update. For any cluster C k , calculate the distance from each point y to C k The sum of the Hamming distances of other points in :

[0031]

[0032] Select the point with the smallest total distance as the new cluster center medoid k, that is:

[0033]

[0034] ④ Cluster iteration. Return to execute ② and ③ until the cluster centers of each cluster no longer change. The final cluster center is used as the representative fault scenario set S of the distribution network to be evaluated. DMG .

[0035] For S DMG Each distribution network fault scenario in (convert the corresponding data points into L DMG , represents the set of fault lines in this scenario), and establishes a distribution network restoration optimization model that considers the dynamic coordinated scheduling of mobile energy storage and maintenance resources. Among them, the constraints are as follows.

[0036] Mobile behavior constraints of mobile energy storage:

[0037]

[0038] In the formula, β m,i,t is a 0-1 variable, β m,i,t =1 means that the mobile energy storage m stops and connects to the node i during time period t, otherwise β m,i,t =0;γ m,t is a 0-1 variable, γ m,t =1 indicates whether the mobile energy storage m is in a mobile state during the period t; M, T, N MESS They represent mobile energy storage, time period, and the set of nodes that mobile energy storage can access; C i,MESS represents the maximum number of mobile energy storage that node i can access; T m,ij is the time it takes for the mobile energy storage m to move from node i to j (in multiples of the step length Δt).

[0039] Operational behavior constraints of mobile energy storage:

[0040]

[0041]

[0042] In the formula, All are 0-1 variables. They respectively represent that the mobile energy storage m is in the charging and discharging states at the node i during the period t; They represent the charging and discharging power of mobile energy storage m at node i in period t, and their upper limit is Q m,i,t is the output reactive power of mobile energy storage m at node i during period t; S m,max is the apparent power capacity of the mobile energy storage m; SOC m,t is the state of charge of the mobile energy storage m at the end of time period t, and its allowable range is [SOC m,min ,SOC m,max ]; is the charging and discharging efficiency of mobile energy storage m; E m,max is the energy storage capacity of mobile energy storage m.

[0043] Maintenance resources (i.e., the combination of human, material and other resources required to complete maintenance) are similar to mobile energy storage, and are both movable resources. Therefore, maintenance resource scheduling includes two parts: mobile behavior and maintenance behavior decision. The mobile behavior constraints are:

[0044]

[0045] In the formula, the meanings of related variables and parameters are the same as above. R represents the set of maintenance resources; L DMG It is a set of failed lines.

[0046] Maintenance behavior constraints of maintenance resources:

[0047]

[0048] In the formula, κ i,t is a 0-1 variable, indicating the state of line i in period t, κ i,t =1 means the line is normal; parameter σ r,i It represents the maintenance rate of the maintenance resource r on the failed line i, reflecting the improvement of the health level of the line i within the unit period Δt (when the health level returns to 1, it is completely restored to normal). This parameter can be simply calculated based on the actual maintenance time; ζ i,t is the health improvement of line i in period t. Specifically, formula (25) indicates that the maintenance resources cannot leave before the line is completely repaired; formulas (26) and (27) indicate that the line can only return to normal when the health level exceeds 1; formula (28) indicates that once the line is repaired, secondary faults are no longer considered in subsequent periods. Here, line i has a similar meaning to line (i, i') below, both representing distribution lines, for ease of expression only.

[0049] Operation constraints of distribution network:

[0050]

[0051] Where P ii',t , Q ii',t is the active and reactive power of the distribution line (i, i') (i.e. the branch from node i to i'); δ i,t is a 0-1 variable, δ i,t =1 means that the load of node i is restored in period t; parameter P L,i,t , Q L,i,t represents the active and reactive loads of node i in time period t; Iverson brackets, when the condition in the brackets is met, the value is 1, otherwise it is 0; v i,t is the voltage amplitude of node i in period t, and its allowable range is [v i,min ,v i,max ],in Treated as a variable; parameter r ii' 、x ii' is the resistance and reactance of branch (i,i'); parameter S ii',max Represents the transmission capacity of line (i,i'). N and L are the nodes and lines of the distribution network respectively.

[0052] The goal of distribution network restoration is to maximize the restored load power while taking into account the dispatch cost of emergency resources. Therefore, the objective function is set as follows:

[0053]

[0054] In the formula, w i is the load weight of node i, which can be determined based on the experience of the power department or by using an objective weighting method; λ1 and λ2 represent the penalty coefficients for the scheduling cost of mobile energy storage and maintenance resources.

[0055] Finally, according to the distribution network restoration optimization results in various scenarios, the expected power outage duration of each node is used as the power outage risk indicator of the distribution network, which is calculated as follows:

[0056]

[0057] In the formula, Index i represents the expected power outage duration of node i; Prob s Indicates the probability corresponding to scene s, which can be expressed by the proportion of samples in the cluster where the cluster center of scene s is located; Index s,i represents the power outage duration of node i in scenario s; δ i,t,s In the scenario s, δ i,t The optimization result can be obtained by solving the above optimization model.

[0058] Due to the adoption of the above technical solution, the present invention has the following advantages:

[0059] First, in terms of fault probability modeling, an innovative coupling mechanism of pole tower and conductor element combination with wind and rain loads was introduced to accurately characterize the failure characteristics of distribution lines under strong wind and rain weather.

[0060] Secondly, in terms of fault scenario generation, an innovative method of Latin hypercube sampling combined with K-medoids clustering based on Hamming distance is adopted to ensure the uniform distribution of samples and achieve effective screening of scenarios.

[0061] Third, in terms of system recovery modeling, the dynamic coordinated scheduling of mobile energy storage and maintenance resources is incorporated into the risk assessment framework for the first time. By establishing a unified optimization model, not only the mobility constraints and timing characteristics of resources are taken into account, but also the dynamic process of energy storage charging and discharging and line repair is described.

[0062] Fourth, in terms of evaluation indicators, an innovative risk quantification method based on the expected power outage duration is proposed. By comprehensively considering the scenario probability and node recovery sequence, it not only reflects the overall risk level of the system, but also reflects the differentiated reliability requirements of different users.

[0063] It can be seen that the present invention brings the following technical effects:

[0064] 1) Embedding the dynamic dispatch of emergency resources such as mobile energy storage and maintenance resources and the optimized recovery process of the distribution network after a distribution network failure into the risk assessment process can more accurately and objectively reflect the actual risk level of the distribution network.

[0065] 2) The established distribution network restoration optimization model considering the dynamic coordinated dispatch of mobile energy storage and maintenance resources is a mixed integer second-order cone optimization model, which can effectively achieve the global optimal solution using existing solver software.

[0066] 3) The proposed method can further derive an emergency decision library for the distribution network. When the actual fault state matches one of the aforementioned scenarios, the corresponding recovery decision can be directly called and implemented, effectively improving the recovery efficiency and resilience of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0068] Figure 1 It is a schematic diagram of the framework of the method flow of the present invention.

[0069] Figure 2 Schematic diagram of the IEEE 37-bus distribution network test system.

[0070] Figure 3 The expected power outage duration indicator calculation result for each node. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described in combination with the accompanying drawings in the embodiments of the present invention. The present invention discloses a distribution network blackout risk assessment method considering dynamic scheduling of emergency resources under strong wind and rain weather, such as Figure 1 As shown, the specific steps are as follows:

[0072] Step 1: Calculate the probability of distribution line failure. Through the power meteorological service provided by the professional meteorological department, obtain the weather forecast data including wind speed and rainfall intensity in the distribution network area, and calculate the failure probability of the distribution line in a specific period in the future.

[0073] Step 2: Generation and reduction of fault scenarios. Based on the distribution line fault probability, Latin hypercube sampling is used to generate the original fault scenario set of the distribution network. The K-medoids method is further used to cluster and reduce the scenario set, and the cluster center is used as the representative fault scenario set.

[0074] Step 3: Power outage risk assessment considering emergency resource scheduling. For each fault scenario s, a distribution network restoration optimization model considering the dynamic coordinated scheduling of mobile energy storage and maintenance resources is established to obtain the optimal restoration decision. The corresponding power outage risk index Index is calculated based on the optimization result. s Traverse all fault scenarios and calculate the final power outage risk index Index.

[0075] The specific steps of calculating the probability of failure of the power distribution line are as follows:

[0076] 1) Calculate the failure probability of tower components. Consider the tower vulnerability as a log-normal distribution. 10,k , Forecast rainfall intensity R k Under this condition, the failure probability of a certain tower k can be calculated by the following formula:

[0077]

[0078] Where P tower,k (Collapse|V 10,k ,R k ) represents the tower k at wind speed V 10,k , rainfall intensity R k The failure probability under V 10,k is the 10-minute average wind speed (m / s) at the tower k at a height of 10 m, R k is the predicted rainfall intensity at tower k (mm / h). In order to consider the equivalent wind speed after the superposition of wind and rain loads, the forecast wind speed can be corrected according to the rainfall intensity. For example, referring to the existing method, it can be calculated as:

[0079]

[0080] In the formula, a 1,k 、b 1,k 、c 1,k 、a 2,k 、c 2,k is a constant, which is determined by fitting the structural test data of the actual tower.

[0081] Among them, μ k , σ k are the mean and variance of the logarithmic distribution (i.e. normal distribution) of the tower vulnerability. Assume that when the wind speed is lower than the tower design wind speed V dsn,k When the wind speed exceeds 2 times the design wind speed, the failure probability of the tower is 0; when the wind speed exceeds 2 times the design wind speed, the failure probability is 1. According to the "3σ" principle, it can be calculated by the following formula:

[0082]

[0083] 2) Calculate the failure probability of conductor elements. First, calculate the failure rate of conductor k between adjacent towers:

[0084]

[0085] In the formula, λ cond,k is the failure rate of conductor k (times / km·h); L k is the length of the conductor k (km); V dsn,k , R dsn,k are the design wind speed and rainfall intensity of conductor k respectively; α k , β k , γ k is a constant. Further, based on the failure rate, the failure probability of conductor k in the future T period can be calculated as:

[0086]

[0087] Where P cond,k (Collapse|V 10,k ,R k ) represents the wire k at wind speed V 10,k , rainfall intensity R k The probability of failure.

[0088] 3) Calculate the failure probability of the entire line. The entire distribution line can be regarded as a series system consisting of poles, towers and conductors. If any component fails, the entire system will fail. Based on the reliability principle of the series system, the failure probability of the entire distribution line can be comprehensively calculated by the failure probability of the poles and conductors, as shown below:

[0089]

[0090] Where P entire,i is the failure probability of line i; sets T(i) and C(i) represent the tower and conductor sets contained in line i respectively.

[0091] The generation and screening of the fault scenarios are specifically performed as follows:

[0092] 1) Generation of the original fault scenario set based on Latin hypercube sampling. According to the failure probability P of each line in the distribution network entire,i , Latin hypercube sampling is used to generate the original fault scenario set of the distribution network. In order to reduce the dimension of the scenario, the failure rate (P entire,i >ε, where ε is the set threshold) to sample the key lines, thereby reducing the number of redundant scenarios. Each scenario is represented by a vector consisting of the state values ​​(0 or 1) of each line.

[0093] 2) Fault scenario screening based on K-medoids clustering. After generating a large number of original fault scenarios of the distribution network, in order to further reduce the number of scenarios and improve the evaluation efficiency, and considering the characteristics of the scenarios as binary (0-1) data samples, the K-medoids clustering method based on Hamming distance is used to cluster the original fault scenario set, and the final cluster center is used as the representative fault scenario set for the subsequent risk assessment process. The specific process of this method is as follows:

[0094] ① Initialization. Determine the number of clusters K2, which represents the number of final scenarios that you want to obtain through reduction. In the data samples corresponding to the original fault scenario set of the distribution network, randomly select K2 data points (i.e., the state value vectors corresponding to the scenarios) as the initial cluster centers.

[0095] ② Data point allocation. Calculate the distance between each data point (i.e. each fault scenario) and all cluster centers. The distance is calculated using the Hamming distance as follows:

[0096]

[0097] In the formula, x and y are two data points, corresponding to two system failure scenarios. According to the Hamming distance, each data point is assigned to the cluster to which the nearest cluster center belongs.

[0098] ③ Cluster center update. For any cluster C k , calculate the distance from each point y to C k The sum of the Hamming distances of other points in :

[0099]

[0100] Select the point with the smallest total distance as the new cluster center medoid k, that is:

[0101]

[0102] ④ Cluster iteration. Return to execute ② and ③ until the cluster centers of each cluster no longer change. The final cluster center is used as the representative fault scenario set S of the distribution network to be evaluated. DMG .

[0103] For S DMG Each distribution network fault scenario in (convert the corresponding data points into L DMG , represents the set of fault lines in this scenario), and establishes a distribution network restoration optimization model that considers the dynamic coordinated scheduling of mobile energy storage and maintenance resources. Among them, the constraints are as follows.

[0104] Mobile behavior constraints of mobile energy storage:

[0105]

[0106] In the formula, β m,i,t is a 0-1 variable, β m,i,t =1 means that the mobile energy storage m stops and connects to the node i during time period t, otherwise β m,i,t =0;γ m,t is a 0-1 variable, γ m,t =1 indicates whether the mobile energy storage m is in a mobile state during the period t; M, T, N MESS They represent mobile energy storage, time period, and the set of nodes that mobile energy storage can access; C i,MESS represents the maximum number of mobile energy storage that node i can access; T m,ij is the time it takes for the mobile energy storage m to move from node i to j (in multiples of the step length Δt).

[0107] Operational behavior constraints of mobile energy storage:

[0108]

[0109]

[0110] In the formula, All are 0-1 variables. They respectively represent that the mobile energy storage m is in the charging and discharging states at the node i during the period t; They represent the charging and discharging power of mobile energy storage m at node i in period t, and their upper limit is Q m,i,t is the output reactive power of mobile energy storage m at node i during period t; S m,max is the apparent power capacity of the mobile energy storage m; SOC m,t is the state of charge of the mobile energy storage m at the end of time period t, and its allowable range is [SOC m,min ,SOC m,max ]; is the charging and discharging efficiency of mobile energy storage m; E m,max is the energy storage capacity of mobile energy storage m.

[0111] Maintenance resources (i.e., the combination of human, material and other resources required to complete maintenance) are similar to mobile energy storage, and are both movable resources. Therefore, maintenance resource scheduling includes two parts: mobile behavior and maintenance behavior decision. The mobile behavior constraints are:

[0112]

[0113] In the formula, the meanings of related variables and parameters are the same as above. R represents the set of maintenance resources; L DMG It is a set of failed lines.

[0114] Maintenance behavior constraints of maintenance resources:

[0115]

[0116] In the formula, κ i,t is a 0-1 variable, indicating the state of line i in period t, κ i,t =1 means the line is normal; parameter σ r,i It represents the maintenance rate of the maintenance resource r on the failed line i, reflecting the improvement of the health level of the line i within the unit period Δt (when the health level returns to 1, it is completely restored to normal). This parameter can be simply calculated based on the actual maintenance time; ζ i,t is the health improvement of line i in period t. Specifically, formula (25) indicates that the maintenance resources cannot leave before the line is completely repaired; formulas (26) and (27) indicate that the line can only return to normal when the health level exceeds 1; formula (28) indicates that once the line is repaired, secondary faults are no longer considered in subsequent periods. Here, line i has a similar meaning to line (i, i') below, both representing distribution lines, for ease of expression only.

[0117] Operation constraints of distribution network:

[0118]

[0119] Where P ii',t , Q ii',t is the active and reactive power of the distribution line (i, i') (i.e. the branch from node i to i'); δ i,t is a 0-1 variable, δ i,t =1 means that the load of node i is restored in period t; parameter P L,i,t , Q L,i,t represents the active and reactive loads of node i in time period t; Iverson brackets, when the condition in the brackets is met, the value is 1, otherwise it is 0; v i,t is the voltage amplitude of node i in period t, and its allowable range is [v i,min ,v i,max ],in Treated as a variable; parameter r ii' 、x ii' is the resistance and reactance of branch (i,i'); parameter S ii',max Represents the transmission capacity of line (i,i'). N and L are the nodes and lines of the distribution network respectively.

[0120] The goal of distribution network restoration is to maximize the restored load power while taking into account the dispatch cost of emergency resources. Therefore, the objective function is set as follows:

[0121]

[0122] In the formula, wi is the load weight of node i, which can be determined based on the experience of the power department or by using an objective weighting method; λ1 and λ2 represent the penalty coefficients for the scheduling cost of mobile energy storage and maintenance resources.

[0123] Finally, according to the distribution network restoration optimization results in various scenarios, the expected power outage duration of each node is used as the power outage risk indicator of the distribution network, which is calculated as follows:

[0124]

[0125] In the formula, Index i represents the expected power outage duration of node i; Prob s Indicates the probability corresponding to scene s, which can be expressed by the proportion of samples in the cluster where the cluster center of scene s is located; Index s,i represents the power outage duration of node i in scenario s; δ i,t,s In the scenario s, δ i,t The optimization result can be obtained by solving the above optimization model.

[0126] Next, the proposed method is verified based on the IEEE 37-node distribution network test system, and the node load value with a load of 0 in the original data is modified to a positive value greater than 0. Assume that according to the meteorological conditions or the distribution of meteorological stations in the distribution network, it can be divided into three areas, such as Figure 2 As shown. The meteorological conditions in each region are considered to be the same. It is assumed that the wind speed and rainfall intensity forecast data for each region in the next 6 hours are as follows: 32m / s, 80mm / h (region 1), 30m / s, 50mm / h (region 2), 25m / s, 40mm / h (region 3). According to operating experience, the line length is uniformly taken as 300m and the span is 50m. According to the aforementioned invention content, the failure probability of the entire line in each region can be calculated as 40.8%, 13.74%, and 0.11%, respectively. Based on the failure probability, Latin hypercube sampling is used to generate 1000 initial scenarios, and the number is reduced to 50 through k-medoids clustering. The cluster centers constitute the final representative set of fault scenarios to be evaluated.

[0127] Further, for each fault scenario, the distribution network restoration optimization model described in the above invention content is established, where the optimization time domain is set to 6h and the step length is 30min; it includes three mobile energy storages with a capacity of 500kW / 1000kWh and three groups of maintenance resources. The maintenance efficiency of each group of maintenance resources is set to 0.5, that is, it takes 2 steps (1h) to repair a faulty line. The access point of the mobile energy storage is set as follows Figure 2 As shown by the blue circle, each mobile energy storage stays at node 3 at the initial moment.

[0128] By adopting the technical solution proposed in the present invention, the power outage risk index of the distribution network is finally calculated as follows: Figure 3 As shown. The results show that after considering the optimal scheduling of emergency resources such as mobile energy storage and maintenance resources, nodes 7, 10, 11, 13-17 are still at a relatively high power outage risk level, with an expected power outage duration of more than 1.25h, among which the expected power outage duration of node 15 is more than 1.75h. By summarizing the recovery process results of each scenario, it can be seen that most of these risk-weak nodes are located in area 1 with more severe weather and are distributed at the edge of the distribution network. Affected by the higher probability of failure, this area is more likely to be split into multiple small-scale load islands with low load levels. In the recovery process, limited maintenance resources will be invested in the repair of "trunk" branches (such as branches 3-6 and 6-22) first, because the repair of these branches will enable more loads in the downstream larger area to quickly restore power supply; similarly, limited mobile energy storage is also more inclined to be deployed in areas 2 and 3, because the probability of line failure in these areas is relatively low, the scale of the formed load islands is relatively larger, and the capacity of mobile energy storage will be fully utilized, that is, more loads will be restored. Combining the above reasons, area 1 has more nodes with higher expected power outage duration. The above results can provide many practical guidance, such as targeted strengthening of edge nodes, including configuring local self-supplied distributed power sources, building interconnection lines, or appropriately increasing their load weights in recovery optimization and ensuring the fairness of the recovery process.

[0129] The technical principle of the present invention can be analyzed from the following aspects:

[0130] 1. Principle of failure probability calculation

[0131] The present invention proposes to model the tower vulnerability as a log-normal distribution, which has clear physical significance. On the one hand, the log-normal distribution can reflect the non-negative characteristics of equipment strength; on the other hand, its skewed characteristics are consistent with the actual observed fault distribution law. By introducing the concept of equivalent wind speed, the effect of rainfall on wind load is cleverly equivalent to wind speed correction, which simplifies the calculation and maintains physical rationality.

[0132] In the calculation of the probability of conductor failure, an exponential model based on failure rate is used. This model form is derived from the stress-strength interference principle in reliability theory, which can accurately describe the interaction between external loads and equipment carrying capacity. By introducing the time dimension, it can also reflect the cumulative effect of failures, making the evaluation results closer to reality.

[0133] 2. Scenario Generation and Screening Principles

[0134] The essence of Latin hypercube sampling is a stratified sampling technique, which ensures uniform distribution of samples in the entire value range by dividing the probability space into equal parts and randomly sampling in each subspace. This method is particularly suitable for the multidimensional random variable sampling problem in the present invention, and can achieve better probability coverage with fewer samples.

[0135] K-medoids clustering selects Hamming distance as the similarity metric, which is based on the binary characteristics of the fault scenario. Compared with traditional K-means, K-medoids is not sensitive to outliers, and the cluster center must be the actual scene, which is more in line with engineering practice. Through iterative optimization, the final cluster center can represent the characteristics of the original scene set to the greatest extent.

[0136] 3. Principle of recovery optimization model

[0137] The model adopts the form of mixed integer programming, unifying discrete scheduling decisions and continuous network status into one framework. The setting of constraints fully considers physical characteristics and engineering practice: the charging and discharging characteristics of mobile energy storage are characterized by power and capacity constraints; the effect of maintenance resources is expressed by the health increment model; and network operation is guaranteed by power flow equations and safety constraints.

[0138] The design of the objective function reflects the multi-objective trade-off idea, which considers both the load recovery benefit and the resource scheduling cost. By introducing the weight coefficient, the importance of different objectives can be flexibly adjusted to adapt to different decision preferences.

[0139] 4. Principles of Risk Indicator Calculation

[0140] The design of the expected power outage duration indicator is based on probability theory and reliability theory. By combining the scenario probability with the recovery sequence, it reflects both the randomness of the system and the dynamic recovery process. The calculation of the indicator takes into account the node differences and can guide differentiated risk management strategies.

[0141] In general, the technical principles of each link of the present invention have a rigorous theoretical basis and clear physical meaning, and the logical relationship between each part is close, forming a complete evaluation system. This systematic technical solution ensures the scientificity and reliability of the evaluation results.

[0142] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A distribution network outage risk assessment method considering dynamic scheduling of emergency resources under strong wind and rain weather, characterized in that: The following steps are involved: Through the power meteorological services provided by professional meteorological departments, weather forecast data including wind speed and rainfall intensity in the distribution network area are obtained to calculate the failure probability of distribution lines in a specific period in the future; Based on the distribution line fault probability, Latin hypercube sampling is used to generate the original fault scenario set of the distribution network. The K-medoids method is further used to cluster and screen the scenario set, and the cluster center is used as the representative fault scenario set. For each fault scenario s, a distribution network restoration optimization model considering the dynamic coordinated scheduling of mobile energy storage and maintenance resources is established to solve the optimal restoration decision and calculate the corresponding power outage risk index Index based on the optimization results. s ; Traverse all fault scenarios and obtain the final power outage risk index Index.

2. The method according to claim 1, characterized in that The distribution line failure probability calculation is specifically as follows: 1) Calculate the failure probability of tower components; consider the vulnerability of the tower as a log-normal distribution, and 10,k , Forecast rainfall intensity R k Under this condition, the failure probability of a certain tower k is calculated by the following formula: Where P tower,k (Co ll apse|V 10,k ,R k ) represents the tower k at wind speed V 10,k , rainfall intensity R k Failure probability under V 10,k is the 10-minute average wind speed (m / s) at the tower k at a height of 10 m, R k is the predicted rainfall intensity at tower k (mm / h); In order to consider the equivalent wind speed after the superposition of wind and rain loads, the forecast wind speed is corrected according to the rainfall intensity; Among them, μ k , σ k are the mean and variance of the logarithmic distribution (i.e., normal distribution) of the tower vulnerability. Assuming that when the wind speed is lower than the tower design wind speed V dsn,k When the wind speed exceeds 2 times the design wind speed, the failure probability of the tower is 0; when the wind speed exceeds 2 times the design wind speed, the failure probability is 1. According to the "3σ" principle, it is calculated by the following formula: 2) Calculate the failure probability of conductor elements; first calculate the failure rate of conductor k between adjacent towers: In the formula, λ cond,k is the failure rate of conductor k (times / km·h); L k is the length of the conductor k (km); V dsn,k , R dsn,k are the design wind speed and rainfall intensity of conductor k respectively; α k , β k , γ k is a constant; further, according to the failure rate, the failure probability of conductor k in the future T period is calculated as: Where P cond,k (Co ll apse|V 10,k ,R k ) represents the wire k at wind speed V 10,k , rainfall intensity R k The probability of failure under 3) Calculate the failure probability of the entire line; the entire distribution line is regarded as a series system consisting of poles, towers and conductors. If any component fails, the entire system will fail. Based on the reliability principle of the series system, the failure probability of the entire distribution line is comprehensively calculated by the failure probability of the poles and conductors, as shown below: Where P entire,i is the failure probability of line i; sets T(i) and C(i) represent the tower and conductor sets contained in line i respectively.

3. The method according to claim 1, characterized in that The generation and screening of the fault scenarios are specifically as follows: 1) Generate the original fault scenario set based on Latin hypercube sampling; according to the failure probability P of each line in the distribution network entire,i , Latin hypercube sampling is used to generate the original fault scenario set of the distribution network; in order to reduce the dimension of the scenario, the failure rate is relatively large (P entire,i >ε, where ε is the set threshold) to sample the key lines, thereby reducing the number of redundant scenarios; each scenario is represented by a vector consisting of the state values ​​(0 or 1) of each line; 2) Fault scenario screening based on K-medoids clustering: After generating a large number of original distribution network fault scenarios, in order to further reduce the number of scenarios and improve the evaluation efficiency, and considering the characteristic that the scenarios are binary (0-1) data samples, the K-medoids clustering method based on Hamming distance is used to cluster the original fault scenario set, and the final cluster center is used as the representative fault scenario set for the subsequent risk assessment process.

4. The method according to claim 3, characterized in that The fault scenario screening based on K-medoids clustering is specifically as follows: 1) Initialization: Determine the number of clusters K2, which represents the number of final scenarios that are expected to be obtained through reduction; randomly select K2 data points (i.e., the state value vectors corresponding to the scenarios) as the initial cluster centers in the data samples corresponding to the original fault scenario set of the distribution network; 2) Data point allocation: Calculate the distance between each data point (i.e., each fault scenario) and all cluster centers. The distance is calculated using the Hamming distance as follows: In the formula, x and y are two data points, corresponding to two system failure scenarios; according to the Hamming distance, each data point is assigned to the cluster to which the nearest cluster center belongs; 3) Cluster center update; for any cluster C k , calculate the distance from each point y to C k The sum of the Hamming distances of other points in : Select the point with the smallest total distance as the new cluster center medoid k, that is: 4) Cluster iteration; return to execute 2) and 3) until the cluster centers of each cluster no longer change; the final cluster center is used as the representative fault scenario set S of the distribution network to be evaluated DMG .

5. The method according to claim 1, characterized in that The distribution network restoration optimization model considering the dynamic coordinated dispatch of mobile energy storage and maintenance resources is specifically: For S DMG Each distribution network fault scenario in (convert the corresponding data points into L DMG , represents the set of fault lines in this scenario), and establishes a distribution network restoration optimization model that considers the dynamic coordinated scheduling of mobile energy storage and maintenance resources; where the constraints are as follows; Mobile behavior constraints of mobile energy storage: In the formula, β m,i,t is a 0-1 variable, β m,i,t =1 means that the mobile energy storage m stops and connects to the node i during time period t, otherwise β m,i,t =0;γ m,t is a 0-1 variable, γ m,t =1 indicates whether the mobile energy storage m is in a mobile state during the period t; M, T, N MESS They represent mobile energy storage, time period, and the set of nodes that mobile energy storage can access; C i,MESS represents the maximum number of mobile energy storage that node i can access; T m,ij is the time it takes for the mobile energy storage m to move from node i to j (in multiples of the step length Δt); Operational behavior constraints of mobile energy storage: In the formula, All are 0-1 variables. They respectively represent that the mobile energy storage m is in the charging and discharging states at the node i during the period t; They represent the charging and discharging power of mobile energy storage m at node i in period t, and their upper limit is Q m,i,t is the output reactive power of mobile energy storage m at node i during period t; S m,max is the apparent power capacity of the mobile energy storage m; SOC m,t is the state of charge of the mobile energy storage m at the end of time period t, and its allowable range is [SOC m,min ,SOC m,max ]; is the charging and discharging efficiency of mobile energy storage m; E m,max is the energy storage capacity of mobile energy storage m; Maintenance resources (i.e., the combination of human, material and other resources required to complete maintenance) are similar to mobile energy storage and are both movable resources. Therefore, maintenance resource scheduling includes two parts: mobile behavior and maintenance behavior decision. The mobile behavior constraints are: In the formula, the meanings of related variables and parameters are the same as above; R represents the set of maintenance resources; L DMG is the set of failed lines; Maintenance behavior constraints of maintenance resources: In the formula, κ i,t is a 0-1 variable, indicating the state of line i in period t, κ i,t =1 means the line is normal; parameter σ r,i It represents the maintenance rate of the maintenance resource r on the failed line i, reflecting the improvement of the health level of the line i within the unit period Δt (when the health level returns to 1, it is completely restored to normal), which is simply calculated based on the actual maintenance time; ζ i,t is the health level improvement of line i in period t; where, formula (25) indicates that the maintenance resources cannot leave before the line is completely repaired; formulas (26) and (27) indicate that the line can only return to normal when the health level exceeds 1; formula (28) indicates that once the line is repaired, secondary faults are no longer considered in subsequent periods; here, line i has a similar meaning to line (i, i') below, both representing distribution lines, just for the convenience of expression; Operation constraints of distribution network: Where P ii',t , Q ii',t is the active and reactive power of the distribution line (i, i') (i.e. the branch from node i to i'); δ i,t is a 0-1 variable, δ i,t =1 means that the load of node i is restored in period t; parameter P L,i,t , Q L,i,t represents the active and reactive loads of node i in time period t; Iverson brackets, when the condition in the brackets is met, the value is 1, otherwise it is 0; v i,t is the voltage amplitude of node i in period t, and its allowable range is [v i,min ,v i,max ],in Treated as a variable; parameter r ii' 、x ii' is the resistance and reactance of branch (i,i'); parameter S ii',max represents the transmission capacity of line (i, i'); N and L are the nodes and lines of the distribution network respectively; The goal of distribution network restoration is to maximize the restored load power while taking into account the dispatch cost of emergency resources; therefore, the objective function is set as follows: In the formula, w i is the load weight of node i, which is determined based on the experience of the power department or by the objective weighting method; λ1 and λ2 represent the penalty coefficients of the scheduling cost of mobile energy storage and maintenance resources.

6. The method according to claim 1, characterized in that The power outage risk indicators are specifically: According to the distribution network restoration optimization results in various scenarios, the expected power outage duration of each node is used as the power outage risk indicator of the distribution network, which is calculated as follows: In the formula, Index i represents the expected power outage duration of node i; Prob s Indicates the probability corresponding to scene s, which is represented by the proportion of samples in the cluster where the cluster center of scene s is located; Index s,i represents the power outage duration of node i in scenario s; δ i,t,s In the scenario s, δ i,t The optimization result is obtained by solving the above optimization model.

Images