Urban power distribution network fault prediction and reconstruction method and system in extreme weather

Abstract

The invention provides an urban power distribution network fault prediction and reconstruction method and system under extreme weather, and relates to the field of power distribution network fault prediction and reconstruction. According to the method, a data-driven power distribution network modeling mode is adopted, namely, a numerical weather forecast model is introduced, a large amount of historical extreme meteorological event data is collected, prediction factors are generated, a perfect data set is formed, statistical transformation and regression are adopted to synthesize a minority oversampling technology to balance the data set, and extreme value distribution of target variables is processed; and key multiple meteorological factors causing the power failure event are identified, so that the influence of extreme weather on the power distribution network is accurately predicted, and the problem of experience dependence of a conventional power distribution network fault model is solved. In a post-disaster power distribution network recovery stage, distributed energy is dispatched by dispatching the distributed energy and based on a mixed integer nonlinear programming model with the purpose of minimizing the reconstruction cost of the power distribution network, the continuous power supply capacity of a power distribution system in a power failure period is met, and power distribution network reconstruction is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network fault prediction and reconstruction, and particularly relates to a method and system for urban distribution network fault prediction and reconstruction under extreme weather conditions. Background Art

[0002] Numerical Weather Prediction (NWP) is a method for predicting weather changes based on mathematical models and physical laws. It predicts the future atmospheric motion state and weather phenomena by numerically calculating the actual situation of the atmosphere and solving the equations of fluid mechanics and thermodynamics. Common NWP models: Global Forecast System (GFS): Developed by the National Weather Service of the United States, based on the Navier-Stokes equations of fluid mechanics, it provides weather forecasts on a global scale. European Centre for Medium-Range Weather Forecasts model (ECMWF): Adopts the four-dimensional variational assimilation method, combines the atmospheric dynamics equations, and simulates the atmospheric and surface physical processes. Weather Research and Forecasting model (WRF): Provides high-resolution local weather forecasts and is commonly used for weather prediction in cities, mountains, etc.

[0003] With the frequent occurrence of extreme weather events, the safety and reliability of urban distribution networks are facing increasing challenges. Traditional distribution network systems are usually designed under normal weather conditions, which makes them often unable to effectively respond to sudden faults when encountering extreme weather, resulting in large-scale power outages, equipment damage, and power supply interruptions, bringing huge impacts to social economy and people's livelihood. Against this background, the methods for distribution network fault prediction and reconstruction under extreme weather have become an important research topic. Although the existing technologies provide some solutions to this problem, there are still many deficiencies.

[0004] In terms of fault prediction, although the existing technologies use various means such as statistical regression models, traditional machine learning models, and deep learning models for fault prediction, these methods generally have the problem of insufficient prediction accuracy. Extreme weather events usually have high uncertainty and complexity, and the occurrence frequency of extreme events is relatively low, which makes the existing machine learning models lack sufficient sample data for effective learning during the training process, resulting in insufficient generalization ability of the models. In addition, most models only focus on the direct relationship between weather factors and faults, ignoring complex factors such as the structural changes of the distribution network, load fluctuations, and system responses. For example, under extreme weather conditions, the topological structure of the distribution network may be adjusted according to the real-time load conditions, and the existing fault prediction methods fail to consider this dynamic change, resulting in certain errors in the prediction of future extreme events.

[0005] In terms of distribution network reconstruction, although existing research has proposed optimization algorithms to improve the distribution network reconstruction strategy under extreme weather, these methods still have many limitations. First of all, most of the existing reconstruction methods assume that the structure of the distribution network is fixed when facing faults, lacking consideration of sudden factors such as equipment damage. This leads to the inability of the reconstruction algorithm to effectively cope with the system vulnerability caused by load transfer and equipment damage during extreme weather events, thus reducing the actual effect of the reconstruction plan. Secondly, many traditional reconstruction methods are only based on local optimization and ignore the global optimization strategy. When facing large-scale and multi-point faults, local optimization is often difficult to provide the optimal recovery plan, affecting the rapid recovery ability of the system.

[0006] In addition, although some studies have proposed reconstruction schemes combining distributed generation and microgrids, attempting to improve the disaster resistance of the distribution network through distributed generation, most of these schemes focus on the optimization of a single system and lack in-depth research on multi-source cooperation. In practical applications, the synergistic effect of microgrids and distributed generation plays a crucial role in the distribution network reconstruction under extreme weather. However, existing technologies have not been able to fully explore the synergistic potential of these technologies. Especially under extreme weather conditions, the distribution network faces a complex situation of multi-point faults and local island operation. Existing reconstruction methods have not fully considered the dynamic cooperation between microgrids, resulting in insufficient synergistic ability of distributed generation and microgrids in some cases. Summary of the Invention

[0007] Aiming at the deficiencies in the existing technology, the present invention provides a method and system for fault prediction and reconstruction of urban distribution networks under extreme weather, which solves the technical problems of low fault prediction accuracy caused by extreme weather in the existing model and insufficient global optimization in the traditional reconstruction method.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] A method for fault prediction and reconstruction of urban distribution networks under extreme weather, comprising the following steps:

[0010] S1. Screen out extreme weather events with damage reaching a preset percentile within a set time period through evaluating power outage records, and generate meteorological prediction factors from the NWP model and supplementary environmental datasets; select the meteorological prediction factors that conform to the current extreme weather as target meteorological prediction factors;

[0011] S2. Perform statistical transformation on the target meteorological prediction factors, use the regression synthetic minority over-sampling technique to synthesize minority-class meteorological prediction factors, and form a dataset with the statistically transformed target meteorological prediction factors;

[0012] S3. Predict the fault probability of each region based on the dataset;

[0013] S4. Obtain the distribution network topology data, distribution network node and line information, and distributed energy data, and determine the actual fault location in the current distribution network based on the fault probability of each region;

[0014] S5. According to the actual fault location, construct a MINLP model, with the goal of minimizing the distribution network reconstruction cost, and determine the objective function of the distribution network reconstruction;

[0015] S6. Under the conditions of applying power flow constraints, node voltage constraints, and line capacity constraints, solve the objective function of the distribution network reconstruction to obtain the optimal distributed energy scheduling plan, and reconstruct the distribution network based on the optimal distributed energy scheduling plan.

[0016] To optimize the above technical solutions, the specific measures taken also include:

[0017] Further, S1 is specifically as follows:

[0018] S1.1. Taking independent geographical area grid cells as the modeling unit, screening out extreme weather events with damage reaching a preset percentile within a set time period through evaluating power outage records, and confirming the start and end times of extreme weather for wind speed, temperature, humidity, and air pressure time series. Organize the data for modeling according to the availability and distribution of regional power outage data;

[0019] S1.2. Use the NWP model to perform weather simulation within the set time period of extreme weather, and extract features including temperature, wind speed, humidity, and precipitation;

[0020] S1.3. Use the digital elevation model and land cover database to obtain the terrain and altitude features around the distribution network facilities, and integrate them with the soil data in the Chinese soil survey database and the soil moisture information simulated by the Weather Research and Forecasting (WRF) model. Incorporate the standardized precipitation index and the standardized precipitation-evapotranspiration index, and form the target meteorological prediction factors together with the features obtained in S1.2.

[0021] Further, S2 is specifically as follows:

[0022] S2.1. Perform natural logarithm transformation on the target meteorological prediction factors, and the formula is as follows:

[0023] V' = ln(V + 1)

[0024] where V is the initial target meteorological prediction factor and V' is the transformed target meteorological prediction factor;

[0025] S2.2. Use the regression synthetic minority over-sampling technique to balance the data set and synthesize minority class meteorological prediction factor samples, specifically:

[0026] Determine the similarity between samples using the distance metric between target meteorological prediction factor samples, and generate minority-class meteorological prediction factor samples based on the similarity between samples; the formula for determining the similarity between samples using the distance metric is as follows:

[0027]

[0028] Among them, D(X,Y) represents the distance between target meteorological prediction factor samples, w x , w y represents the distance weight of different meteorological prediction factor samples, x i , y i represents different meteorological prediction factor samples, s(x i ,y i ) represents the distance metric method of different meteorological prediction factor samples, r represents the weight controlling the distance calculation, and N' represents the number of samples;

[0029] S2.3. The minority-class meteorological prediction factor samples and the target meteorological prediction factors after statistical transformation form a data set.

[0030] Furthermore, S3 is specifically as follows:

[0031] S3.1. Predict the original fault prediction probability for each region based on the data set, and use the deviation function to correct the deviation of the original fault prediction probability to obtain the first fault prediction probability. The deviation function is:

[0032]

[0033] Among them, P represents the original fault prediction probability, P' represents the corrected fault prediction probability, T represents the threshold, and g represents the scaling factor;

[0034] S3.2. Use the differential evolution algorithm to optimize the hyperparameters of the gradient boosting decision tree model, and search for the optimal hyperparameter combination in the parameter space through meteorological prediction factor proxy samples, and adjust the meteorological prediction factors in the data set in an iterative manner. The formula is as follows:

[0035] y t+1 =y t +F×(y p -y q )

[0036] Among them, y t+1 represents the next-generation proxy meteorological prediction factor sample, y t represents the current-generation proxy meteorological prediction factor sample. The proxy meteorological prediction factor sample in the initial iteration comes from the meteorological prediction factors in the data set obtained in S2, y p , y qTwo randomly selected contemporary meteorological prediction factor agents are denoted as, and F represents the scaling factor;

[0037] When the maximum number of iterations is reached or the prediction result reaches the threshold, the iteration terminates, and the optimal meteorological prediction factor sample is obtained. Based on the optimal meteorological prediction factor sample, the second fault prediction probability of each region is predicted;

[0038] S3.3. Select the optimal fault prediction probability from the first fault prediction probability and the second fault prediction probability.

[0039] Furthermore, S5 is specifically as follows:

[0040] Construct a MINLP model with the aim of minimizing the distribution network reconstruction cost to obtain the objective function of the distribution network reconstruction:

[0041]

[0042] Among them, represents the distributed energy cost, represents the node containing distributed energy, represents the cost of strengthening the line between nodes n and j, represents the line between nodes n and j that needs to be strengthened, represents the cost of adding a new line between nodes n and j, represents the line between nodes n and j, du represents the duration of the extreme weather event, and Y represents the number of years of the extreme weather event occurring, represents the distributed energy maintenance cost, represents the maintenance cost of the line between nodes n and j, represents the line state between nodes n and j, K t represents the operating cost of the standby distributed generator at the node, represents the supply power of node n at time t, represents the direct economic loss caused by the unsupplied power, represents the distributed energy supply power of node n at time t, and VOLL represents the economic value of the unsupplied load, represents the unsupplied power of node n at time t, N represents the number of nodes, and ST represents the duration of the extreme weather.

[0043] Furthermore, in S6, the power flow constraint is specifically as follows:

[0044]

[0045] Among them, represents the active power and reactive power of node n at time t, represents the active power and reactive power of the distributed energy at nodes n and j at time t, Indicates the active power and reactive power of the line between nodes n and j at time t. Indicates the active power and reactive power of the load between nodes n and j at time t, A n,j Indicates the line connection matrix, where N represents the number of nodes. Indicates the active power and reactive power supplied by node n at time t.

[0046] The node voltage constraint and line capacity constraint are specifically as follows:

[0047]

[0048] Among them, Indicates the upper limit of line capacity. Indicates the upper limit of node capacity, V n,t Indicates the node voltage, V min Indicates the lower limit of node voltage, V max Indicates the upper limit of node voltage.

[0049] The present invention also proposes an urban distribution network fault prediction and reconstruction system under extreme weather, including:

[0050] A meteorological data acquisition module, which is used to screen out extreme weather events with damage reaching a preset percentile within a set time period by evaluating power outage records, and generate meteorological prediction factors from the NWP model and supplementary environmental datasets; select the meteorological prediction factors that conform to the current extreme weather as the target meteorological prediction factors.

[0051] A data preprocessing module, which is used to perform statistical transformation on the target meteorological prediction factors, use the regression synthetic minority over-sampling technique to synthesize minority-class meteorological prediction factors, and form a dataset with the statistically transformed target meteorological prediction factors.

[0052] A prediction module, which is used to predict the fault probability of each region based on the dataset.

[0053] A distribution network data acquisition module, which is used to obtain the distribution network network topology data, distribution network node and line information, and distributed energy data, and determine the actual fault location in the current distribution network based on the fault probability of each region.

[0054] A reconstruction model building module, which is used to construct a MINLP model according to the actual fault location, with the goal of minimizing the distribution network reconstruction cost, and determine the objective function of the distribution network reconstruction.

[0055] A reconstruction model solving module, which is used to solve the objective function of the distribution network reconstruction under the conditions of applying power flow constraints, node voltage constraints, and line capacity constraints, obtain the optimal distributed energy scheduling scheme, and reconstruct the distribution network based on the optimal distributed energy scheduling scheme.

[0056] The beneficial effects of the present invention are as follows:

[0057] The present invention introduces the deep combination of big data and machine learning technologies into the fault prediction of the distribution network under extreme weather conditions, and makes full use of multi-source heterogeneous data such as meteorological data, distribution network structure information, and historical fault records. By introducing a statistical regression model and an improved machine learning algorithm, the model can deeply explore the complex correlations between weather characteristics, load fluctuations, and distribution network equipment, effectively overcoming the sensitivity of traditional models to scarce data samples and skewed distributions, thereby improving the prediction ability for complex faults caused by extreme weather. Compared with traditional methods, the method of the present invention avoids prediction biases caused by sparse sample distributions by establishing a dynamically adjusted probability model, making the fault prediction results closer to the actual operation scenario.

[0058] The present invention introduces a distributed energy collaborative control mechanism into the distribution network reconstruction strategy. Different from the limitations of traditional distribution network reconstruction methods based only on static structure optimization, the present invention fully considers the load changes and economic efficiency of equipment dynamic response under extreme weather conditions. In specific implementation, through the flexible access of distributed generation units, the method can adjust the reconstruction plan in real time, thereby more reasonably allocate key resources and optimize the reconstruction path.

[0059] The multi-stage optimization model proposed by the present invention combines advanced heuristic algorithms, significantly improving the calculation efficiency and global search ability. Different from the problem that traditional optimization algorithms are prone to fall into local optimal solutions, the present invention realizes faster optimization convergence while considering multi-dimensional optimization objectives by introducing a distribution network fault prediction method based on a numerical weather prediction (NWP) model and a distribution network reconstruction method based on a mixed-integer non-linear programming model (MINLP), providing theoretical support and technical guarantee for the efficient reconstruction of the distribution network under extreme weather conditions. Brief Description of the Drawings

[0060] Figure 1 It is a flowchart of the method for fault prediction and reconstruction of the urban distribution network under extreme weather provided by the embodiment of the present invention.

[0061] Figure 2 It is a module diagram of the system for fault prediction and reconstruction of the urban distribution network under extreme weather provided by the embodiment of the present invention.

[0062] Figure 3 It is a reconstructed diagram of the IEEE 33-node distribution network provided by the embodiment of the present invention. Detailed Embodiment

[0063] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0064] Embodiment 1

[0065] The present invention proposes a method for fault prediction and reconstruction of urban distribution networks under extreme weather conditions. The flowchart of this method is as Figure 1 shown, including the following steps:

[0066] S1. Screen out extreme weather events with damage reaching a preset percentile within a set time period through the evaluation of power outage records, and generate meteorological prediction factors from the NWP model and supplementary environmental datasets; select the meteorological prediction factors that conform to the current extreme weather as the target meteorological prediction factors. Specifically, S1 is as follows:

[0067] S1.1. Taking independent geographical area grid units as the modeling unit, screen out extreme weather events with damage reaching a preset percentile (set at the 95th percentile in this embodiment) within a set time period (set at 48 hours in this embodiment) through the evaluation of power outage records, confirm the start and end times of extreme weather for wind speed, temperature, humidity, and air pressure time series, and organize the data for modeling according to the availability and distribution of regional power outage data.

[0068] S1.2. Use the NWP model to simulate the weather within the set time period of extreme weather (set at 48 hours in this embodiment), and extract features including temperature, wind speed, humidity, and precipitation; such as the wind energy exposed to trees:

[0069] E wind = r × A leaf × V wind 3

[0070] where E wind represents wind energy, ρ represents air density, A leaf represents leaf area, and V wind represents wind speed.

[0071] S1.3. Use the digital elevation model and land cover database to obtain the terrain and elevation features around the distribution network facilities, and integrate them with the soil data in the Chinese soil survey database and the soil moisture information simulated by the Weather Research and Forecasting (WRF) model, incorporate the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI), and form the target meteorological prediction factors together with the features obtained in S1.2.

[0072]

[0073] Among them, P w represents precipitation, ET represents evapotranspiration, mP w represents the mean of precipitation, sP w represents the standard deviation of precipitation.

[0074] S2. Statistically transform the target meteorological prediction factors to reduce the skewness problem of their distribution, ensure that the target data is more operable during model training, use the regression synthetic minority over-sampling technique (SMOTE) to synthesize minority-class meteorological prediction factors, balance the dataset, enhance the representativeness of rare high-impact events in the dataset, and ensure that the model can handle the data skewness problem; form a dataset with the statistically transformed target meteorological prediction factors; S2 is specifically as follows:

[0075] S2.1. Perform a natural logarithm transformation on the target meteorological prediction factors to reduce the long-tail and zero-inflation problems, adjust the data distribution, and obtain a target variable more suitable for modeling. The formula is as follows:

[0076] V' = ln(V + 1)

[0077] Among them, V is the initial target meteorological prediction factor, and V' is the transformed target meteorological prediction factor;

[0078] S2.2. Use the regression synthetic minority over-sampling technique to balance the dataset and synthesize minority-class meteorological prediction factor samples. Specifically:

[0079] Determine the similarity between samples using the distance metric between target meteorological prediction factor samples, and generate minority-class meteorological prediction factor samples based on the similarity between samples; the formula for determining the similarity between samples using the distance metric is expressed as follows:

[0080]

[0081] Among them, D(X, Y) represents the distance between target meteorological prediction factor samples, w x , w y represents the distance weight of different meteorological prediction factor samples, x i , y i represent different meteorological prediction factor samples, s(x i , y i ) represents the distance metric method of different meteorological prediction factor samples, r represents the weight controlling the distance calculation, and N' represents the number of samples;

[0082] S2.3. Combine the minority-class meteorological prediction factor samples with the statistically transformed target meteorological prediction factors to form a dataset.

[0083] S3. Predict the failure probability of each region based on the dataset; specifically, S3 is as follows:

[0084] S3.1. Predict the original failure prediction probability of each region based on the dataset. There are many choices for the prediction method here, such as Bayesian neural network, data analysis method, and support vector machine.

[0085] Use the deviation function to correct the deviation of the original failure prediction probability to obtain the first failure prediction probability. The deviation function is:

[0086]

[0087] where P represents the original failure prediction probability, P' represents the corrected failure prediction probability, T represents the threshold, and g represents the scaling factor;

[0088] S3.2. Use the differential evolution algorithm to optimize the hyperparameters of the gradient boosting decision tree model, and search for the optimal hyperparameter combination in the parameter space through the proxy samples of meteorological prediction factors, and adjust the meteorological prediction factors in the dataset iteratively. The formula is as follows:

[0089] y t+1 = y t + F × (y p - y q )

[0090] where y t+1 represents the next-generation proxy meteorological prediction factor sample, y t represents the current-generation proxy meteorological prediction factor sample. The proxy meteorological prediction factor sample at the initial iteration comes from the meteorological prediction factors in the dataset obtained in S2, y p , y q represent two randomly selected current-generation proxy meteorological prediction factor samples, and F represents the scaling multiple;

[0091] When the maximum number of iterations is reached or the prediction result reaches the threshold, the iteration terminates, and the optimal meteorological prediction factor sample is obtained. Predict the second failure prediction probability of each region based on the optimal meteorological prediction factor sample. There are many choices for the prediction method here, such as Bayesian neural network, data analysis method, and support vector machine.

[0092] S3.3. Select the optimal failure prediction probability from the first failure prediction probability and the second failure prediction probability.

[0093] S4. Obtain the distribution network topology data, distribution network node and line information, and distributed energy data, and determine the actual failure location in the current distribution network based on the failure probability of each region;

[0094] S5. Construct a MINLP (Mixed Integer Nonlinear Programming) model based on the actual fault location, and determine the objective function for the reconfiguration of the distribution network with the goal of minimizing the reconfiguration cost of the distribution network. Specifically, S5 is as follows:

[0095] Construct a MINLP (Mixed Integer Nonlinear Programming) model, aiming to minimize the reconfiguration cost of the distribution network, and obtain the objective function for the reconfiguration of the distribution network:

[0096]

[0097] Among them, represents the cost of distributed energy, represents the node containing distributed energy, represents the cost of strengthening the line between nodes n and j, represents the line between nodes n and j that needs to be strengthened, represents the cost of adding a new line between nodes n and j, represents the line between nodes n and j, du represents the duration of the extreme weather event, and Y represents the number of years of the extreme weather event occurrence, represents the maintenance cost of distributed energy, represents the maintenance cost of the line between nodes n and j, represents the existing state of the line between nodes n and j, K t represents the operating cost of the standby distributed generator at node, represents the supply power of node n at time t, represents the direct economic loss caused by the unsupplied power, represents the distributed energy supply power of node n at time t, and VOLL represents the economic value of the unsupplied load, represents the unsupplied power of node n at time t, N represents the number of nodes, and ST represents the duration of the extreme weather.

[0098] S6. Under the conditions of applying power flow constraints, node voltage constraints, and line capacity constraints, solve the objective function for the reconfiguration of the distribution network to obtain the optimal scheduling plan for distributed energy, and reconfigure the distribution network based on the optimal scheduling plan for distributed energy.

[0099] The power flow constraint is specifically:

[0100]

[0101] Among them, represents the active power and reactive power of node n at time t, represents the active power and reactive power of distributed energy at nodes n and j at time t, represents the active power and reactive power of the line between nodes n and j at time t, Denote the active and reactive power of the load between nodes n and j at time t, A n,j Denote the line connection matrix, N denotes the number of nodes, Denote the active and reactive power supplied by node n at time t.

[0102] The node voltage constraint and line capacity constraint are specifically:

[0103]

[0104] Among them, Denote the upper limit of line capacity, Denote the upper limit of node capacity, V n,t Denote the node voltage, V min Denote the lower limit of node voltage, V max Denote the upper limit of node voltage.

[0105] Example Two

[0106] The present invention proposes an urban distribution network fault prediction and reconstruction system under extreme weather corresponding to the method of Example One. The module diagram of the system is as Figure 2 shown, including:

[0107] A meteorological data acquisition module, which is used to screen out extreme weather events with damage reaching a preset percentile within a set time period by evaluating power outage records, and generate meteorological prediction factors from the NWP model and supplementary environmental datasets; select the meteorological prediction factors that meet the current extreme weather as the target meteorological prediction factors;

[0108] A data preprocessing module, which is used to perform statistical transformation on the target meteorological prediction factors, use the regression synthetic minority over-sampling technique to synthesize minority-class meteorological prediction factors, and form a dataset with the statistically transformed target meteorological prediction factors;

[0109] A prediction module, which is used to predict the fault probability of each region based on the dataset;

[0110] A distribution network data acquisition module, which is used to obtain distribution network network topology data, distribution network node and line information, and distributed energy data, and determine the actual fault location in the current distribution network based on the fault probability of each region;

[0111] A reconstruction model building module, which is used to construct a MINLP model according to the actual fault location, with the goal of minimizing the distribution network reconstruction cost, and determine the objective function of the distribution network reconstruction;

[0112] The reconstruction model solving module is used to solve the objective function of the distribution network reconstruction under the conditions of applying power flow constraints, node voltage constraints, and line capacity constraints, obtain the optimal scheduling scheme for distributed energy, and reconstruct the distribution network based on the optimal scheduling scheme for distributed energy.

[0113] The implementation manners of each module and module functions in the system are completely consistent with the steps of the method in the first embodiment, so details are not described herein again.

[0114] The present invention can effectively reconstruct the IEEE33 distribution network after suffering from extreme weather disasters by constructing a MINLP model for it, solving the optimal reconstruction parameters. The reconstructed IEEE33-node distribution network is as Figure 3 shown.

[0115] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this application can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0116] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A method for predicting and reconstructing faults in urban distribution networks under extreme weather conditions, characterized in that, Including the following steps: S1. Screen out extreme weather events with damage reaching a preset percentile within a set time period through evaluating power outage records, and generate meteorological prediction factors from the NWP model and supplementary environmental datasets; Select the meteorological prediction factors that conform to the current extreme weather as the target meteorological prediction factors; S2. Conduct statistical transformation on the target meteorological prediction factors, use the regression synthetic minority over-sampling technique to synthesize minority-class meteorological prediction factors, and form a dataset with the statistically transformed target meteorological prediction factors; S3. Predict the failure probability of each region based on the dataset; S4. Obtain the distribution network topology data, distribution network node and line information, and distributed energy data, and determine the actual failure locations in the current distribution network based on the failure probability of each region; S5. According to the actual failure locations, construct a MINLP model, with the goal of minimizing the distribution network reconstruction cost, and determine the objective function of the distribution network reconstruction; S6. Under the conditions of applying power flow constraints, node voltage constraints, and line capacity constraints, solve the objective function of the distribution network reconstruction to obtain the optimal distributed energy scheduling plan, and reconstruct the distribution network based on the optimal distributed energy scheduling plan.

2. The method for predicting and reconstructing faults in urban distribution networks under extreme weather conditions according to claim 1, wherein Specifically, S1 is as follows: S1.

1. Taking independent geographical region grid cells as the modeling unit, screen out extreme weather events with damage reaching a preset percentile within a set time period through evaluating power outage records, confirm the start and end times of the extreme weather for wind speed, temperature, humidity, and air pressure time series, and organize the data for modeling according to the availability and distribution of regional power outage data; S1.

2. Use the NWP model to simulate the weather within the set time period of the extreme weather, and extract features including temperature, wind speed, humidity, and precipitation; S1.

3. Use the digital elevation model and land cover database to obtain the terrain and elevation features around the distribution network facilities, integrate them with the soil data in the Chinese soil survey database and the soil moisture information simulated by the Weather Research and Forecasting (WRF) model, incorporate the standardized precipitation index and the standardized precipitation-evapotranspiration index, and form the target meteorological prediction factors with the features obtained in S1.

2.

3. The method for predicting and reconstructing faults in urban distribution networks under extreme weather according to claim 1, wherein, Specifically, S2 is as follows: S2.

1. Conduct natural logarithm transformation on the target meteorological prediction factors, and the formula is as follows: V￠=ln(V+1) where V is the initial target meteorological prediction factor, and V' is the transformed target meteorological prediction factor; S2.

2. Use the regression synthetic minority over-sampling technique to balance the dataset and synthesize minority-class meteorological prediction factor samples. Specifically: Determine the similarity between samples using the distance metric between target meteorological prediction factor samples, and generate minority-class meteorological prediction factor samples based on the similarity between samples. The formula for determining the similarity between samples using the distance metric is as follows: Among them, D(X,Y) represents the distance between target meteorological prediction factor samples, w x , w y represents the distance weight of different meteorological prediction factor samples, x i , y i represent different meteorological prediction factor samples, s(x i ,y i ) represents the distance measurement method of different meteorological prediction factor samples, r represents the weight for controlling distance calculation, and N' represents the number of samples; S2.

3. The minority-class meteorological prediction factor samples and the statistically transformed target meteorological prediction factors form a dataset.

4. The method for predicting and reconstructing urban distribution network faults under extreme weather according to claim 1, wherein Specifically, S3 is as follows: S3.

1. Predict the original failure prediction probability of each region based on the dataset, and use the deviation function to correct the deviation of the original failure prediction probability to obtain the first failure prediction probability. The deviation function is: Among them, P represents the original fault prediction probability, P' represents the corrected fault prediction probability, T represents the threshold, and g represents the scaling factor; S3.

2. Optimize the hyperparameters of the gradient boosting decision tree model using the differential evolution algorithm, and search for the optimal hyperparameter combination in the parameter space through the proxy samples of meteorological prediction factors, and adjust the meteorological prediction factors in the dataset in an iterative manner. The formula is as follows: y t+1 = y t + F × (y p - y q ) Among them, y t+1 represents the next-generation proxy meteorological prediction factor sample, and y t represents the current-generation proxy meteorological prediction factor sample. The proxy meteorological prediction factor sample at the initial iteration comes from the meteorological prediction factors in the dataset obtained from S2. y p , y q represents two randomly selected current-generation meteorological prediction factor proxy samples, and F represents the scaling factor; When the maximum number of iterations is reached or the prediction result reaches the threshold, the iteration terminates, and the optimal meteorological prediction factor sample is obtained. Based on the optimal meteorological prediction factor sample, the second fault prediction probability of each region is predicted; S3.

3. Select the optimal fault prediction probability from the first fault prediction probability and the second fault prediction probability.

5. The method for predicting and reconstructing faults in an urban distribution network under extreme weather conditions according to claim 1, wherein, S5 is specifically as follows: Construct a MINLP model with the aim of minimizing the distribution network reconstruction cost to obtain the objective function of the distribution network reconstruction: Among them, represents the cost of distributed energy, represents containing distributed energy nodes, represents the cost of strengthening the line between nodes n and j, represents the line between nodes n and j that needs to be strengthened, represents the cost of adding a new line between nodes n and j, represents the line between nodes n and j, du represents the duration of extreme weather events, and Y represents the number of years of extreme weather events, represents the maintenance cost of distributed energy, represents the maintenance cost of the line between nodes n and j, represents the existing state of the line between nodes n and j, K t represents the operating cost of the standby distributed generator at the node, represents the supply power of node n at time t, represents the direct economic loss caused by the unsupplied power, represents the distributed energy supply power of node n at time t, and VOLL represents the economic value of the unsupplied load, represents the unsupplied power of node n at time t, N represents the number of nodes, and ST represents the duration of extreme weather.

6. The method for predicting and reconstructing faults in urban distribution networks under extreme weather according to claim 1, wherein In S6, the power flow constraint is specifically: Among them, represents the active power and reactive power of node n at time t, represents the active power and reactive power of distributed energy of nodes n and j at time t, represents the active power and reactive power of the line between nodes n and j at time t, represents the active power and reactive power of the load between nodes n and j at time t, A n,j represents the line connection matrix, and N represents the number of nodes, represents the active power and reactive power supplied by node n at time t; The node voltage constraint and the line capacity constraint are specifically: Among them, represents the upper limit of line capacity, represents the upper limit of node capacity, V n,t represents the node voltage, V min represents the lower limit of node voltage, V max represents the upper limit of node voltage.

7. An urban distribution network fault prediction and reconstruction system under extreme weather conditions, characterized in that Including: A meteorological data acquisition module, which is used to screen out extreme weather events whose damage reaches the preset percentile within a set time period through evaluating power outage records, and generate meteorological prediction factors from the NWP model and the supplementary environmental dataset; Select the meteorological prediction factors that conform to the current extreme weather as the target meteorological prediction factors; A data preprocessing module, which is used to perform statistical transformation on the target meteorological prediction factors, use the regression synthetic minority over-sampling technique to synthesize minority class meteorological prediction factors, and form a dataset with the statistically transformed target meteorological prediction factors; A prediction module, which is used to predict the fault probability of each region based on the dataset; A distribution network data acquisition module, which is used to acquire the distribution network network topology data, the distribution network node and line information, and the distributed energy data, and determine the actual fault location in the current distribution network based on the fault probability of each region; A reconstruction model building module, which is used to construct a MINLP model according to the actual fault location, with the goal of minimizing the distribution network reconstruction cost, and determine the objective function of the distribution network reconstruction; A reconstruction model solving module, which is used to solve the objective function of the distribution network reconstruction under the conditions of applying the power flow constraint, the node voltage constraint, and the line capacity constraint, obtain the optimal distributed energy scheduling scheme, and reconstruct the distribution network based on the optimal distributed energy scheduling scheme.

Images