Early warning and treatment method for the contradiction between tree lines and power lines in rural distribution networks considering severe convective weather

Abstract

The present invention discloses a tree-line conflict early warning and processing method for township distribution networks considering severe convective weather. Based on SVM, a severe convective weather-tree-line conflict mapping model for each area of the township distribution network is established, and the zoning early warning of the tree barrier grounding risk is realized. Based on the risk early warning results, the obstacle removal resource constraints, the obstacle removal team working time constraints and other conditions are comprehensively considered, and the overall economic investment and power outage loss are minimized. A double-layer optimization processing model for the tree-line conflict in the township distribution network taking into account the influence of severe convective weather is established to decide on the obstacle removal plan for each conflict-prone point. The method fully considers the regional correlation characteristics of severe convective weather and the obstacle removal conditions of the township power supply company, and can effectively explore the mapping relationship between severe convective weather and tree-line conflicts, provide a theoretical basis for the prevention and control of tree barrier grounding risks, and properly weigh the obstacle removal costs and power outage losses of each conflict-prone point, and give the most economical ex ante obstacle removal plan.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent power distribution, and more specifically, relates to a method for warning and handling the contradiction between tree and line in rural distribution networks considering severe convective weather. Background Technique

[0002] According to statistics, among many fault records, grounding faults caused by the contradiction between tree and line account for a large proportion. The main reasons are as follows: First, the terrain and landform of our country are complex, and rural areas are relatively scattered, resulting in rural distribution networks needing to cross mountainous areas, hilly areas, etc.; Second, overhead bare conductors are widely used as the carrier for power transmission in rural distribution networks. In order to save laying costs, "straightening" operations are often carried out, resulting in the lines passing through areas with a relatively high vegetation coverage rate. Although rural power supply companies will regularly patrol the lines and timely trim the trees that are too high and wide, it can only ensure the reliable operation of the distribution lines under normal weather conditions. Once encountering severe convective weather such as strong winds and heavy rains, due to factors such as line galloping, tree swaying, or even bending and overturning, the air between the tree and the line is extremely easy to break down, triggering grounding faults, which have a huge impact on the power supply of local areas or even the entire distribution line. In recent years, with the increasingly severe situation of ecological damage and global warming, severe convective weather such as strong winds and short-term heavy precipitation occurs frequently, posing a great threat to the safe operation of distribution networks, especially rural distribution networks. Therefore, it is necessary to study the risk warning and optimization treatment technology for the contradiction between tree and line in rural distribution networks considering the influence of severe convective weather to help the high-quality development of the power industry. Summary of the Invention

[0003] Aiming at the above defects or improvement requirements of the prior art, the present invention provides a method for warning and handling the contradiction between tree and line in rural distribution networks considering severe convective weather, aiming to effectively explore the mapping relationship between severe convective weather and the contradiction between tree and line, and give the most economical pre-obstacle-clearing plan.

[0004] To achieve the above object, according to the first aspect of the present invention, a method for warning and handling the contradiction between tree and line in rural distribution networks considering severe convective weather is provided, characterized by including:

[0005] S1, dividing the area according to the administrative division information of R townships to obtain R areas;

[0006] S2, establishing R SVM classification models, respectively preprocessing the historical severe convective meteorological monitoring information of the R areas to obtain R characteristic variables, using the R characteristic variables and the tree obstacle grounding fault flag of the i-th area under the historical severe convective weather as the training set, and training the i-th SVM classification model; i = 1, 2,..., R;

[0007] S3. Input the characteristic variables obtained after preprocessing the severe convective weather forecast information for each region into the corresponding trained SVM classification models respectively, to obtain the prediction results of tree obstacle grounding faults in each region and the corresponding fault probabilities.

[0008] S4. Use the tree obstacle grounding risk probability of the region where the tree-line conflict prone points are located as their fault probabilities; establish an optimization processing model for tree-line conflicts including a decision-making layer and a dispatching layer.

[0009] The decision-making layer includes a pre-clearance decision-making model, whose optimization objective is to minimize the sum of the pre-day clearance cost, the planned power outage loss caused by pre-day clearance, the clearance cost after a fault occurs in the distribution network during the day, and the power outage loss caused by the fault and clearance in the distribution network during the day. The constraint conditions include the number constraint of clearance operation points. Among them, the pre-day is normal weather, and the day is severe convective weather. The clearance cost after a fault occurs in the distribution network during the day and the power outage loss caused by the fault and clearance in the distribution network during the day are obtained according to the fault probabilities of each tree-line conflict prone point.

[0010] The dispatching layer includes a clearance team dispatching model, whose optimization objective is to minimize the sum of the pre-day clearance cost and the planned power outage loss caused by pre-day clearance. The constraint conditions include clearance resource constraint, power outage loss constraint, clearance duration constraint, working status constraint of clearance teams, working duration constraint of clearance teams, and repair duration constraint of clearance teams.

[0011] S5. The decision-making layer randomly generates the initial clearance decision results of each tree-line conflict prone point within the scope of the constraint conditions of the pre-clearance decision-making model and sends them to the dispatching layer. The dispatching layer solves the clearance team dispatching model according to the clearance decision results to obtain the work plan of the clearance teams, and feeds back the corresponding clearance cost of the work plan of the clearance teams to the decision-making layer. The decision-making layer combines the feedback clearance cost and the objective function of the pre-clearance decision-making model, randomly generates new clearance decision results, and after exhausting all the clearance decision results of the tree-line conflict prone points, obtains the optimal pre-clearance plan result.

[0012] According to the second aspect of the present invention, there is provided a tree-line conflict warning and processing method system for rural distribution networks considering severe convective weather, including: a computer-readable storage medium and a processor.

[0013] The computer-readable storage medium is used to store executable instructions.

[0014] The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method as described in the first aspect.

[0015] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0016] The township distribution network tree-line conflict early warning and handling method considering severe convective weather provided by the present invention aims at the risk early warning and obstacle clearing requirements of the tree-line conflict in the township distribution network before the severe convective weather arrives. Considering the regional correlation characteristics of the severe convective weather and the obstacle clearing conditions of the township power supply company, a set of "zoned early warning - overall weighing - optimized handling" mechanism is proposed. First, considering the regional correlation characteristics of the severe convective weather, the meteorological monitoring information of the entire power supply area of the township distribution network is used as the data to be classified, and whether a tree fault occurs and grounds is used as the label. Based on SVM, a mapping model of severe convective weather - tree-line conflict in each area of the township distribution network is established, realizing the zoned early warning of the risk of tree fault grounding. Secondly, based on the risk early warning results, comprehensively considering conditions such as the constraint of obstacle clearing resources and the working time constraint of the obstacle clearing team, with the goal of minimizing the overall economic investment and power outage loss, a two-layer optimized handling model of the tree-line conflict in the township distribution network considering the influence of severe convective weather is established to decide the obstacle clearing plan for each prone point of conflict. This design fully considers the regional correlation characteristics of the severe convective weather and the obstacle clearing conditions of the township power supply company, can effectively explore the mapping relationship between the severe convective weather and the tree-line conflict, provides a theoretical basis for the prevention and control of the risk of tree fault grounding, and properly weighs the obstacle clearing cost and power outage loss of each prone point of conflict, giving the most economical prior obstacle clearing plan. Description of the Drawings

[0017] Figure 1 It is the overall modeling flow chart of the present invention.

[0018] Figure 2 It is the structure diagram of the two-layer optimized handling model of the tree-line conflict provided by the present invention.

[0019] Figure 3 It is the topology, regional division result and location map of the prone points of the tree-line conflict of the improved IEEE 33-node system provided by the present invention.

[0020] Figure 4 It is the process diagram of the input data formation of the tree fault grounding risk early warning model provided by the present invention.

[0021] Figure 5 It is the relationship diagram of the minimum rest duration and continuous working duration of each obstacle clearing team provided by the present invention. Detailed Embodiment

[0022] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0023] At present, scholars at home and abroad have carried out a large number of studies on the risk early warning technology of distribution networks, providing an important basis for the operation and transformation of power grids. However, most of them only focus on urban distribution networks, rarely involving the actual operation of rural distribution networks, and do not consider the impact of tree-line conflicts. In terms of tree-fault monitoring and modeling, existing research can help line patrol personnel accurately control the safe distance between trees and conductors. However, most of them analyze from the growth characteristics of the trees themselves and do not consider the impact of meteorological factors on tree-faults.

[0024] Different from general sudden grounding faults, tree-fault grounding is an accumulative fault that can be predicted and pre-defended. Therefore, it is not possible to allow trees to grow wantonly. After determining the risk of touching the line by combining meteorological forecast information, it should be cleared as much as possible to avoid power outages. On the other hand, considering the tight time before the arrival of severe convective weather and the limited tree-clearing resources of rural power supply companies, planned power outages are also required during the pre-clearance process, which will also cause power outage losses. Therefore, it is necessary to combine the fault probabilities of tree-line conflict-prone points, weigh the losses of planned power outages and pre-clearance costs, as well as the losses of fault power outages and post-treatment costs, and make a proper decision on whether to take pre-clearance measures. For the conflict points with high risk of tree-fault grounding and huge power outage losses after the fault, priority should be given to treatment; on the contrary, for the conflict points with low probability of tree-fault grounding, small power outage losses after the fault and high pre-clearance costs, under the conditions of scarce tree-clearing resources and tight time, treatment can be postponed. Mathematically speaking, its essence is an optimization problem, but there is still little research on this problem. Based on this, the embodiments of the present invention provide a method for warning and handling tree-line conflicts in rural distribution networks, as Figure 1 shown, including:

[0025] S1, dividing the area according to the administrative division information of R townships to obtain R areas.

[0026] Specifically, on the premise of ensuring that all areas can independently obtain meteorological monitoring data, combined with administrative division information and the location of meteorological stations, comprehensively determine the number of regional divisions R of the rural distribution network and the boundaries of each area.

[0027] S2, establishing R SVM classification models, respectively preprocessing the historical severe convective meteorological monitoring information of the R areas to obtain R characteristic variables, using the R characteristic variables and the tree-fault grounding fault flag of the i-th area under the historical severe convective meteorology as the training set, and training the i-th SVM classification model; i = 1, 2,..., R.

[0028] Specifically, step S2 includes:

[0029] S21, obtaining the historical severe convective meteorological monitoring data and tree-fault grounding fault records of each area;

[0030] S22. Establish R SVM classification models, and use the data and fault records obtained in step S21 to train the R models respectively.

[0031] It can be understood that since each region has its own geographical and meteorological environment characteristics, respective SVM classification models are established for each region.

[0032] Preferably, align the severe convective weather monitoring information at each moment in the R regions according to the time scale, perform data coding and normalization, and obtain R characteristic variables.

[0033] Specifically, in step S2, the training process of the model is as follows:

[0034] Each model uses whether a tree fault grounding fault occurs in the local area as the class flag, where a fault is recorded as 1, and otherwise as -1; uses the severe convective weather monitoring information in the full power supply range of the rural distribution network as the data to be classified, with 6 variables in each region and a total of 6R variables in the R regions. First, align the severe convective weather monitoring information collected in each region according to the time scale, perform data coding on the information described in words such as wind direction and wind force level, and integrate the sampled values at each moment to form a local characteristic variable combination; second, perform normalization on the characteristic variable combinations in the R regions respectively, convert them into data between [0, 1], and integrate the characteristic variables of the meteorological information in the R regions to form the data to be classified; finally, integrate the tree fault grounding fault flag variables in each region with the data to be classified respectively, and input them into the corresponding SVM classification model, then the mapping relationship between severe convective weather and tree fault grounding faults can be mined by region.

[0035] That is, when training the i-th SVM classification model, the characteristic variables of the meteorological information in the R regions are used, but only the fault flag variable in the i-th region is used.

[0036] S3. Input the characteristic variables obtained after preprocessing the severe convective weather forecast information in each region into the corresponding trained SVM classification models respectively to obtain the prediction results of tree fault grounding faults in each region and the corresponding fault probabilities.

[0037] Preferably, both the monitoring information and the forecast information include temperature, humidity, wind speed, rainfall, wind direction and wind force level.

[0038] Specifically, input the meteorological forecast information before the arrival of severe convective weather into the R models respectively to obtain the tree fault grounding prediction results and the corresponding reliability probabilities in each region of the rural distribution network, and realize the early warning of the risk of tree-line conflict.

[0039] Use the tree fault grounding risk probability in each region (i.e., the fault probability of tree fault grounding faults in each region) as the fault probability of the tree-line conflict prone points in the region.

[0040] The above R SVM classification models are the tree obstacle grounding risk early warning models provided by the present invention.

[0041] S4. Use the tree obstacle grounding risk probability of the area where the tree-line conflict prone points are located as their failure probabilities; establish an optimized processing model for tree-line conflicts including a decision-making layer and a dispatching layer.

[0042] Specifically, in combination with the tree removal conditions of the township power supply company and the locations of the tree-line conflict prone points, set the parameters of the decision-making layer and the dispatching layer of the optimized processing model for tree-line conflicts respectively.

[0043] The decision-making layer includes a pre-fault tree removal decision model, and its optimization objective is to minimize the sum of the pre-fault tree removal cost, the planned power outage loss caused by pre-fault tree removal, the tree removal cost after a fault occurs in the distribution network during the day, and the power outage loss caused by the distribution network fault and tree removal during the day. The constraint conditions include the number constraint of tree removal operation points; among them, the pre-fault period is normal weather, and the in-day period is severe convective weather; the tree removal cost after a fault occurs in the distribution network during the day and the power outage loss caused by the distribution network fault and tree removal during the day are obtained according to the failure probabilities of each tree-line conflict prone point.

[0044] Specifically, the pre-fault tree removal decision model: the optimization objective is to minimize the sum of the overall economic input and power outage loss in the past two days (i.e., pre-fault and in-day). Among them, the economic input includes two categories: the pre-fault tree removal cost under normal weather (i.e., the pre-fault tree removal cost) and the tree removal cost after a fault occurs in the distribution network under severe convective weather (i.e., the tree removal cost after a fault occurs in the distribution network during the day), and the power outage loss also includes two categories: the planned power outage loss caused by tree removal under normal weather (i.e., the planned power outage loss caused by pre-fault tree removal) and the power outage loss faced during the distribution network fault and tree removal process under severe convective weather (i.e., the power outage loss caused by the distribution network fault and tree removal during the day).

[0045] That is, the objective function of the pre-fault tree removal decision model is:

[0046] min(C loss-extr +C clear-extr +C loss-norm +C clear-norm )(1)

[0047]

[0048]

[0049] In the above formula, C loss-extr is the power outage loss caused by the distribution network fault and the fault handling process under severe convective weather, and C clear-extr is the handling cost after the distribution network fault under severe convective weather. Both are related to the risk assessment results of each conflict prone point; Closs-norm For the planned power outage loss caused by line clearance under normal weather conditions, C clear-norm For the pre - clearance cost under normal weather conditions, both of which are obtained from the feedback of the lower - layer model (i.e., the line clearance team scheduling model). n is the total number of tree - line conflict prone points; P i extr is the failure probability of the tree - line conflict prone point i under severe convective weather; is the power outage loss per unit time of the tree - line conflict prone point i; T i fail-extr is the power outage duration of the tree - line conflict prone point i under severe convective weather; u i is a flag variable indicating whether the tree - line conflict prone point i takes line clearance operations, u i = 1 indicates that the pre - clearance measure is taken at the conflict prone point i, and u i = 0 indicates that the treatment measure is not taken at the conflict prone point i; is the line clearance cost per unit time under severe convective weather; T i clear-extr is the line clearance duration of the tree - line conflict prone point i under severe convective weather.

[0050] The constraint condition of the objective function is the number constraint of line clearance operation points:

[0051]

[0052] In the above formula, n max is the upper limit of the number of pre - clearance operation points.

[0053] This constraint condition can ensure that the optimal tree - line conflict handling result output by the model is within the range allowed by the actual line clearance conditions and economic conditions.

[0054] The scheduling layer includes a line clearance team scheduling model, and its optimization goal is to minimize the sum of the pre - clearance cost and the planned power outage loss caused by pre - clearance on the day before.

[0055] Specifically, the optimization goal of the line clearance team scheduling model is to minimize the sum of the line clearance operation cost on the day (i.e., the pre - clearance cost) and the planned power outage loss (i.e., the planned power outage loss caused by pre - clearance on the day before).

[0056] That is, the objective function of the line clearance team scheduling model is:

[0057] min(C clear-norm + C loss-norm ) (5)

[0058]

[0059]

[0060] In the above formula, N is the number of obstacle removal teams; T is the duration of a day, i.e., 24 hours; is the unit time working cost of the obstacle removal team j under normal weather conditions; is the working flag variable of the obstacle removal team j at time t, indicating that the obstacle removal team j is in a working state at the potential conflict point i at time t, otherwise it is in a rest state or in a working state at other potential conflict points; is the flag variable indicating whether obstacle removal measures are taken at the potential conflict point i at time t, indicating that the obstacle removal operation is in progress at the potential conflict point i at time t, then indicates that the potential conflict point i is not in an obstacle removal state at time t, and its relationship with is as shown in formula (26).

[0061] The constraint conditions of the objective function include obstacle removal resource constraint, power outage loss constraint, obstacle removal duration constraint, working state constraint of obstacle removal teams, working duration constraint of obstacle removal teams, and rest duration constraint of obstacle removal teams. Among them,

[0062] Obstacle removal resource constraint: This constraint condition is used to ensure that the resources such as manpower and material resources consumed for obstacle removal within the same time period do not exceed the maximum available resources.

[0063] Power outage loss constraint: This constraint condition is used to ensure that the planned power outage loss caused by obstacle removal within the same time period does not exceed the maximum tolerable loss.

[0064] Obstacle removal duration constraint: This constraint condition is used to ensure that the potential tree-line conflict points to be processed can be properly cleared and the safety hazards can be completely eliminated.

[0065] Working state constraint of obstacle removal teams: This constraint condition is used to ensure that at most one tree-line conflict point can be processed by the same obstacle removal team within the same time period, and when an obstacle removal measure is taken for a certain tree-line conflict point, there is exactly one obstacle removal team performing this task.

[0066] Working duration constraint of obstacle removal teams: This constraint condition is used to ensure that the working duration of each obstacle removal team does not exceed its capacity limit.

[0067] Rest duration constraint of obstacle removal teams: This constraint condition is used to ensure that each obstacle removal team can get sufficient rest after working for a period of time.

[0068] That is, the obstacle removal resource constraint:

[0069] The resources such as manpower and material resources consumed for obstacle removal within the same time period cannot exceed the maximum available resources.

[0070]

[0071] In the above formula, N max is the maximum number of work teams for obstacle clearance allowed in the same time period.

[0072] Power outage loss constraint:

[0073] The planned power outage loss caused by obstacle clearance within the same time period cannot exceed the maximum tolerable loss.

[0074]

[0075] In the above formula, C max is the maximum power outage loss allowed in the same time period.

[0076] Obstacle clearance duration constraint:

[0077]

[0078] In the above formula, T i,min is the minimum obstacle clearance duration at the tree-line conflict prone point i.

[0079] Obstacle clearance work team status constraint:

[0080]

[0081] This constraint indicates that: within the same time period, the same obstacle clearance work team can handle at most one tree-line conflict prone point.

[0082]

[0083] This constraint indicates that: when obstacle clearance operations are carried out at the tree-line conflict prone point i, there is exactly one work team performing the obstacle clearance task here.

[0084] Obstacle clearance work team working duration constraint:

[0085]

[0086] In the above formula, is the working duration of the obstacle clearance work team j at the tree-line conflict prone point i, is the remaining working duration of the obstacle clearance work team j at the tree-line conflict prone point i, is the maximum working duration of the obstacle clearance work team j. This constraint indicates that: the j-th obstacle clearance work team can work at most hours.

[0087] Obstacle clearance work team rest duration constraint:

[0088] After working for a period of time, each obstacle clearance work team must rest for at least a certain duration before it can continue to work.

[0089]

[0090] In the above formula, is the rest duration of the obstacle removal team j, is the minimum rest duration of the obstacle removal team j, and f pc (·) is a proportional function.

[0091] S5. The decision-making layer randomly generates the initial obstacle removal decision results of each tree-line conflict prone point within the constraint range of the pre-event obstacle removal decision model and sends them to the dispatching layer. The dispatching layer solves the obstacle removal team dispatching model according to the obstacle removal decision results to obtain the work plans of the obstacle removal teams, and feeds back the corresponding obstacle removal costs to the decision-making layer; The decision-making layer combines the obtained obstacle removal costs and the objective function of the pre-event obstacle removal decision model, randomly generates new obstacle removal decision results, and after exhausting all the obstacle removal decision results of the tree-line conflict prone points, obtains the optimal pre-event obstacle removal plan results.

[0092] Specifically, the above tree-line conflict optimization processing model including the decision-making layer and the dispatching layer can be solved by using exhaustive methods, particle swarm optimization algorithms, bacterial foraging algorithms, DDQN algorithms, etc., and calling commercial optimization software such as cplex, gurobi, mosek, etc.

[0093] Preferably, the particle swarm optimization algorithm is used to solve the pre-event obstacle removal decision model. As Figure 1-2 shown, step S5 includes:

[0094] S51. The decision-making layer uses the particle swarm optimization algorithm to initialize the positions and velocities of each particle within the constraint range of the pre-event obstacle removal decision model.

[0095] S52. Each particle transmits its position parameters to the dispatching layer. The dispatching layer solves the obstacle removal team dispatching model according to the transmitted parameters to obtain the work plans of each obstacle removal team, and feeds back the corresponding obstacle removal costs to the decision-making layer.

[0096] Preferably, the commercial optimization software gurobi is used to solve the obstacle removal team dispatching model.

[0097] Specifically, each particle transmits its position parameters to the dispatching layer. The dispatching layer solves the established obstacle removal team dispatching model and the transmitted parameters with the help of the commercial optimization software gurobi, obtains the work plans of each obstacle removal team, and feeds back the corresponding obstacle removal costs to the decision-making layer.

[0098] S53. The decision-making layer combines the obstacle removal costs and the objective function of the pre-event obstacle removal decision model to calculate the fitness of each particle.

[0099] S54. According to the fitness of each particle, update the individual optimal position and the current global optimal position, and combine the inertia factor and the acceleration constant to update the positions and velocities of each particle.

[0100] In S55, steps S52 - S54 are repeated until the set maximum number of iterations is reached or the result converges, and the optimal pre - obstacle - clearing plan result is output.

[0101] In summary, the method provided by the present invention proposes a set of "zoning early warning - overall trade - off - optimization processing" mechanisms. Among them, the tree - obstacle grounding risk early - warning model uses the meteorological monitoring information of the entire power - supply area of the rural distribution network as the data to be classified, and whether a tree - obstacle grounding occurs as the label. Based on SVM, the mapping relationship between severe convective weather and tree - obstacle grounding risk is mined, realizing the zoning early warning of tree - obstacle grounding risk; the tree - line conflict optimization - processing model is divided into a decision - making layer and a dispatching layer. The decision - making layer establishes a pre - obstacle - clearing decision - making model with the minimum overall economic input and power - outage loss during the day - ahead and intra - day (normal weather during the day - ahead and severe convective weather during the intra - day) as the objective function, judges whether pre - obstacle - clearing measures should be taken for each tree - line conflict prone point, and transmits the corresponding judgment results to the dispatching layer; the dispatching layer, based on the given pre - obstacle - clearing instructions, establishes a clearing - team dispatching model with the minimum clearing loss and operation cost on the day as the objective function, solves the work plans of each clearing team, and feeds back the corresponding pre - obstacle - clearing costs to the decision - making layer. At this time, the decision - making layer combines the cost results transmitted by the dispatching layer for comparison and optimization, and explores new tree - line conflict optimization - processing schemes. The double - layer model iterates repeatedly until the optimal tree - line conflict processing result is obtained under the premise of meeting the constraint conditions.

[0102] The method provided by the present invention is further described below with a specific example.

[0103] Embodiment 1:

[0104] Refer to Figure 1 , a method for early warning and handling of tree - line conflicts in rural distribution networks considering severe convective weather. This method takes the rural distribution network shown in Figure 3 as the implementation object (wherein, the grid network structure adopts an improved IEEE 33 - node system, there are branch - line circuit breakers at nodes 1 and 5, sectional switches at nodes 6, 13, and 28, and 10 tree - line conflict prone points are set, numbered ① - ⑩ respectively), and is carried out in the following steps in sequence:

[0105] Step 1: On the premise of ensuring that all regions can independently obtain meteorological monitoring data, combined with administrative division information and the positions of meteorological stations, comprehensively determine the number of regional divisions R of the rural distribution network and the boundaries of each region; the finally determined number of regional divisions R = 8, and the boundaries of each region are shown in Figure 3 .

[0106] Step 2: Obtain the historical severe convective weather monitoring data (including six variables: temperature, humidity, wind speed, rainfall, wind direction, and wind force level) and tree obstacle grounding fault records for each region. Considering that severe convective weather has a short duration, low occurrence frequency, and is concentrated in summer, to prevent its data characteristics from being overwhelmed by a large number of normal weather conditions, the present invention first counts the occurrence times of severe convective weather in eight regions of the township from 2018 to 2020 in summer, and intercepts the meteorological monitoring data for a total of three days before and after each day of severe convective weather occurrence for splicing and integration (for example, it can be arranged according to the steps shown in Figure 4 ), and jointly constitutes the undetermined input data set of the SVM with the tree obstacle grounding records at the corresponding time scale. The data time sampling accuracy is 1h, and there are a total of 1115 groups of data in three years. The undetermined input data set obtained by splicing and integration is a severely imbalanced data set, in which the data with tree obstacle grounding faults accounts for less than 10%. To improve the balance of the input samples and avoid affecting the training results of the SVM classification model, the present invention uses the SMOTE algorithm to "interpolate" to generate some minority class samples (i.e., tree obstacle grounding faults occur), replacing the majority class samples (i.e., no faults occur), as the final input data of the SVM classifier. It should be particularly noted that this replacement process does not change the size of the data set.

[0107] Step 3: Establish R SVM classification models, and use the data and fault records obtained in Step 2 to train the R models respectively; each model uses whether there is a tree obstacle grounding fault in the local area as the class label, with a fault recorded as 1 and otherwise recorded as -1; use the severe convective weather monitoring information in the entire power supply range of the township distribution network as the data to be classified, with 6 variables in each region and a total of 6R variables in R regions. First, align the severe convective weather monitoring information collected in each region according to the time scale, and perform data encoding on the information described in words such as wind direction and wind force level, and integrate the sampled values at each moment to form a local feature variable combination; secondly, perform normalization on the feature variable combinations of the R regions respectively, convert them into data between [0,1], and arrange and integrate them to form the data to be classified; finally, integrate the tree obstacle grounding fault flag variables in each region with the data to be classified respectively, and input them into the corresponding SVM classification model, and the mapping relationship between severe convective weather and tree obstacle grounding can be mined by region.

[0108] Step 4: Input the meteorological forecast information before the severe convective weather into the R models respectively, obtain the tree obstacle grounding prediction results and the corresponding reliability probabilities for each region of the township distribution network, and realize the early warning of the risk of tree-line conflict;

[0109] Step 5: Combine the tree removal conditions of the township power supply company and the locations of the prone points of tree-line conflicts, and set the decision-making layer and dispatching layer parameters of the tree-line conflict optimization processing model respectively; considering the limited tree removal resources of the township power supply company, the present invention takes the upper limit n of the number of pre-removal operation pointsmax = 5, the number of line clearance teams N = 3, the maximum number of line clearance teams allowed to work during the same period N max = 2, the maximum power outage loss C max = 10,000 yuan. The unit - time power outage loss and the minimum line clearance duration of each prone - to - contradiction point are shown in Table 1. The unit - time working cost and the maximum working duration of the line clearance teams are shown in Table 2. The relationship between the minimum rest duration and the continuous working duration is shown in Figure 5 .

[0110] Table 1 Unit - time power outage loss and minimum line clearance duration of each prone - to - contradiction point

[0111]

[0112] Table 2 Unit - time working cost and maximum working duration of each line clearance team

[0113]

[0114] Step 6: The decision - making layer uses the particle swarm optimization algorithm to initialize the positions and velocities of each particle within the constraint conditions of the pre - event line clearance decision - making model. The pre - event line clearance decision - making model takes the minimum of the sum of the overall economic input and power outage loss in the past two days as the optimization goal. Among them, the economic input includes two major categories: the pre - event line clearance cost under normal weather and the line clearance cost after the distribution network fails under severe convective weather. The power outage loss also includes two major categories: the planned power outage loss caused by line clearance under normal weather and the power outage loss faced during the distribution network failure and line clearance process under severe convective weather.

[0115] min(C loss-extr +C clear-extr +C loss-norm +C clear-norm )(15)

[0116]

[0117]

[0118] In the above formula, C loss-extr is the power outage loss caused by the distribution network failure and the fault handling process under severe convective weather, C clear-extr is the handling cost after the distribution network failure under severe convective weather, and both are related to the risk assessment results of each prone - to - contradiction point; C loss-norm is the planned power outage loss generated by line clearance under normal weather conditions, C clear-norm is the pre - event line clearance cost under normal weather conditions, and both are obtained by feedback from the lower - layer model. n is the total number of tree - line prone - to - contradiction points; P i extr is the failure probability of the tree - line prone - to - contradiction point i under severe convective weather; is the power outage loss per unit time at the tree-line conflict prone point i; T i fail-extr is the power outage duration at the tree-line conflict prone point i under severe convective weather; u i is a flag variable indicating whether the tree-line conflict prone point i takes clearance operation, u i = 1 means that the clearance measure is taken in advance at the conflict prone point i, u i = 0 means that the governance measure is not taken at the conflict prone point i; is the clearance cost per unit time under severe convective weather; T i clear-extr is the clearance duration at the tree-line conflict prone point i under severe convective weather.

[0119] The constraint condition of the objective function is the number constraint of clearance operation points:

[0120]

[0121] In the above formula, n max is the upper limit of the number of clearance operation points in advance. This constraint condition can ensure that the optimal tree-line conflict handling result output by the model is within the range allowed by the actual clearance conditions and economic conditions.

[0122] Step 7: Each particle transmits its position parameters to the scheduling layer. The scheduling layer, according to the established clearance team scheduling model and the transmitted parameters, uses the commercial optimization software gurobi to solve, obtains the work plans of each clearance team, and feeds back the corresponding clearance costs to the decision-making layer; the clearance team scheduling model takes the minimum sum of the clearance operation cost and the planned power outage loss on the same day as the optimization goal:

[0123] min(C clear-norm +C loss-norm ) (19)

[0124]

[0125]

[0126] In the above formula, N is the number of clearance teams; T is the duration of a day, that is, 24 hours; is the unit time working cost of the clearance team j under normal weather conditions; is the working flag variable of the clearance team j at time t, means that the clearance team j is in a working state at the conflict prone point i at time t, otherwise it is in a rest state or in a working state at other conflict prone points; is the flag variable indicating whether the clearance measure is taken at the conflict prone point i at time t, means that the clearance operation is being carried out at the conflict prone point i at time t, It indicates that the potential trouble point i during the t period is not in the obstacle clearance state, and its relationship with is shown in Equation (40).

[0127] The constraint conditions of the objective function include obstacle clearance resource constraint, power outage loss constraint, obstacle clearance duration constraint, working status constraint of the obstacle clearance team, working duration constraint of the obstacle clearance team, and repair duration constraint of the obstacle clearance team. Among them,

[0128] Obstacle clearance resource constraint:

[0129] The resources such as manpower and material resources consumed for obstacle clearance within the same period cannot exceed the maximum available resource volume.

[0130]

[0131] In the above formula, N max is the maximum number of obstacle clearance teams allowed to work within the same period.

[0132] Power outage loss constraint:

[0133] The planned power outage loss caused by obstacle clearance within the same period cannot exceed the maximum tolerable loss volume.

[0134]

[0135] In the above formula, C max is the maximum allowable power outage loss volume within the same period.

[0136] Obstacle clearance duration constraint:

[0137]

[0138] In the above formula, T i,min is the minimum obstacle clearance duration of the potential trouble point i of the tree-line conflict. This constraint condition is used to ensure that the potential trouble points of the tree-line conflict to be processed can be properly cleared and the potential safety hazards can be completely eliminated.

[0139] Working status constraint of the obstacle clearance team:

[0140]

[0141] This constraint condition indicates that: within the same period, the same obstacle clearance team can handle at most one potential trouble point of the tree-line conflict.

[0142]

[0143] This constraint condition indicates that: when the potential trouble point i of the tree-line conflict undergoes obstacle clearance operations, there is one and only one team performing the obstacle clearance task here.

[0144] Working duration constraint of the obstacle clearance team:

[0145]

[0146] In the above formula, is the working hours of the obstacle removal team j at the tree-line conflict prone point i, is the remaining working hours of the obstacle removal team j at the tree-line conflict prone point i, is the maximum working hours of the obstacle removal team j. This constraint indicates that: the j-th obstacle removal team can work at most hours.

[0147] Obstacle removal team repair time constraint:

[0148] After working for a period of time, each obstacle removal team must rest for at least a certain period of time before continuing to work.

[0149]

[0150] In the above formula, is the rest time of the obstacle removal team j, is the minimum rest time of the obstacle removal team j, f pc (·) is a proportional function.

[0151] Step 8: The decision-making layer combines the obtained obstacle removal cost and the objective function of the pre-obstacle removal decision-making model to calculate the fitness of each particle;

[0152] Step 9: According to the fitness of each particle, update the individual optimal position and the current global optimal position, and then combine the inertia factor and the acceleration constant to update the position and velocity of each particle;

[0153] Step 10: Repeat Steps 7 to 9 until the set maximum number of iterations is reached or the result converges, and output the optimal pre-obstacle removal plan result.

[0154] To verify the effectiveness of the tree obstacle grounding risk warning model in the method of the present invention, in this embodiment, 1115 groups of data enhanced by SMOTE are divided into 900 groups of training samples and 215 groups of detection samples. First, the optimal parameters of each region SVM classifier are solved by means of the training samples, and then the classification effect corresponding to the optimal parameters is solved based on the detection samples. The results are shown in Table 3.

[0155] Table 3 Classification effect of the tree obstacle grounding risk warning model based on SVM for detection samples

[0156] Area code Classification accuracy rate / % Precision rate / % Recall rate / % F1 score A 92.5581％ 85.8696％ 96.3415％ 0.9080 B 91.6279％ 92.7083％ 89.0000％ 0.9082 C 88.8372％ 92.7083％ 83.9623％ 0.8812 D 88.3721％ 90.6250％ 84.4660％ 0.8744 E 89.3023％ 94.1748％ 85.0877％ 0.8940 F 86.9767％ 84.6939％ 86.4583％ 0.8557 G 88.8372％ 87.5000％ 87.5000％ 0.8750 H 86.9767％ 85.4369％ 87.1287％ 0.8627

[0157] As can be seen from Table 3, the tree fault grounding risk warning model based on SVM has a good classification effect on the detection samples. The classification accuracy rates of the 8 regions are all above 86.9767%, and the highest can reach 92.5581%. The precision rate and recall rate are both at a relatively high level, and the F1 score is also above 0.8557, which can accurately realize the tree fault grounding risk warning.

[0158] To further verify the classification effect of the model on actual data, the meteorological monitoring data of a severe convective weather occurrence day in August 2020 was intercepted and input into the SVM classifiers of each region. The output risk warning results were compared with the actual tree fault grounding fault records. The results are shown in Table 4 (where, during the time periods from 1 to 7 and 13 to 24, no tree fault grounding faults occurred in the 8 regions of the township's distribution network, and the output results of the SVM classifiers were all 0 and not shown). As can be seen from the table, for this severe convective weather occurrence day, the classification accuracy rate of the tree fault grounding risk warning model in regions A and C is 95.8333%, and the classification accuracy rates of the models in the other six regions are all 100%. The comprehensive accuracy rate can reach 98.9583%. Therefore, it can be determined that the tree fault grounding risk warning model for the township distribution network proposed in the present invention can effectively excavate the mapping relationship between severe convective weather and the contradiction between trees and lines, and realize the zonal warning of tree fault grounding risk.

[0159] Table 4 Comparison between the model output results and the actual fault records

[0160]

[0161] To verify the effectiveness of the tree-line contradiction optimization model in the method of the present invention, the following three scenarios are set in this embodiment: Scenario 1, without considering the risk warning results, no prior tree clearing measures are taken for all tree-line contradiction prone points; Scenario 2, also without considering the risk warning results, only the n max tree-line contradiction prone points with the largest power outage losses due to faults are selected for prior tree clearing treatment (for two contradiction points with the same power outage losses, the one with the lower minimum tree clearing duration is selected, and the final selection results are ①, ⑤, ⑥, ⑨, ③); Scenario 3, that is, the model proposed in the present invention, combines the fault risks of each contradiction prone point, weighs the planned power outage losses, prior tree clearing costs, power outage losses due to faults and post-treatment costs, and properly decides whether to take prior tree clearing measures. Except for the control conditions, the other parameters in the above three scenarios are kept consistent.

[0162] Based on the risk warning results during a severe convective weather occurrence period in the summer of 2020, a scenario comparison analysis is carried out (the fault probabilities of each contradiction prone point are 0.8, 0.8, 0.4, 0.5, 0.55, 0.3, 0.3, 0.75, 0.2, and 0.6 respectively). The system costs under each scenario are shown in Table 5.

[0163] Table 5 Comparison of System Costs under Different Scenarios

[0164] Item Scenario 1 Scenario 2 Scenario 3 <![CDATA[Planned power outage loss C loss-norm / yuan]]> 0 86100 47760 <![CDATA[Pre - clearance cost C clear-norm / yuan]]> 0 86900 49600 <![CDATA[Fault power outage loss C loss-extr / yuan]]> 97893.4 19551.4 35259.4 <![CDATA[Post-treatment cost C clear-extr / yuan]]> 132750 79000 73750 Total cost C / yuan 230643.4 271551.4 206369.4

[0165] As can be seen from Table 5, in Scenario 1, no pre - obstacle - clearing measures are taken, so there is no need to bear any planned power outage losses and pre - obstacle - clearing costs. However, because it does not deal with the prone - to - contradiction points with high risk of tree - obstacle grounding, after the strong convective weather arrives, it needs to bear more fault - power - outage losses and post - handling costs, so the total cost is relatively high. Compared with Scenario 1, the fault - power - outage losses and post - handling costs in Scenario 2 have decreased significantly. This is because Scenario 2 has carried out pre - obstacle - clearing treatment on the 5 prone - to - contradiction points with the largest fault - power - outage losses, eliminating some grounding risks. Therefore, when the strong convective weather comes, the power - outage losses caused by tree - obstacle grounding are smaller, and the number of fault points to be processed is reduced, and the post - handling costs are lower. However, because it does not consider the risk - warning results and only ranks and clears each prone - to - contradiction point based on the fault - power - outage losses, it needs to bear many unnecessary pre - obstacle - clearing costs and planned power - outage losses, resulting in an increase rather than a decrease in its total cost. In contrast, the total cost of Scenario 3 is the lowest. This is because it can combine the risk - warning results of each prone - to - contradiction point, weigh the planned power - outage losses and pre - obstacle - clearing costs, as well as the fault - power - outage losses and post - handling costs, and then give the pre - obstacle - clearing plan with the best economy after comparison.

[0166] The final decision result output by Scenario 3 is: only conduct pre - obstacle - clearing treatment on the prone - to - contradiction points ①, ②, and ⑧, and do not take any measures for the remaining prone - to - contradiction points. To verify the accuracy of the model output results, the planned power - outage losses, pre - obstacle - clearing costs, fault - power - outage losses, and post - handling costs of each prone - to - contradiction point are listed in Table 6 respectively. Among them, the pre - obstacle - clearing costs will vary due to different clearing teams, so a range is given in the table. As can be seen from Table 6, only the sum of the planned power - outage losses and pre - obstacle - clearing costs of the prone - to - contradiction points ①, ②, and ⑧ is lower than the sum of their fault - power - outage losses and post - handling costs. Therefore, it is the most economical to only take pre - obstacle - clearing measures for these three tree - line prone - to - contradiction points. The calculation results are consistent with the model output results, verifying the accuracy of the model output results.

[0167] From the above analysis, it can be seen that the optimized processing model for tree - line contradictions in rural distribution networks proposed in the present invention can properly weigh the obstacle - clearing costs and power - outage losses of each prone - to - contradiction point and give the pre - obstacle - clearing plan with the best economy. In addition, because the power - outage losses are considered in the cost calculation, the proposed model can also take into account the power supply reliability during the process of optimizing the tree - line contradictions.

[0168] Table 6 Comparison of Cost Situations of Each Prone - to - Contradiction Point and Verification of the Accuracy of Model Output Results

[0169]

[0170]

[0171] In summary, the method for warning and handling the tree-line conflict of the rural distribution network considering severe convective weather proposed by the present invention is effective and reasonable.

[0172] An embodiment of the present invention provides a system for warning and handling the tree-line conflict of the rural distribution network, including: a computer-readable storage medium and a processor;

[0173] The computer-readable storage medium is used to store executable instructions;

[0174] The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method described in any of the above embodiments.

[0175] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for warning and handling the contradiction between power lines and trees in rural distribution networks considering severe convective weather, characterized in that, Including: S1, dividing into R regions according to the administrative division information of R townships; S2, establishing R SVM classification models, respectively preprocessing the historical severe convective weather monitoring information of the R regions to obtain R feature variables, using the R feature variables and the tree fault grounding fault flag of the i-th region under the historical severe convective weather as the training set, and training the i-th SVM classification model; i = 1, 2, …, R; S3, inputting the feature variables obtained by preprocessing the severe convective weather forecast information of each region into the corresponding trained SVM classification models respectively to obtain the prediction results of tree fault grounding faults in each region and the corresponding fault probabilities; S4, using the tree fault grounding risk probability of the region where the tree-line conflict prone point is located as its fault probability; establishing a tree-line conflict optimization processing model including a decision-making layer and a dispatching layer; The decision-making layer includes a pre-clearance decision-making model, and its optimization goal is to minimize the sum of the pre-day clearance cost, the planned power outage loss caused by pre-day clearance, the clearance cost after a fault occurs in the in-day distribution network, and the power outage loss caused by the in-day distribution network fault and clearance; the constraint conditions include the number constraint of clearance operation points; among them, the pre-day is normal weather, and the in-day is severe convective weather; the clearance cost after a fault occurs in the in-day distribution network and the power outage loss caused by the in-day distribution network fault and clearance are obtained according to the fault probabilities of each tree-line conflict prone point; The dispatching layer includes a clearance team dispatching model, and its optimization goal is to minimize the sum of the pre-day clearance cost and the planned power outage loss caused by pre-day clearance; the constraint conditions include clearance resource constraint, power outage loss constraint, clearance duration constraint, clearance team working status constraint, clearance team working duration constraint, and clearance team repair duration constraint; S5, the decision-making layer randomly generates the initial clearance decision results of each tree-line conflict prone point within the scope of the constraint conditions of the pre-clearance decision-making model and sends them to the dispatching layer. The dispatching layer solves the clearance team dispatching model according to the clearance decision results to obtain the clearance team work plan, and feeds back the corresponding clearance cost to the decision-making layer; the decision-making layer combines the feedback clearance cost and the objective function of the pre-clearance decision-making model, randomly generates new clearance decision results, and after exhausting all the clearance decision results of the tree-line conflict prone points, obtains the optimal pre-clearance plan result.

2. The method according to claim 1, characterized in that Aligning the severe convective weather monitoring information of each moment in the R regions according to the time scale, and performing data coding and normalization to obtain R feature variables.

3. The method according to claim 1 or 2, characterized in that, The monitoring information and the forecast information both include temperature, humidity, wind speed, rainfall, wind direction, and wind force level.

4. The method according to claim 1, characterized in that, The cost of fault clearance after a fault occurs in the intraday distribution network The power outage losses caused by the faults and fault clearing in the intraday distribution network Among them, n is the total number of tree-line conflict prone points; P i extr is the failure probability of the tree-line conflict prone point i under severe convective weather; is the obstacle removal cost per unit time under severe convective weather; T i clear-extr is the obstacle removal duration of the tree-line conflict prone point i under severe convective weather; u i is a flag variable indicating whether obstacle removal operations are carried out at the tree-line conflict prone point i, u i = 1 indicates that the pre-emptive obstacle removal measure is taken at the conflict prone point i, u i = 0 indicates that no treatment measures are taken at the conflict prone point i; is the power outage loss per unit time of the tree-line conflict prone point i; T i fail-extr is the power outage duration of the tree-line conflict prone point i under severe convective weather; The aforesaid cost of clearing obstacles before a certain date The planned power outage losses caused by the previous obstacle clearance Where N is the number of obstacle removal teams, and T is the duration of a day, i.e., 24 hours; is the unit time working cost of obstacle removal team j under normal weather conditions; is the working flag variable of obstacle removal team j at time t, indicating that obstacle removal team j is in a working state at the potential conflict point i at time t, otherwise it is in a resting state or in a working state at other potential conflict points; is the flag variable indicating whether obstacle removal measures are taken at the potential conflict point i at time t, indicating that obstacle removal operations are being carried out at the potential conflict point i at time t, while indicating that the potential conflict point i is not in an obstacle removal state at time t.

5. The method according to claim 1, wherein The number of obstacle clearing operation points is restricted as follows: where n max is the upper limit of the number of obstacle clearing operation points in advance; The described obstacle removal resource constraint is as follows: where N max is the maximum number of obstacle removal teams allowed to work during the same period; The power outage loss constraint is as follows: where C max is the maximum allowable power outage loss during the same period; The clearing time limit is where T i,min is the minimum clearing time for the tree-line conflict prone point i; The work status constraints of the obstacle removal team are as follows: The working hours of the obstacle removal team are restricted as follows: Among them, is the working hours of the obstacle removal team j at the tree-line conflict prone point i, is the remaining working hours of the obstacle removal team j at the tree-line conflict prone point i, is the maximum working hours of the obstacle removal team j; this constraint indicates that the jth obstacle removal team can work at most hours; The repair duration constraint of the obstacle removal team is as follows: Among them, is the rest duration of the obstacle removal team j, is the minimum rest duration of the obstacle removal team j, f pc (·) is a proportional function.

6. The method according to claim 1, wherein Using the particle swarm algorithm to solve the pre-clearance decision-making model, step S5 includes: S51, the decision-making layer uses the particle swarm optimization algorithm to initialize the positions and velocities of each particle within the scope of the constraint conditions of the pre-clearance decision-making model; S52, each particle transmits its position parameters to the dispatching layer. The dispatching layer solves the clearance team dispatching model according to the transmitted parameters to obtain the clearance team work plans of each clearance team, and feeds back the corresponding clearance cost to the decision-making layer; S53. The decision-making layer calculates the fitness of each particle by combining the obstacle removal cost and the objective function of the pre-obstacle removal decision-making model. S54. According to the fitness of each particle, update the individual optimal position and the current global optimal position, and combine the inertia factor and the acceleration constant to update the position and velocity of each particle. S55. Repeat steps S52 - S54 until the set maximum number of iterations is reached or the result converges, and output the optimal pre-obstacle removal plan result.

7. The method according to claim 6, wherein Use the optimization software gurobi to solve the obstacle removal team scheduling model.

8. A system for warning and handling the contradiction between tree lines in a rural distribution network considering severe convective weather, characterized in that, It includes: A computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the method according to any one of claims 1 - 7.

Images