Differentiated disaster prevention decision-making method for power distribution network

By building multiple disaster prediction models and dynamic Bayesian network models, accurately predicting and making decisions for different natural disasters, the problems of insufficient data utilization and weak dynamic response capabilities of traditional distribution network disaster prevention decision-making methods are solved, and disasters are improved Accuracy of response and disaster prevention capabilities of the distribution network.

CN118966411APending Publication Date: 2024-11-15STATE GRID FUJIAN POWER ELECTRIC CO ECONOMIC RESEARCH INSTITUTE +1

Patent Information

Application Number
CN202410975293.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Traditional distribution network disaster prevention decision-making methods lack systematic disaster prediction and emergency response mechanisms. They rely on a single data source, have weak dynamic response capabilities and low decision-making accuracy, and cannot effectively reduce the impact of natural disasters on the distribution network.

Method used

By collecting and preprocessing historical disaster data, various disaster prediction models are constructed, such as typhoons, floods, ice and snow, natural fires and earthquake prediction models, combining space-time attention fusion network, long-term and short-term memory network, meta-learning model and multimodal attention Force network, predicts the probability and impact of disasters, and constructs a dynamic Bayesian network model to make disaster decisions.

Benefits of technology

Accurate prediction and response strategies for different natural disasters have been achieved, the accuracy and response capabilities of the distribution network in disaster prevention decisions have been improved, and the failure rate has been effectively reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966411A_ABST
    Figure CN118966411A_ABST
Patent Text Reader

Abstract

The invention relates to a power distribution network differential disaster prevention decision-making method, which comprises the following steps: collecting various historical disaster data in a power distribution network region, preprocessing the various historical disaster data, and respectively constructing corresponding training sets; respectively constructing various disaster prediction models, inputting the corresponding training sets into the disaster prediction models of the corresponding types for training to obtain various trained disaster prediction models, and performing disaster prediction through the various trained disaster prediction models; and meanwhile, constructing a power distribution network disaster prevention decision model, inputting disaster prediction results output by various disaster prediction models into the power distribution network disaster prevention decision model, and outputting a disaster prevention decision of the power distribution network by the power distribution network disaster prevention decision model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a distribution network differentiated disaster prevention decision-making method, and belongs to the technical field of distribution network planning. Background Art

[0002] As an important part of the power system, the distribution network undertakes the key task of power transmission and distribution. Typical natural disasters include earthquakes, typhoons, floods, ice and snow disasters, etc. These disasters can cause a series of problems such as the collapse of distribution lines, damage to substations, and power outages, affecting social production and life.

[0003] Traditional distribution network disaster prevention decisions mainly rely on historical experience and expert judgment, and often lack a systematic disaster prediction and emergency response mechanism. Specific limitations include: Insufficient data utilization: Traditional methods usually rely only on a single data source and cannot fully integrate data resources from meteorology, geology, distribution network operation status, etc., resulting in incomplete decision-making basis. Weak dynamic response capabilities: Traditional decision-making plans lack real-time monitoring and response capabilities for the evolution of disasters, and often take measures passively after the disaster occurs, missing the best time to respond. Low decision-making accuracy: Due to the lack of accurate prediction of the impact of disasters, traditional decision-making methods perform poorly when dealing with disasters of different types and intensities, and cannot effectively reduce the impact of disasters on distribution networks.

[0004] The invention patent with the announcement number "CN109558990B" discloses a distribution network disaster prevention backbone grid planning method based on the Steiner tree model. This method first considers the power supply to important users of the distribution network under typical disaster conditions, takes the construction cost of lines and towers and the minimum risk of disasters as the optimization goal, and establishes the Steiner tree model of the distribution network disaster prevention backbone grid based on network connectivity constraints and flow constraints; uses the binary particle swarm algorithm to solve the optimization model, obtains the optimal backbone grid solution, and realizes the power supply of the power grid to important users under disasters. The invention takes into account the comprehensive cost of differentiated transformation of the backbone grid, ensures the power supply of the power grid to important loads during disasters, and improves the power grid's ability to provide power supply under disasters.

[0005] However, the above scheme only proposes a planning method for the disaster prevention backbone grid of the distribution network, which is of no help for the disaster prevention decision-making of the distribution network that has been completed. At the same time, the planning method of the disaster prevention backbone grid does not plan according to the characteristics of each natural disaster, and the planned line may not necessarily be the optimal line. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention proposes a distribution network differentiated disaster prevention decision-making method.

[0007] The technical solution of the present invention is as follows:

[0008] On the one hand, the present invention provides a distribution network differentiated disaster prevention decision-making method, comprising the following steps:

[0009] Collect various historical disaster data in the distribution network area, and construct corresponding training sets after preprocessing various historical disaster data;

[0010] Construct various disaster prediction models respectively, input the corresponding training sets into the corresponding types of disaster prediction models for training, obtain the trained disaster prediction models, and perform disaster prediction through the trained disaster prediction models;

[0011] At the same time, a distribution network disaster prevention decision-making model is constructed, and the disaster prediction results output by various disaster prediction models are input into the distribution network disaster prevention decision-making model, and the distribution network disaster prevention decision-making model outputs the distribution network disaster prevention decision.

[0012] As a preferred embodiment of the present invention, the various disaster data include historical typhoon, flood, ice and snow, natural fire and earthquake data;

[0013] The various disaster prediction models include typhoon prediction models, flood prediction models, ice and snow prediction models, natural fire prediction models and earthquake prediction models.

[0014] As a preferred embodiment of the present invention, the typhoon prediction model is constructed based on a spatiotemporal attention fusion network model;

[0015] The spatiotemporal attention fusion network includes an input layer, a spatial attention layer, a spatiotemporal fusion layer and an output layer, wherein the spatial attention layer is used to identify important areas of the input feature data samples, and the spatiotemporal fusion layer is used to fuse the output data of the spatial attention layer with multi-dimensional data to obtain the predicted path of the typhoon;

[0016] During the training of the typhoon prediction model, the model error C is calculated by the following formula: r :

[0017]

[0018] Where: C represents the absolute error of the typhoon center; R represents the typhoon radius; α p , β p Respectively represent the longitude and latitude of the typhoon forecast value; α r , β r They represent the longitude and latitude of the actual typhoon value respectively; kh represents the time difference between the typhoon forecast value and the actual typhoon value;

[0019] When the model error is less than the preset threshold, the trained typhoon prediction model is obtained.

[0020] As a preferred embodiment of the present invention, the flood prediction model is constructed based on a long short-term memory network, as shown in the following formula:

[0021] P flood (t) = σ(W f ·[h t-1 x t ]+b f )

[0022] Where: P flood (t) represents the probability of flood occurrence in the next t days; σ represents the activation function; W f represents the weight matrix; h t-1 represents the hidden state at time step t-1; x t represents the input feature vector at time step t; b f Represents the bias term.

[0023] As a preferred embodiment of the present invention, the ice and snow prediction model is constructed based on a meta-learning model;

[0024] Input historical ice and snow data into the meta-learning model for training, and use the trained meta-learning model to predict future temperatures and snowfall;

[0025] Construct an energy balance model to calculate the net energy change E of snow cover s , as shown in the following formula:

[0026] E s =I s -Q s -H s -E q

[0027] Where: I s represents solar radiation; Q s Represents reflected radiation; H s Indicates sensible heat exchange; E q represents latent heat exchange;

[0028] The ice and snow disaster scoring model S is constructed by integrating multiple factors, as shown in the following formula:

[0029]

[0030] Among them: α1, α2, α3 are weight coefficients respectively; Represents meteorological disaster factor; G(E s ) represents the energy balance factor; H(T s , W) represents the surface condition factor;

[0031] Finally, by judging whether S is greater than the preset threshold, it is determined whether an ice and snow disaster has occurred.

[0032] As a preferred embodiment of the present invention, the natural fire prediction model is constructed based on a multi-scale convolutional neural network model, including an input layer, a multi-scale convolutional layer, a feature fusion layer, a fully connected layer and an output layer, as shown in the following formula:

[0033] P fire =σ(W o concat(f1, f2, ..., f s )+b o )

[0034] Where: P fire represents the probability of fire occurrence; σ represents the activation function; W o represents the output layer weight; concat(f1, f2, ..., f s ) represents f extracted by multi-scale convolutional layer s The features are connected through the feature fusion layer; b o Represents the output layer bias.

[0035] As a preferred embodiment of the present invention, the earthquake prediction model is constructed based on a multimodal attention network, including an input layer, a multimodal feature extraction layer, an attention mechanism layer, a feature fusion layer and an output layer;

[0036] The multimodal feature extraction layer is constructed based on a convolutional neural network and a long short-term memory network to perform independent feature extraction on each modality;

[0037] The attention mechanism layer is used to perform weighted aggregation between features of each modality;

[0038] The feature fusion layer fuses the output of the attention mechanism layer to form a unified feature representation;

[0039] The final model outputs the probability of earthquake occurrence P through the output layer earthquake , as shown in the following formula:

[0040] P earthquake =σ(W k ·f att +b k )

[0041] Where: σ represents the activation function; W k represents the weight of the output layer; f att represents the modal features after weighted aggregation; b k Represents the bias of the output layer.

[0042] As a preferred embodiment of the present invention, the distribution network disaster prevention decision model is constructed based on a dynamic Bayesian network model;

[0043] Perform multimodal time series preprocessing on the input disaster prediction data;

[0044] A dynamic Bayesian network is constructed based on the preprocessed disaster prediction data. The dependency relationship between natural disasters and the distribution network status is calculated through the dynamic Bayesian network, and the disaster response strategy is dynamically inferred based on the dependency relationship.

[0045] Optimize the distribution network operation strategy based on the inference results.

[0046] On the other hand, the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method described in any embodiment of the present invention is implemented.

[0047] In yet another aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0048] The present invention has the following beneficial effects:

[0049] 1. The present invention proposes different prediction models for different natural disasters, selects appropriate algorithms for prediction according to the characteristics of each natural disaster, and makes the prediction results more accurate;

[0050] 2. The present invention proposes a distribution network disaster prevention decision-making model, and based on the prediction results of various natural disasters, proposes corresponding disaster response strategies, which effectively reduces the failure rate of the distribution network in natural disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0054] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0055] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0056] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0057] Embodiment 1:

[0058] See also Figure 1 , a distribution network differentiated disaster prevention decision-making method, comprising the following steps:

[0059] Collect various historical disaster data in the distribution network area, and construct corresponding training sets after preprocessing various historical disaster data;

[0060] Construct various disaster prediction models respectively, input the corresponding training sets into the corresponding types of disaster prediction models for training, obtain the trained disaster prediction models, and perform disaster prediction through the trained disaster prediction models;

[0061] At the same time, a distribution network disaster prevention decision-making model is constructed, and the disaster prediction results output by various disaster prediction models are input into the distribution network disaster prevention decision-making model, and the distribution network disaster prevention decision-making model outputs the distribution network disaster prevention decision.

[0062] As a preferred implementation of this embodiment, the various disaster data include historical typhoon, flood, ice and snow, natural fire and earthquake data;

[0063] The various disaster prediction models include typhoon prediction models, flood prediction models, ice and snow prediction models, natural fire prediction models and earthquake prediction models.

[0064] As a preferred implementation of this embodiment, the typhoon prediction model is constructed based on a spatiotemporal attention fusion network model;

[0065] The spatiotemporal attention fusion network includes an input layer, a spatial attention layer, a spatiotemporal fusion layer and an output layer, wherein the spatial attention layer is used to identify important areas of the input feature data samples, and the spatiotemporal fusion layer is used to fuse the output data of the spatial attention layer with multi-dimensional data to obtain the predicted path of the typhoon;

[0066] In this embodiment, the training set of the input model is composed of typhoon historical meteorological images and historical typhoon data, wherein the historical typhoon data includes two-dimensional and three-dimensional data, the two-dimensional data includes the intensity, position, air pressure and wind speed of the typhoon, and the three-dimensional data includes the wind field information, vertical velocity and vorticity of different height layers of the typhoon;

[0067] In this embodiment, the spatial attention layer learns from the historical meteorological images of typhoons and marks the area near the eye of the typhoon in the image as an important area. The spatiotemporal fusion layer combines the output data of the spatial attention layer with the two-dimensional and three-dimensional data of the historical typhoon data to predict the specific location of the typhoon in the future and the change of the typhoon intensity.

[0068] During the training of the typhoon prediction model, the model error C is calculated by the following formula: r :

[0069]

[0070] Among them: C represents the absolute error of the typhoon center, C is a hyperparameter, which is set manually; R represents the typhoon radius; α p , β p Respectively represent the longitude and latitude of the typhoon forecast value; α r , β r They represent the longitude and latitude of the actual typhoon value respectively; kh represents the time difference between the typhoon forecast value and the actual typhoon value;

[0071] When the model error is less than the preset threshold, the trained typhoon prediction model is obtained.

[0072] As a preferred implementation of this embodiment, the flood prediction model is constructed based on a long short-term memory network, as shown in the following formula:

[0073] P flood (t) = σ(W f ·[h t-1 x t ]+b f )

[0074] Where: P flood (t) represents the probability of flood occurrence in the next t days; σ represents the activation function; W f represents the weight matrix; h t-1 represents the hidden state at time step t-1; x t represents the input feature vector at time step t; b f represents the bias term;

[0075] In this embodiment, the training set of the input model includes historical rainfall, rainfall intensity, river water level, basin area, soil moisture, flood flow, peak flow, runoff and flood frequency, wherein the calculation formula of the peak flow is as follows:

[0076]

[0077] Among them: F top represents the peak flow; G represents the runoff coefficient, which reflects the flow-generating capacity of the basin; O represents the rainfall intensity; A represents the basin area;

[0078] The calculation formula of the runoff is shown as follows:

[0079]

[0080] Where: Q represents runoff; p represents rainfall; s represents the maximum potential rainfall retention;

[0081] The flood frequency calculation formula is as follows:

[0082]

[0083] Among them: x represents flood flow; μ represents location parameter; δ represents scale parameter;

[0084] The above data are constructed as a training set and input into the flood prediction model for training. The trained flood prediction model outputs the probability of flood occurrence in the next few days.

[0085] As a preferred implementation of this embodiment, the ice and snow prediction model is constructed based on a meta-learning model;

[0086] Input historical ice and snow data into the meta-learning model for training, and use the trained meta-learning model to predict future temperatures and snowfall;

[0087] The meta-learning model includes two steps: meta-learning and meta-updating;

[0088] The goal of the meta-training is to find an initial model parameter that performs well on multiple tasks;

[0089] The goal of the meta-update is to enable the model to quickly adapt to new tasks;

[0090] In this embodiment, the training set input to the meta-learning model includes historical air temperature, historical snowfall, historical wind speed, and historical surface temperature;

[0091] Construct an energy balance model to calculate the net energy change E of snow cover s , as shown in the following formula:

[0092] Es =I s -Q s -H s -E q

[0093] Where: I s represents solar radiation; Q s Represents reflected radiation; H s Indicates sensible heat exchange; E q represents latent heat exchange;

[0094] The solar radiation I s The calculation formula is:

[0095] I s =S0·cos(α)·(1-a)

[0096] Where: S0 represents the solar constant, specifically 1361W / m 2 ; θ represents the solar altitude angle; a represents the surface albedo, which is usually 0.7 to 0.9;

[0097] The reflected radiation Q s The calculation formula is:

[0098] Q s =a·I s

[0099] The heat exchange H s The calculation formula is:

[0100] H s =C h ·W·(T p -T s )

[0101] Where: C h represents the sensible heat exchange coefficient; W represents the wind speed;

[0102] The latent heat exchange E q The calculation formula is:

[0103] E q =L v ·E

[0104] Where: L v represents the latent heat of vaporization of water; E represents the evaporation amount, which can be set through experience;

[0105] The ice and snow disaster scoring model S is constructed by integrating multiple factors, as shown in the following formula:

[0106]

[0107] Among them: α1, α2, and α3 are weight coefficients, which are determined by regression of historical ice and snow data; represents the meteorological disaster factor, is the predicted temperature, The predicted snowfall is combined with the predicted temperature and snowfall to calculate the risk of ice and snow disasters; G(E s ) represents the energy balance factor, which evaluates the energy impact of snow melting or accumulation; H(T s , W) represents the surface condition factor, which takes into account the effects of surface temperature and wind speed on snow and ice;

[0108] Finally, by judging whether S is greater than the preset threshold, it is determined whether an ice and snow disaster has occurred.

[0109] As a preferred implementation of this embodiment, the natural fire prediction model is constructed based on a multi-scale convolutional neural network model, including an input layer, a multi-scale convolutional layer, a feature fusion layer, a fully connected layer and an output layer, as shown in the following formula:

[0110] P fire =σ(W o concat(f1, f2, ..., f s )+b o )

[0111] Where: P fire represents the probability of fire occurrence; σ represents the activation function; W o represents the output layer weight; concat(f1, f2, ..., f s ) represents f extracted by multi-scale convolutional layer s The features are connected through the feature fusion layer; b o represents the output layer bias;

[0112] In this embodiment, the training set of the input model includes: satellite images or drone images (providing ground cover and combustible information of the fire area), meteorological data (including temperature, humidity, wind speed and rainfall), geographic data (including geographical elements such as terrain, vegetation type, roads, and water sources), historical fire data (including the location, time and cause of past fires), remote sensing data (including vegetation index, combustible humidity, etc.);

[0113] The calculation formula of the vegetation index NDVI is:

[0114]

[0115] Among them: NIR represents the near-infrared band reflectance in remote sensing data; RED represents the red band reflectance in remote sensing data;

[0116] The calculation formula of the combustible humidity FM is:

[0117] FM=b1·T b b2

[0118] Where: b1 and b2 are empirical coefficients, which are set manually; T b is the microwave brightness in microwave remote sensing data;

[0119] The multi-scale convolution layer includes multiple layers, each of which uses multiple convolution kernels of different sizes to extract multi-scale features from image data. Convolution kernels of different scales capture the characteristics of natural fires occurring in different spatial ranges. The multi-scale convolution layer can be expressed by the following formula:

[0120]

[0121] Where: f s l represents the feature of the sth scale of the lth layer; σ represents the activation function; W s l represents the convolution kernel of the lth layer and the sth scale; x represents the input data; represents the bias of the s-th scale of the l-th layer;

[0122] After extracting the features of satellite images or drone images through the multi-scale convolutional layer, meteorological, geographical, and remote sensing data are used as additional features and fused through the feature fusion layer to enhance the accuracy of fire prediction.

[0123] As a preferred implementation of this embodiment, the earthquake prediction model is constructed based on a multimodal attention network, including an input layer, a multimodal feature extraction layer, an attention mechanism layer, a feature fusion layer and an output layer;

[0124] The training set of the input model includes earthquake precursor data (including microseismic, geomagnetic, groundwater level, stress and strain data), geological structure data (including fault lines, stratigraphic maps, seismic wave velocity distribution) and historical earthquake data (including magnitude, focal depth, epicenter location, earthquake time);

[0125] The multimodal feature extraction layer is constructed based on a convolutional neural network and a long short-term memory network, and performs independent feature extraction on each mode (i.e., earthquake precursors, geological structures, and historical data), as shown in the following formula:

[0126] f m =CNN m (X m )

[0127] f m =LSTM m (X m )

[0128] The above formula indicates that for each modality m, convolutional neural network CNN and long short-term memory network LSTM are used to extract features respectively;

[0129] The attention mechanism layer is used to perform weighted aggregation between the features of each modality through a multi-head attention mechanism, identify key signals, and control potential earthquake precursor signals by learning the importance weights between different modalities. For each modal feature f m , calculate the attention weights and perform weighted aggregation:

[0130] a m =softmax(W m ·f m +b m )

[0131] f att =∑ma m ·f m

[0132] Among them: a m represents the attention weight of modality m; W m and b m They represent the attention weight parameters, which are optimized through training;

[0133] The feature fusion layer fuses the output of the attention mechanism layer to form a unified feature representation;

[0134] The final model outputs the probability of earthquake occurrence P through the output layer earthquake , as shown in the following formula:

[0135] P earthquake =σ(W k ·f att +b k )

[0136] Where: σ represents the activation function; W k represents the weight of the output layer; f att represents the modal features after weighted aggregation; b k Represents the bias of the output layer.

[0137] As a preferred implementation of this embodiment, the distribution network disaster prevention decision model is constructed based on a dynamic Bayesian network model;

[0138] The training set input to the model includes the prediction data output by the above-mentioned disaster prediction models, the status data of the distribution network (including voltage, current, load, topology structure, etc.), and the historical fault data of the distribution network (i.e., the fault records under various disaster conditions, including fault location, type, recovery time, etc.);

[0139] Perform multimodal time series preprocessing on the input disaster prediction data, specifically combining multimodal input including disaster data and distribution network status data, and extracting time series features through sliding window technology;

[0140] Based on the pre-processed disaster prediction data, a dynamic Bayesian network is constructed. The dependency relationship between natural disasters and the distribution network status is calculated through the dynamic Bayesian network, and the disaster response strategy is dynamically inferred based on the dependency relationship. The specific steps are as follows:

[0141] Define the node set X t ={D t , G t}, indicating the state at time t, including the natural disaster state D t and the distribution network status G t ;

[0142] Define the state transition probability:

[0143] P(X t+1 |X t )=P(D t+1 |D t )·P(G t+1 |G t, , D t )

[0144] Calculate the conditional probability based on the historical fault data of the distribution network and estimate the dependency between nodes:

[0145]

[0146] Use Bayesian inference to calculate the posterior probability of disaster response strategies:

[0147]

[0148] Where: S represents the strategy set, and the strategy with the maximum posterior probability is selected as the response plan;

[0149] Optimize the distribution network operation strategy based on the inference results (such as adjusting the load, enabling backup power supply, etc.).

[0150] Embodiment 2:

[0151] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the present invention when executing the program.

[0152] Embodiment three:

[0153] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in any embodiment of the present invention is implemented.

[0154] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0155] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0156] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0157] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0158] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A distribution network differentiated disaster prevention decision-making method, characterized in that: The following steps are involved: Collect various historical disaster data in the distribution network area, and construct corresponding training sets after preprocessing various historical disaster data; Construct various disaster prediction models respectively, input the corresponding training sets into the corresponding types of disaster prediction models for training, obtain the trained disaster prediction models, and perform disaster prediction through the trained disaster prediction models; At the same time, a distribution network disaster prevention decision-making model is constructed, and the disaster prediction results output by various disaster prediction models are input into the distribution network disaster prevention decision-making model, and the distribution network disaster prevention decision-making model outputs the distribution network disaster prevention decision.

2. A distribution network differentiated disaster prevention decision-making method according to claim 1, characterized in that: The various disaster data include historical typhoon, flood, ice and snow, natural fire and earthquake data; The various disaster prediction models include typhoon prediction models, flood prediction models, ice and snow prediction models, natural fire prediction models and earthquake prediction models.

3. A distribution network differentiated disaster prevention decision-making method according to claim 2, characterized in that: The typhoon prediction model is constructed based on a spatiotemporal attention fusion network model; The spatiotemporal attention fusion network includes an input layer, a spatial attention layer, a spatiotemporal fusion layer and an output layer, wherein the spatial attention layer is used to identify important areas of the input feature data samples, and the spatiotemporal fusion layer is used to fuse the output data of the spatial attention layer with multi-dimensional data to obtain the predicted path of the typhoon; During the training of the typhoon prediction model, the model error C is calculated by the following formula: r : Where: C represents the absolute error of the typhoon center; R represents the typhoon radius; α p , β p Respectively represent the longitude and latitude of the typhoon forecast value; α r , β r They represent the longitude and latitude of the actual typhoon value respectively; kh represents the time difference between the typhoon forecast value and the actual typhoon value; When the model error is less than the preset threshold, the trained typhoon prediction model is obtained.

4. The method for differentiated disaster prevention decision-making in distribution network according to claim 2, characterized in that: The flood prediction model is constructed based on the long short-term memory network, as shown in the following formula: P flood ()=σ(W f ·[h t-1 x t ]+ f ) Where: P flood () represents the probability of flood occurrence in the next t days; σ represents the activation function; W f represents the weight matrix; h t-1 represents the hidden state at time step t-1; x t represents the input feature vector at time step t; b f Represents the bias term.

5. The method for differentiated disaster prevention decision-making in distribution network according to claim 2, characterized in that: The ice and snow prediction model is constructed based on a meta-learning model; Input historical ice and snow data into the meta-learning model for training, and use the trained meta-learning model to predict future temperatures and snowfall; Construct an energy balance model to calculate the net energy change E of snow cover s , as shown in the following formula: E s =I s -Q s -H s -E q Where: I s represents solar radiation; Q s Represents reflected radiation; H S Indicates sensible heat exchange; E q represents latent heat exchange; The ice and snow disaster scoring model S is constructed by integrating multiple factors, as shown in the following formula: Among them: α1, α2, α3 are weight coefficients respectively; Represents meteorological disaster factor; G(E s ) represents the energy balance factor; H(T s ,) represents the surface condition factor; Finally, by judging whether S is greater than the preset threshold, it is determined whether an ice and snow disaster has occurred.

6. The method for differentiated disaster prevention decision-making in distribution network according to claim 2, characterized in that: The natural fire prediction model is constructed based on a multi-scale convolutional neural network model, including an input layer, a multi-scale convolutional layer, a feature fusion layer, a fully connected layer, and an output layer, as shown in the following formula: P fire =σ(W o ·concat(f1,f2,…,f s )+b o ) Where: P fire represents the probability of fire occurrence; σ represents the activation function; W o represents the output layer weight; concat(f1,f2,…,f s ) represents f extracted by multi-scale convolutional layer s The features are connected through the feature fusion layer; b o Represents the output layer bias.

7. The method for differentiated disaster prevention decision-making in distribution network according to claim 2, characterized in that: The earthquake prediction model is constructed based on a multimodal attention network, including an input layer, a multimodal feature extraction layer, an attention mechanism layer, a feature fusion layer, and an output layer; The multimodal feature extraction layer is constructed based on a convolutional neural network and a long short-term memory network to perform independent feature extraction on each modality; The attention mechanism layer is used to perform weighted aggregation between features of each modality; The feature fusion layer fuses the output of the attention mechanism layer to form a unified feature representation; The final model outputs the probability of earthquake occurrence P through the output layer earthquake , as shown in the following formula: P earthquake =σ(W k ·f att +b k ) Where: σ represents the activation function; W k represents the weight of the output layer; f att Represents the modal features after weighted aggregation; b k Represents the bias of the output layer.

8. The method for differentiated disaster prevention decision-making in distribution network according to claim 2, characterized in that: The distribution network disaster prevention decision model is constructed based on a dynamic Bayesian network model; Perform multimodal time series preprocessing on the input disaster prediction data; A dynamic Bayesian network is constructed based on the preprocessed disaster prediction data. The dependency relationship between natural disasters and the distribution network status is calculated through the dynamic Bayesian network, and the disaster response strategy is dynamically inferred based on the dependency relationship. Optimize the distribution network operation strategy based on the inference results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Method and system for public network meteorological data application to power grid disaster prevention and reduction

    CN103383744A

  • Method for natural disaster data information mining and early warning response analysis based on GIS power grid

    CN103914951A

  • LSTM-based medium and small river short-term flood forecasting method

    CN109615011A

  • Power system elasticity evaluation method and system under extreme disaster condition

    CN113343467A

  • Power grid disaster management platform

    CN115691055A

Cited By

  • Diabetes blood sugar intelligent monitoring method and system based on artificial intelligence and medium

    CN119694598A

  • Artificial intelligence-based intelligent monitoring method and system for blood glucose of diabetes and medium

    CN119694598B

  • Power distribution network fault prediction and reinforcement method and system considering multi-disaster coupling

    CN122736271A