A method for predicting the situation of rainstorm-induced disasters with risk dependence of urban distribution network and drainage network

Abstract

The present invention discloses a method for predicting the situation of rainstorm disaster accidents in which the urban distribution network and the drainage network are risk-dependent, comprising the following steps: S1. Obtaining t S1. Use the relevant data of urban distribution network and drainage network at all times to solve the distribution of urban waterlogging under heavy rain, and map the solved urban waterlogging distribution to the flooding depth of power equipment based on the latitude and longitude information of power equipment; S2. Judge whether a large-scale power outage will occur based on the critical conditions that cause large-scale power outages; S3. Consider the uncertain parameters that cause urban waterlogging and power outages, predict the probability of power outages through the power outage probability density calculation model, and use the multi-source data-driven method to roll over the model power outage risk distribution results.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method for predicting the situation of rainstorm-induced disaster accidents in which the risks of urban distribution networks and drainage water networks are interdependent. Background Art

[0002] Heavy rainfall can easily cause rainwater to overflow in the drainage network, forming waterlogging, causing local power transmission and transformation equipment to fail or short-circuit and trip, and drainage and flood control pump stations to be forced to stop due to power outages, leading to further expansion of the scope of urban flooding, which in turn causes waterlogging accidents and power failures to interact and cascade between the drainage network and the power grid, exacerbating the risk of heavy rain and flood disasters and large-scale power outages in the power grid. Research on distribution network accident simulation and power outage location prediction methods under extreme rainfall can effectively improve the accuracy and scientific nature of power grid disaster prevention and mitigation work, strengthen extreme weather response capabilities, and reduce the risk of power outages.

[0003] At present, most of the research on the location and prediction of power outage accidents under extreme rainfall only obtains the distribution of urban waterlogging by solving the two-dimensional hydrological and hydrodynamic model of the urban surface, and uses the power grid geographic information system data to map the distribution equipment and the calculated flooding depth to obtain the power outage risk distribution of the substation area, without further analyzing the impact of the power outage and shutdown of urban drainage equipment on the evolution of the power grid rainstorm disaster accident chain, and the problem of the early waterlogging time of distribution equipment caused by the cross-network interactive propagation of urban waterlogging and power failures. In addition, the difference in the time scale of waterlogging evolution and power grid failures is not considered, the series of emergency measures taken by the power grid for waterlogging failures are not considered, and the problems of voltage crossing and disconnection of distributed wind and solar units are not involved, resulting in low power outage prediction accuracy. Summary of the invention

[0004] In view of this, the present invention provides a method for predicting the situation of rainstorm disaster accidents that are dependent on the risks of urban distribution networks and drainage water networks, so as to at least solve the problem of low accuracy in predicting the probability of power outage accidents in distribution station areas under the spatiotemporal differentiation of existing rainstorm disaster risks in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] The method for predicting the situation of rainstorm disaster accidents with the risk dependence of urban distribution network and drainage network includes the following steps:

[0007] S1. Acquisition The relevant data of urban distribution network and drainage network are used to solve the distribution of urban waterlogging under heavy rain weather. Based on the longitude and latitude information of power equipment, the solved distribution of urban waterlogging is mapped to the flooding depth of power equipment.

[0008] S2. Determine whether a large-scale power outage will occur based on the critical conditions that cause a large-scale power outage; wherein, a multi-time-scale power-drainage network fault evolution model is established to quantify the critical conditions that cause a large-scale power outage;

[0009] S3. Consider the accidents that may cause urban flooding and power outages X The power outage probability density calculation model is used to predict the occurrence of power outage accidents. X probability.

[0010] Preferably, the specific contents of constructing the power outage accident probability density calculation model in S3 include:

[0011] Obtained from the probability space of uncertain parameters Representative discrete points , establish a power outage accident probability density calculation model, solve the power outage accident probability density calculation model to obtain the joint probability density function of each representative discrete point , and indivual The cumulative summation results in the power outage accident The probability density evolution expression of is used to quantify the power outage risk; the power outage accident probability density calculation model is as follows:

[0012] (1)

[0013] in, Represents a representative discrete point Power outage With uncertain parameters The joint probability density function between is the time variable, Indicates A power outage at a distribution network node Indicates the network configuration node number. Indicates the total number of distribution network nodes. express Derivative with respect to time.

[0014] Preferably, the specific content of S1 includes:

[0015] S11. Construct a one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model, including a two-dimensional urban surface hydrological and hydrodynamic model and a one-dimensional hydrodynamic model of the drainage network;

[0016] Among them, the two-dimensional hydrological and hydrodynamic model of the urban surface is:

[0017] (2)

[0018] (3)

[0019] (4)

[0020] (5)

[0021] (6)

[0022] In the formula, Indicates time, and Respectively represent the longitude and latitude of the city’s geographical coordinates; represents the conservation vector, and Respectively and The flux vector in the direction, , and They represent the surface slope source term, ground friction source term and overflow source term respectively; For water depth, and They are and The flow velocity value in the direction, is the acceleration due to gravity, is the terrain elevation, is the Manning roughness coefficient; and Respectively indicate along Direction and The bottom slope source term in the direction of and Respectively indicate along and The friction source term in the direction; is the rainfall, is the flow rate into the drainage well, is the pumping flow rate of the pump station, is the evaporation of rainwater, is the amount of water that infiltrates into the soil;

[0023] The one-dimensional hydrodynamic model of the drainage network is:

[0024] (7)

[0025] In the formula, is the cross-sectional area of ​​water in the drainage pipe, is the cross-sectional flow rate, The amount of water discharged into the river through the drainage pipe. is the pipeline water depth, and are the bottom slope and friction slope of the drainage pipe, Indicates the spatial position of the pipeline;

[0026] S12. Solve for the flow rate into the drainage well and pumping station discharge flow , providing parameters for solving the one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model:

[0027] The urban drainage network is coupled through flow exchange between the drainage well and the surface, and the flow into the drainage well is solved by the one-dimensional hydrodynamic model of the drainage network. ;

[0028] Pumping station discharge flow The solution:

[0029] (8)

[0030] in, Indicates the active power consumed by the pump station, is the energy loss coefficient of the pumping station, is the water density, is the static head of the pump;

[0031] S13. Aiming at the hydrological and hydrodynamic model of urban waterlogging, the regional calculation unit is simulated by arbitrary triangular unstructured grid division, and the Godunov format is used to construct the Riemann problem to solve the numerical flux of the water flow grid microelement. The second-order spatial reconstruction of the MUSCL-TVD format and the second-order discretization of the time direction of the Runge-Kutta format are performed to obtain the distribution of water depth of urban waterlogging under heavy rain weather, and the water depth of urban waterlogging is mapped to the flooding depth of power equipment based on the latitude and longitude information of power equipment.

[0032] Preferably, the specific contents of "quantifying the critical conditions that cause large-scale power outages by establishing a multi-time-scale power-drainage water network fault evolution model" in S2 include:

[0033] S21. Determine whether there is equipment failure based on whether the submerged depth of the power equipment reaches the flood prevention height;

[0034] S22. If a fault occurs, jump to S23 to perform distribution network transient analysis. If no fault occurs, jump to S24 to perform distribution network steady-state analysis.

[0035] S23. Select a short-time transient analysis scale and perform transient simulation on the distribution network fault, determine the transient fault according to the voltage waveform in the simulation result, and urgently remove the existing transient fault;

[0036] S24. Select the long-term steady-state analysis scale, calculate the dynamic flow of the urban distribution network, and optimize the flow for the results with faults: if the flow does not converge, remove the corresponding fault and re-perform S23 and S24; if the flow converges, update the urban distribution network topology and the drainage equipment to pump out the accumulated water power, and output The load loss at the moment and enter S25;

[0037] S25. Determine whether the rainstorm has stopped at the current moment: If the rainstorm is still continuing, The power grid and drainage equipment data are updated at all times as At the initial condition, repeat S21-S25; if the rainstorm stops, jump to S26 to quantify the critical condition;

[0038] S26. Draw a total load reduction change graph based on the load loss in each time period to quantify the critical conditions that trigger urban waterlogging and cascading failures of distribution equipment caused by waterlogging, leading to large-scale power outages.

[0039] Preferably, we obtain Representative discrete points The specific contents include:

[0040] make Represents all uncertain parameters, including the uncertainty of the spatial and temporal distribution of heavy rain , User load uncertainty and uncertainty in wind and solar power generation output :

[0041] (9)

[0042] in is the total number of random variables in the electricity-water risk dependent network, is the number of uncertainty parameters of the temporal and spatial distribution of heavy rain, is the number of user load uncertainty parameters, is the number of uncertainty parameters of wind and solar power generator output, For the The uncertainty parameters of the temporal and spatial distribution of rainstorms, For the User load uncertainty parameters, For the The uncertainty parameters of wind and solar power unit output, Indicates Parameter pairs;

[0043] In the uncertain parameter space Select a random parameter set as the initial point set:

[0044] (10)

[0045] in represents the number of initial point sets, Indicates indivual ;

[0046] Let the joint probability density function of all uncertain parameters in the initial point set be , then the probability of each initial point is used to calculate its probability in the uncertain parameter space The corresponding volume To describe:

[0047] (11)

[0048] Genetic algorithm is used to process the initial point set to ensure that the GF deviation is minimized and obtain Representative discrete points:

[0049] (12).

[0050] Optimally, the uncertainty of the temporal and spatial distribution of heavy rain It directly affects the state of the drainage network and indirectly affects the operation of the distribution network through urban waterlogging, which can be expressed as:

[0051] (13)

[0052] In the formula, is the number of rainfall samples; , and is the bandwidth of the kernel density estimate;

[0053] User load uncertainty In line with Gaussian distribution, the output of wind and solar power units is uncertain It conforms to the Weibull distribution, which directly affects the distribution network flow analysis and indirectly affects the drainage network through the shutdown of pump station nodes.

[0054] Preferably, the specific contents of establishing the power outage accident probability density calculation model include:

[0055] Considering the interactive propagation of cascading faults between the urban distribution network and the drainage network, the urban distribution network and drainage network model are simplified as follows:

[0056] (14)

[0057] (15)

[0058] In the formula, Indicates the area where the urban distribution network is shut down. Representation Node Power outage situation; and are deterministic parameters of the distribution network and drainage network, among which Represents the safe and stable operation parameters of the urban distribution network. Represents drainage network parameters such as rainfall and river water levels; Indicates the waterlogging and outage of power distribution equipment; Indicates that the urban distribution network outage areas are respectively and The corresponding transformation relationship of Indicates the depth of water accumulation in urban waterlogging. express The city is deeply flooded at the location; Represents the power outage and shutdown of drainage equipment. The depth of waterlogging is respectively and The corresponding transformation relationship of

[0059] Taking into account the urban distribution network, drainage network and fault evolution time factors, an electric-water coupled power system model under heavy rain weather is established:

[0060] (16)

[0061] In the formula, Indicates power outage and distribution network outage area , urban waterlogging depth and time Transformation relationship between them;

[0062] After hiding the deterministic parameters of the distribution network and drainage network, the power outage accident Further simplified expression:

[0063] (17)

[0064] Then the component form of a single distribution network node is expressed as:

[0065] (18)

[0066] remember time The corresponding evolution region is , then the probability space and state space of the electricity-water risk dependent network are described as follows:

[0067] (19)

[0068] (20)

[0069] In the formula, express The probability corresponding to the evolving region over time; , Respectively and The power outage status of the system at all times; yes time Region boundaries of state space; express The overall rate of change over time is expressed as ; is the unit vector on the boundary of the state space region; is the infinitesimal area element on the boundary of the state space;

[0070] Combining the Gaussian divergence theorem and simplifying the probability space and state space of the electricity-water risk dependent network, the generalized probability density calculation equation for power outage accidents is obtained:

[0071] (twenty one)

[0072] In the formula express With uncertain parameters The joint probability density function between Indicates power outage The time derivatives of the components in are denoted by ;

[0073] The uncertainty parameter probability space is selected Representative discrete points Calculate system parameters in the probability density calculation model of power outage accidents ,Will The corresponding Consider it as a known quantity and directly substitute it into the relevant power outage accident The generalized probability density calculation equation of is used to obtain the probability density calculation model of power outage accidents.

[0074] Preferably, the power outage probability density calculation model is solved to obtain the joint probability density function of each representative discrete point: , and indivual The cumulative summation results in the power outage accident The specific contents of the probability density evolution expression to quantify the power outage risk include:

[0075] The initial conditions of the power outage probability density calculation model are expressed as:

[0076] (twenty two)

[0077] In the formula, is the Dirac function; represents the initial moment; Indicates a deterministic initial blackout event; Indicates uncertain parameters Take representative discrete values The corresponding joint probability density function value when ;

[0078] Establish the research area and The spatiotemporal difference network is used to solve the probability density calculation model of power outage accidents using the finite difference method with total variation non-increasing format:

[0079] (twenty three)

[0080] In the formula, the accident variable and time Respectively and The grid lines are evenly divided, express No. Grid and No. Grid The value at ; and Respectively represent division and The grid line spacing; express and The grid ratio of express Overall rate of change over time abbreviation of; ; , ; ; The flux limiter is expressed as:

[0081] (twenty four)

[0082] in, is a unit step function; is a Roe-Sweby low dissipation flux limiter with ; Satisfy the same rules;

[0083] By solving the accident probability density calculation model, we can obtain the joint probability density function of each representative discrete point. Finally, the power outage risk is quantified by summing up all the situations, and the power outage accident is obtained. The probability density evolution expression of is as follows:

[0084] (25).

[0085] Preferably, after constructing the power outage accident probability density calculation model, S3 also includes error correction of the model-driven output results, and the specific contents include:

[0086] (1) Obtain equipment to collect multi-source historical data from the air, space, and ground, obtain the deviation between the model result and the actual probability from the accident probability density calculation model, and pre-process the data;

[0087] (2) Divide the dataset into a training set for training the neural network model, a validation set for adjusting model parameters and preventing overfitting, and a test set for the final evaluation of model performance;

[0088] (3) A real-time correction strategy for power outage risk probability prediction based on a large language model (LLM) is constructed, with multi-source data as input and risk probability prediction error as output: first, the prompting project of the LLM is constructed; second, the parameters of all layers of the LLM are frozen and two learnable neural network modules are embedded on this basis. The first neural network module is located between the multi-source meteorological data and the LLM data input interface, and is used to align the input multi-source data with the text data received by the LLM. Its output result is connected in series with the prompting project as the LLM text input; the second neural network module is located between the text output of the LLM and the risk probability prediction error, and is used to map the LLM output to the time series data of the risk probability prediction error;

[0089] (4) Simplify the real-time correction strategy of power outage risk probability prediction based on LLM into a deviation learning function ,before The multi-source monitoring data of the step length predicts the probability error of power outage risk prediction driven by the future A step length model, which is specifically expressed as follows:

[0090] (26)

[0091] Where: ; ; express Nodes in Time-varying space-ground multi-source data, writing ,in Indicates Nodes in Moment Multi-source data features, written ; express Nodes in The prediction probability error at time t is written as ;

[0092] (5) The relevant data of the urban distribution network and drainage network and the short-term rainfall forecast data are input into the power outage risk model to predict the power outage probability. The real-time air-space-ground multi-source monitoring data is input into the trained large model network to obtain the prediction probability error. The model risk distribution results are rolled corrected, and the model parameters are regularly fine-tuned using new data. The rolling correction formula for the power outage probability is expressed as follows:

[0093] (27)

[0094] In the formula, and Respectively represent the rolling correction process The power outage probability after time correction and the power outage probability predicted by the model.

[0095] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method for predicting the situation of rainstorm disaster accidents in which the urban distribution network and the drainage network are risk-dependent, which has the following beneficial effects:

[0096] Firstly, the present invention reveals the positive feedback superposition and expansion mechanism of the interactive propagation of urban waterlogging accidents and power failures under heavy rain disasters. By considering the impact of power failure and shutdown of urban drainage equipment on the evolution of the accident chain of heavy rain disasters in the power grid, a simulation model of the interactive propagation of urban waterlogging and power grid failures under heavy rainfall is established, and the critical conditions for triggering urban waterlogging and cascading failures of distribution equipment caused by waterlogging and causing large-scale power outages are quantified, which effectively improves the accuracy and scientific nature of the power grid's disaster prevention and mitigation work; secondly, a probability density calculation model for accidents with dependent risks of power-water conservancy networks under heavy rainfall is proposed, and a mapping between state space and probability space is established through the joint probability density constructed by the waterlogged state variables of the distribution network and the spatiotemporal random variables of rainfall intensity. The probability density calculation equations for wind and solar power unit disconnection and power outages in distribution stations due to waterlogging were derived based on the principle of conservation of probability density. A spatiotemporal difference network was established and the finite difference method with a total variation non-increasing format was used to solve the problem. The probability of wind and solar power unit disconnection and power outages in distribution stations was quantitatively evaluated compared with traditional methods. The calculation burden is smaller, the power outage time is advanced, and the prediction accuracy is higher. Finally, a power outage location prediction method driven by the collaboration of multi-source data and models is proposed. This method mines the correlation between the real-time monitoring data of the integrated air-ground-space and the prediction error of the risk model, and performs a rolling correction on the model risk distribution results based on multi-source data, so as to further improve the accuracy of power outage prediction in urban distribution stations under heavy rain. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0098] Figure 1 It is a flow chart of a method for predicting the situation of rainstorm disaster accidents in which the urban distribution network and the drainage water network are risk-dependent, as disclosed in the present invention;

[0099] Figure 2 A schematic diagram of interactive propagation of urban waterlogging and power grid failures under a rainstorm disaster disclosed in an embodiment of the present invention;

[0100] Figure 3 A diagram showing changes in total load reduction in a distribution network under heavy rainfall according to an embodiment of the present invention;

[0101] Figure 4 It is the quantified result of the probability of power outage accident in the system disclosed in the embodiment of the present invention. DETAILED DESCRIPTION

[0102] 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.

[0103] The present invention provides a method for predicting the situation of rainstorm disaster accidents in which the urban distribution network and the drainage network are risk-dependent. Figure 1 As shown, the following steps are included:

[0104] S1. Acquisition The relevant data of urban distribution network and drainage network are used to solve the distribution of urban waterlogging under heavy rain weather. Based on the longitude and latitude information of power equipment, the solved distribution of urban waterlogging is mapped to the flooding depth of power equipment.

[0105] S2. Determine whether a large-scale power outage will occur based on the critical conditions that cause a large-scale power outage; wherein, a multi-time-scale power-drainage network fault evolution model is established to quantify the critical conditions that cause a large-scale power outage;

[0106] S3. Consider the accidents that may cause urban flooding and power outages X The power outage probability density calculation model is used to predict the occurrence of power outage accidents. X probability.

[0107] In order to further implement the above technical solution, the specific contents of constructing the power outage accident probability density calculation model in S3 include:

[0108] Obtained from the probability space of uncertain parameters Representative discrete points , establish a power outage accident probability density calculation model, solve the power outage accident probability density calculation model to obtain the joint probability density function of each representative discrete point , and indivual The cumulative summation results in the power outage accident The probability density evolution expression of is used to quantify the power outage risk; the power outage accident probability density calculation model is as follows:

[0109] (1)

[0110] in, Represents a representative discrete point Power outage With uncertain parameters The joint probability density function between is the time variable, Indicates A power outage at a distribution network node Indicates the network configuration node number. Indicates the total number of distribution network nodes. express Derivative with respect to time.

[0111] In order to further implement the above technical solution, the specific contents of S1 include:

[0112] S11. Construct a one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model, including a two-dimensional urban surface hydrological and hydrodynamic model and a one-dimensional hydrodynamic model of the drainage network;

[0113] Among them, the two-dimensional hydrological and hydrodynamic model of the urban surface is:

[0114] (2)

[0115] (3)

[0116] (4)

[0117] (5)

[0118] (6)

[0119] In the formula, Indicates time, and Respectively represent the longitude and latitude of the city’s geographical coordinates; represents the conservation vector, and Respectively and The flux vector in the direction, , and They represent the surface slope source term, ground friction source term and overflow source term respectively; For water depth, and They are and The flow velocity value in the direction, is the acceleration due to gravity, is the terrain elevation, is the Manning roughness coefficient; and Respectively indicate along Direction and The bottom slope source term in the direction of and Respectively indicate along and The friction source term in the direction; is the rainfall, is the flow rate into the drainage well, is the pumping flow rate of the pump station, is the rainfall, which can be obtained through weather forecast information. is the evaporation of rainwater, which can be estimated by Meyer's empirical formula, is the amount of water that infiltrates into the soil, which can be solved using Horton’s infiltration formula;

[0120] The one-dimensional hydrodynamic model of the drainage network is:

[0121] (7)

[0122] In the formula, is the cross-sectional area of ​​water in the drainage pipe, is the cross-sectional flow rate, The amount of water discharged into the river through the drainage pipe. is the pipeline water depth, and are the bottom slope and friction slope of the drainage pipe, Indicates the spatial position of the pipeline;

[0123] S12. Solve for the flow rate into the drainage well and pumping station discharge flow , providing parameters for solving the one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model:

[0124] The urban drainage network is coupled through the flow exchange between the drainage well and the surface, and the flow into the drainage well Solved by one-dimensional hydrodynamic model of drainage pipe network;

[0125] Pumping station discharge flow The solution:

[0126] (8)

[0127] in, Indicates the active power consumed by the pump station, is the energy loss coefficient of the pumping station, is the water density, is the static head of the pump;

[0128] S13. Aiming at the hydrological and hydrodynamic model of urban waterlogging, the regional calculation unit is simulated by arbitrary triangular unstructured grid division, and the Godunov format is used to construct the Riemann problem to solve the numerical flux of the water flow grid microelement. The second-order spatial reconstruction of the MUSCL-TVD format and the second-order discretization of the time direction of the Runge-Kutta format are performed to obtain the distribution of water depth of urban waterlogging under heavy rain weather, and the water depth of urban waterlogging is mapped to the flooding depth of power equipment based on the latitude and longitude information of power equipment.

[0129] In order to further implement the above technical solutions, urban flooding under heavy rainfall will cause waterlogging of power distribution equipment, which will cause power outages in the substation area and shutdown of drainage pump stations, leading to the spread of waterlogging and causing large-scale power outages in the urban distribution network. The interactive propagation process of waterlogging accidents and power failures under heavy rain disasters is as follows: Figure 2 As shown in Figure 2, the specific contents of "Quantifying the critical conditions that cause large-scale power outages by establishing a multi-time scale power-drainage network fault evolution model" in S2 include:

[0130] (1) Input The relevant data of urban distribution network and drainage network are collected at all times. The urban waterlogging distribution is solved according to the one-dimensional / two-dimensional coupled hydrological and hydrodynamic model. The water depth of power equipment is obtained by combining the longitude and latitude information of the power grid geographic information system software platform. .

[0131] (2) When the water depth of the power equipment reaches the flood prevention height, a fault occurs, and it is judged whether there is an equipment fault: if no fault occurs, jump directly to step (4) to perform a steady-state analysis of the distribution network; if a fault occurs, jump to step (3) to perform a transient analysis of the distribution network, and then jump to step (4) to perform a steady-state analysis after the system remains stable and the faults such as new energy disconnection are eliminated.

[0132] (3) 20ms is selected as the short-term transient analysis scale, and the electromechanical transient model of the synchronous generator is combined with the second-generation new energy general dynamic model to construct an instantaneous transient simulation model of urban distribution network faults. Based on the resulting voltage waveform, problems such as distribution network instability, severe line overload, and wind and solar unit disconnection are judged to ensure that the chain fault simulation meets the needs.

[0133] (4) Steady-state analysis First, the dynamic power flow of the urban distribution network is calculated, and the power flow is optimized for the results with faults: If the power flow does not converge, the corresponding fault is removed and steps (3) and (4) are repeated; if the power flow converges, the grid structure and drainage equipment power are updated, and the output The result at the moment goes to step (5). The long-term steady-state analysis scale is selected as 5 minutes, which usually takes into account the general overload, heavy load disconnection and power outage of the power line during the flood diffusion process, and includes a short time scale. The power flow optimization aims to minimize the total load reduction of each node, and considers the power flow constraints, line voltage capacity constraints, generator active and reactive power climbing constraints, and load reduction constraints for calculation.

[0134] (5) Determine whether the rainstorm has stopped at the current moment: if the rainstorm has stopped, jump to step (6) to quantify the critical condition; if the rainstorm is still continuing, The power grid and drainage equipment data obtained at the moment are used as Repeat steps (1) to (5) for the initial conditions at time t.

[0135] (6) Based on the load loss change rate index in each time period, the critical conditions that trigger urban waterlogging and cascading failures of distribution equipment, leading to large-scale power outages, are quantified.

[0136] It should be noted that:

[0137] In this embodiment, the short time scale considers the instability of the distribution network, severe line overload, and wind and solar unit disconnection caused by the instantaneous failure of equipment due to waterlogging; the long time scale considers the general overload of power lines, heavy load disconnection, and power outage of pump stations during the spread of waterlogging, and includes the short time scale.

[0138] In order to further implement the above technical solutions, the rainfall distribution has strong spatiotemporal uncertainty, which leads to the time-varying probability distribution of urban waterlogging and power outage accidents, and the accident risk is interactively propagated in the urban distribution network and the drainage network. An accident probability density calculation model with risk dependence between the two networks under heavy rainfall can be established to quantitatively evaluate the probability distribution of power outage risk; obtain the probability space of uncertain parameters Representative discrete points The specific contents include:

[0139] To ensure uniformity of symbols, Represents all uncertain parameters, including the uncertainty of the spatial and temporal distribution of heavy rain , User load uncertainty and uncertainty in wind and solar power generation output :

[0140] (9)

[0141] in is the total number of random variables in the electricity-water risk dependent network, is the number of uncertainty parameters of the temporal and spatial distribution of heavy rain, is the number of user load uncertainty parameters, is the number of uncertainty parameters of wind and solar power generator output, For the The uncertainty parameters of the temporal and spatial distribution of rainstorms, For the User load uncertainty parameters, For the The uncertainty parameters of wind and solar power unit output, Indicates Parameter pairs;

[0142] In the uncertain parameter space Select a random parameter set as the initial point set:

[0143] (10)

[0144] in represents the number of initial point sets, Indicates indivual ;

[0145] The probability space is optimally divided using Voronoi regions: the representative discrete points are used to form a Delaunay triangulation network through the angle discrimination method, and then the circumcenter of each triangulation network is calculated and recorded, and the triangle list is traversed to connect the circumcenters of adjacent triangles to form a Voronoi diagram with triangle vertices as generators. Here, the joint probability density function of all uncertain parameters is recorded as , then the probability of each discrete point can be used to calculate its probability in the uncertain parameter space The corresponding volume To describe:

[0146] (11)

[0147] Genetic algorithm is used to process the initial point set to ensure that the GF deviation is minimized and obtain Representative discrete points:

[0148] (12).

[0149] The newly generated representative discrete points can be used to effectively solve the power outage accident probability density calculation model and quantitatively assess the power outage risk.

[0150] In order to further implement the above technical solutions, the uncertainty of the temporal and spatial distribution of heavy rain It directly affects the state of the drainage network and indirectly affects the operation of the distribution network through urban waterlogging, which can be expressed as:

[0151] (13)

[0152] In the formula, is the number of rainfall samples; , and is the bandwidth of the kernel density estimate;

[0153] User load uncertainty In line with Gaussian distribution, the output of wind and solar power units is uncertain It conforms to the Weibull distribution, which directly affects the distribution network flow analysis and indirectly affects the drainage network through the shutdown of pump station nodes.

[0154] In order to further implement the above technical solution, the specific contents of establishing the probability density calculation model of power outage accidents include:

[0155] Considering the interactive propagation of cascading faults between the urban distribution network and the drainage network, the urban distribution network and drainage network model are simplified as follows:

[0156] (14)

[0157] (15)

[0158] In the formula, Indicates the area where the urban distribution network is shut down. Representation Node Power outage situation; and are deterministic parameters of the distribution network and drainage network, among which Represents the safe and stable operation parameters of the urban distribution network. Represents drainage network parameters such as rainfall and river water levels; Indicates the outage of power distribution equipment due to flooding. The equipment failure probability under historical rainfall events and the actual failure condition analysis can be combined to select the representative value of the typical failure probability as the critical condition for judging whether a failure occurs; Indicates that the urban distribution network outage areas are respectively and The corresponding transformation relationship of Indicates the depth of water accumulation in urban waterlogging. express The city is deeply flooded at the location; Represents the power outage and shutdown of drainage equipment. The depth of waterlogging is respectively and The corresponding transformation relationship of

[0159] Taking into account the urban distribution network, drainage network and fault evolution time factors, an electric-water coupled power system model under heavy rain weather is established:

[0160] (16)

[0161] In the formula, Indicates power outage and distribution network outage area , urban waterlogging depth and time Considering the interactive cascade propagation law of urban waterlogging accidents and power failures under heavy rain disasters, it can be judged that the depth of urban waterlogging, the fault status of power distribution equipment, the calculation results of distribution network flow, the power failure and shutdown of drainage equipment all have certain uncertainties, and their uncertainties are essentially still caused by These three types of uncertainties arise. Therefore, after hiding the deterministic parameters of the distribution network and the drainage network, equation (16) is further simplified as follows:

[0162] (17)

[0163] This formula summarizes the overall evolution of power outage accidents over time, and its component form with respect to distribution network nodes can be expressed as:

[0164] (18)

[0165] In the electricity-water interdependent risk network, there are no existing random factors disappearing, and no new random factors being added. It can be regarded as a conservative random system and can be described from two equivalent perspectives: random events and state space. If we directly model and analyze its probability, it will involve high-dimensional stochastic differential equations that are difficult to solve. However, from a probability perspective, observing the evolution of combined events under each uncertain representative point can achieve decoupling and dimensionality reduction of the equation and simplify the complex calculation process.

[0166] remember time The corresponding evolution region is , then the probability space and state space of the electricity-water risk dependent network are described as follows:

[0167] (19)

[0168] (20)

[0169] In the formula, express The probability corresponding to the evolving region over time; , Respectively and The power outage status of the system at all times; yes time Region boundaries of state space; express The overall rate of change over time is expressed as ; is the unit vector on the boundary of the state space region; is the infinitesimal area element on the boundary of the state space;

[0170] Combining the Gaussian divergence theorem and simplifying the probability space and state space of the electricity-water risk dependent network, the generalized probability density calculation equation for power outage accidents is obtained:

[0171] (twenty one)

[0172] In the formula express With uncertain parameters The joint probability density function between Indicates power outage The time derivatives of the components in are denoted by ;

[0173] Based on the probability density calculation equation of power outage accidents, combined with the model coefficients calculated from the representative points of uncertain parameters, The joint probability density function is obtained. Usually, the analytical solution of this equation is difficult to obtain. The finite difference method with total variation non-increasing format can be used to solve the above problems and quantify the power outage risk.

[0174] The uncertainty parameter probability space is selected Representative discrete points Calculate system parameters in the probability density calculation model of power outage accidents ,Will The corresponding Substituting the known variables directly into the generalized probability density calculation equation of the power outage accident variables will result in the power outage accident probability density calculation model.

[0175] In order to further implement the above technical solution, the power outage probability density calculation model is solved to obtain the joint probability density function under each representative discrete point. , and indivual The cumulative summation results in the power outage accident The specific contents of the probability density evolution expression to quantify the power outage risk include:

[0176] The initial conditions of the power outage probability density calculation model are expressed as:

[0177] (twenty two)

[0178] In the formula, is the Dirac function; represents the initial moment; Indicates a deterministic initial blackout event; Indicates uncertain parameters Take representative discrete values The corresponding joint probability density function value when ;

[0179] Establish the research area and The spatiotemporal difference network is used to solve the probability density calculation model of power outage accidents using the finite difference method with total variation non-increasing format:

[0180] (twenty three)

[0181] In the formula, the accident variable and time Respectively and The grid lines are evenly divided, express No. Grid and No. Grid The value at ; express and The grid ratio of ; , ; ; The flux limiter is expressed as:

[0182] (twenty four)

[0183] in, is a unit step function; is a Roe-Sweby flux limiter with relatively small dissipation. ; Satisfy the same rules;

[0184] By solving the accident probability density calculation model, we can obtain the joint probability density function of each representative discrete point. Finally, the power outage risk is quantified by summing up all the situations, and the power outage accident is obtained. The probability density evolution expression of is as follows:

[0185] (25).

[0186] In order to further implement the above technical solution, considering the influence of factors such as the accuracy of the power outage probability density calculation model, the accuracy of the differential algorithm solution, and uncertain parameters, there is a certain difference between the predicted power outage probability and the actual power outage probability. Therefore, a power outage location prediction method driven by multi-source data and models is proposed, and the error correction of the model-driven output results is performed. After S3 constructs the power outage probability density calculation model, it also includes error correction of the model-driven output results. The specific contents include:

[0187] (1) Collect historical data from multiple sources including air, ground and space from meteorological satellites, rainfall radars, rain gauges, distribution network sensors and other equipment, obtain the deviation between the model results and the actual probability from the accident probability density calculation model, and perform pre-processing such as data cleaning and normalization on the data.

[0188] (2) The dataset is divided into a training set for training the neural network model, a validation set for adjusting model parameters and preventing overfitting, and a test set for final evaluation of model performance, accounting for 80%, 10%, and 10% of the total data, respectively.

[0189] (3) Construct a real-time correction strategy for power outage risk probability prediction based on a large language model (LLM). The model takes multi-source data as input and risk probability prediction error as output. First, the prompt project of LLM is constructed, including specifying the tasks that LLM needs to complete, describing the output and output data, and the topology of the distribution network. Secondly, the parameters of all layers of LLM are frozen and two learnable neural network modules are embedded on this basis. The first module is located between the multi-source meteorological data and the LLM data input interface. Its main function is to align the input multi-source data with the text data received by LLM. The output of this module is connected in series with the prompt project as the LLM text input. The second module is located between the text output of LLM and the risk probability prediction error. Its main function is to map the LLM output to the time series data of the risk probability prediction error. Since the model freezes the parameters of all layers of LLM, only a small amount of time series data and a few training cycles are needed to adapt LLM before it can be used.

[0190] (4) The real-time correction strategy for power outage risk probability prediction based on LLM can be simplified as a deviation learning function ,before The multi-source monitoring data of the step length predicts the probability error of power outage risk prediction driven by the future A step length model, which is specifically expressed as follows:

[0191] (26)

[0192] Where: ; ; express Nodes in The multi-source data of space, land and sky at any time can be written as ,in Indicates Nodes in Moment D Multi-source data features can be written as ; express Nodes in The prediction probability error at time can be written as .

[0193] (5) Input the parameters of the urban distribution network-drainage network and short-term rainfall forecast data into the power outage risk model to predict the power outage probability, input the real-time space-ground multi-source monitoring data into the trained large model network to obtain the prediction probability error, perform rolling correction on the model risk distribution results, and regularly use new data to fine-tune the model parameters. The rolling correction formula for the power outage probability can be expressed as follows:

[0194] (27)

[0195] In the formula, and Respectively represent the rolling correction process The power outage probability after time correction and the power outage probability predicted by the model.

[0196] The present invention will be further described below through specific experiments:

[0197] The present invention is used in the prediction of rainstorm disaster accidents with the risk of urban distribution network-drainage network failure. Figure 1 The process shown is implemented in Python. The total load reduction of the distribution network changes under the interactive propagation of urban waterlogging and power grid faults as shown in the figure below: Figure 3 As shown in the figure, the quantified results of the probability of power outage accidents under heavy rainfall are as follows: Figure 4 The comparison of the results of different prediction methods is shown in Table 1:

[0198]

[0199] pass Figure 3 The exponential growth law of waterlogging propagation, waterlogging degree of power distribution equipment and power outage damage can be observed. The load reduction ratio is the ratio of the load reduction before a certain moment to the load reduction during the entire rainstorm period. The maximum point of total load reduction change rate within the load reduction ratio less than 20% can be selected as the critical condition for triggering large-scale power outages in the city. Figure 4 It can be seen that due to the consideration of the positive feedback superposition and expansion mechanism of the interactive propagation of urban waterlogging and power grid faults, the power outage accident is earlier than the traditional method, and the prediction accuracy is higher. From Table 1, it can be seen that compared with other methods, the prediction error of the waterlogging fault time of the present invention is reduced by more than 47%, and the accuracy of the power outage area is improved by more than 7.5%.

[0200] In summary, the present invention has the following beneficial effects:

[0201] 1. The positive feedback superposition and expansion mechanism of the interactive propagation of urban waterlogging accidents and power failures under heavy rain disasters was revealed. Considering the impact of power outage and shutdown of urban drainage equipment on the evolution of the power grid rainstorm disaster accident chain, a simulation model of the interactive propagation of urban waterlogging and power grid failures under heavy rainfall was established, and the exponential growth law of waterlogging propagation, the degree of waterlogging of distribution equipment, and power outage losses was discovered. The critical conditions that trigger the cascading failure of urban waterlogging and waterlogging of distribution equipment and cause large-scale power outages were quantified.

[0202] 2. A probability density calculation model for the risk-dependent accidents of the power-water network under heavy rainfall is proposed. The mapping between state space and probability space is established based on the joint probability density of the distribution network flooding state variables and the spatiotemporal random variables of rainfall intensity. The probability space is optimally partitioned using the Voronoi region. Based on the principle of probability density conservation, the probability density calculation equations for wind and solar generator disconnection and power outages in distribution stations due to flooding are derived. A spatiotemporal differential network is established and the finite difference method with a total variation non-increasing format is used to solve the problem, and the probability of wind and solar generator disconnection and power outages in distribution stations is quantitatively evaluated.

[0203] 3. A method for locating and predicting power outages in distribution substations driven by the collaboration of multi-source data and models is proposed. The short-term rainfall forecast data is input into the accident probability density calculation model to predict the risk probability of power outages in distribution substations. By mining the correlation between the real-time monitoring data of the integrated space-ground and air-space and the prediction error of the risk model, a real-time correction strategy for risk probability prediction driven by multi-source data based on a large language model is proposed. The multi-source monitoring data of the space-ground and air-space, such as meteorological satellites, rainfall radars, and rain gauges, is used as input, and the prediction error of the risk model is used as output. The data-driven method is used to perform rolling correction on the model risk distribution results to further improve the accuracy of power outage prediction.

[0204] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for predicting the situation of rainstorm disaster accidents with risk dependence between urban distribution network and drainage network, characterized in that: The following steps are involved: S1. Acquisition The relevant data of urban distribution network and drainage network are used to solve the distribution of urban waterlogging under heavy rain weather. Based on the longitude and latitude information of power equipment, the solved distribution of urban waterlogging is mapped to the flooding depth of power equipment. S2. Determine whether a large-scale power outage will occur based on the critical conditions that cause a large-scale power outage; wherein, a multi-time-scale power-drainage network fault evolution model is established to quantify the critical conditions that cause a large-scale power outage; S3. Consider the accidents that may cause urban flooding and power outages The power outage probability density calculation model is used to predict the occurrence of power outage accidents. probability; The specific contents of constructing the power outage probability density calculation model in S3 include: Obtained from the probability space of uncertain parameters Representative discrete points , establish a power outage accident probability density calculation model, solve the power outage accident probability density calculation model to obtain the joint probability density function of each representative discrete point , and indivual The cumulative summation results in the power outage accident The probability density evolution expression of is used to quantify the power outage risk; the power outage accident probability density calculation model is as follows: (1) in, Represents a representative discrete point Power outage With uncertain parameters The joint probability density function between is the time variable, Indicates A power outage at a distribution network node Indicates the network configuration node number. Indicates the total number of distribution network nodes. express Derivative with respect to time.

2. The method for predicting the situation of rainstorm disaster accidents with risk dependence of urban distribution network and drainage network according to claim 1 is characterized in that: The specific contents of S1 include: S11. Construct a one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model, including a two-dimensional urban surface hydrological and hydrodynamic model and a one-dimensional hydrodynamic model of the drainage network; Among them, the two-dimensional hydrological and hydrodynamic model of the urban surface is: (2) (3) (4) (5) (6) In the formula, Indicates time, and Respectively represent the longitude and latitude of the city’s geographical coordinates; represents the conservation vector, and Respectively and The flux vector in the direction, and They represent the surface slope source term, ground friction source term and overflow source term respectively; For water depth, and They are and The flow velocity value in the direction, is the acceleration due to gravity, is the terrain elevation, is the Manning roughness coefficient; and Respectively indicate along Direction and The bottom slope source term in the direction of and Respectively indicate along and The friction source term in the direction; is the rainfall, is the flow rate into the drainage well, is the pumping flow rate of the pump station, is the evaporation of rainwater, is the amount of water that infiltrates into the soil; The one-dimensional hydrodynamic model of the drainage network is: (7) In the formula, is the cross-sectional area of ​​water in the drainage pipe, is the cross-sectional flow rate, The amount of water discharged into the river through the drainage pipe. is the pipeline water depth, and are the bottom slope and friction slope of the drainage pipe, Indicates the spatial position of the pipeline; S12. Solve for the flow rate into the drainage well and pumping station discharge flow , providing parameters for solving the one-dimensional and two-dimensional coupled waterlogging hydrological and hydrodynamic model: The urban drainage network is coupled through flow exchange between the drainage well and the surface, and the flow into the drainage well is solved by the one-dimensional hydrodynamic model of the drainage network. ; Pumping station discharge flow The solution: (8) in, Indicates the active power consumed by the pump station, is the energy loss coefficient of the pumping station, is the water density, is the static head of the pump; S13. Aiming at the hydrological and hydrodynamic model of urban waterlogging, the regional calculation unit is simulated by arbitrary triangular unstructured grid division, and the Godunov format is used to construct the Riemann problem to solve the numerical flux of the water flow grid microelement. The second-order spatial reconstruction of the MUSCL-TVD format and the second-order discretization of the time direction of the Runge-Kutta format are performed to obtain the distribution of water depth of urban waterlogging under heavy rain weather, and the water depth of urban waterlogging is mapped to the flooding depth of power equipment based on the latitude and longitude information of power equipment.

3. The method for predicting the situation of rainstorm-induced disaster accidents with risk dependence of urban distribution network and drainage network according to claim 1 is characterized in that: The specific contents of "Quantifying the critical conditions that cause large-scale power outages by establishing a multi-time scale power-drainage network fault evolution model" in S2 include: S21. Determine whether there is equipment failure based on whether the submerged depth of the power equipment reaches the flood prevention height; S22. If a fault occurs, jump to S23 to perform distribution network transient analysis. If no fault occurs, jump to S24 to perform distribution network steady-state analysis. S23. Select a short-time transient analysis scale and perform transient simulation on the distribution network fault, determine the transient fault according to the voltage waveform in the simulation result, and urgently remove the existing transient fault; S24. Select the long-term steady-state analysis scale, calculate the dynamic flow of the urban distribution network, and optimize the flow for the results with faults: if the flow does not converge, remove the corresponding fault and re-perform S23 and S24; if the flow converges, update the urban distribution network topology and the drainage equipment to pump out the accumulated water power, and output The load loss at the moment and enter S25; S25. Determine whether the rainstorm has stopped at the current moment: If the rainstorm is still continuing, The power grid and drainage equipment data are updated at all times as At the initial condition, repeat S21-S25; if the rainstorm stops, jump to S26 to quantify the critical condition; S26. Draw a total load reduction change graph based on the load loss in each time period, and quantify the critical conditions that trigger urban waterlogging and cascading failures of distribution equipment caused by waterlogging, leading to large-scale power outages.

4. The method for predicting the situation of rainstorm-induced disaster accidents with risk dependence of urban distribution network and drainage network according to claim 1 is characterized in that: Obtained from the probability space of uncertain parameters Representative discrete points The specific contents include: make Represents all uncertain parameters, including the uncertainty of the spatial and temporal distribution of heavy rain , User load uncertainty and uncertainty in wind and solar power generation output : (9) in is the total number of random variables in the electricity-water risk dependent network, is the number of uncertainty parameters of the temporal and spatial distribution of heavy rain, is the number of user load uncertainty parameters, is the number of uncertainty parameters of wind and solar power generator output, For the The uncertainty parameters of the temporal and spatial distribution of rainstorms, For the User load uncertainty parameters, For the The uncertainty parameters of wind and solar power unit output, Indicates Parameter pairs; In the uncertain parameter space Select a random parameter set as the initial point set: (10) in represents the number of initial point sets, Indicates indivual ; Let the joint probability density function of all uncertain parameters in the initial point set be , then the probability of each initial point is used to calculate its probability in the uncertain parameter space The corresponding volume To describe: (11) Genetic algorithm is used to process the initial point set to ensure Minimize the deviation and obtain Representative discrete points: (12)。 5. The method for predicting the situation of rainstorm disaster accidents with risk dependence of urban distribution network and drainage network according to claim 4 is characterized in that: Uncertainty in temporal and spatial distribution of heavy rain It directly affects the state of the drainage network and indirectly affects the operation of the distribution network through urban waterlogging, which can be expressed as: (13) In the formula, is the number of rainfall samples; and is the bandwidth of the kernel density estimate; and Respectively represent the longitude and latitude of the city’s geographical coordinates; User load uncertainty In line with Gaussian distribution, the output of wind and solar power units is uncertain It conforms to the Weibull distribution, which directly affects the distribution network flow analysis and indirectly affects the drainage network through the shutdown of pump station nodes.

6. The method for predicting the situation of rainstorm-induced disaster accidents with risk dependence of urban distribution network and drainage network according to claim 4 is characterized in that: The specific contents of establishing the probability density calculation model of power outage accidents include: Considering the interactive propagation of cascading faults between the urban distribution network and the drainage network, the urban distribution network and drainage network model are simplified as follows: (14) (15) In the formula, Indicates the area where the urban distribution network is shut down. Representation Node Power outage situation; and are deterministic parameters of the distribution network and drainage network, among which Represents the safe and stable operation parameters of the urban distribution network. including rainfall and river levels; Indicates the waterlogging and outage of power distribution equipment; Indicates that the urban distribution network outage areas are respectively and The corresponding transformation relationship of Indicates the depth of water accumulation in urban waterlogging. express The city is deeply flooded at the location; Represents the power outage and shutdown of drainage equipment. The depth of waterlogging is respectively and The corresponding transformation relationship of Taking into account the urban distribution network, drainage network and fault evolution time factors, an electric-water coupled power system model under heavy rain weather is established: (16) In the formula, Indicates power outage and distribution network outage area , urban waterlogging depth and time Transformation relationship between them; After hiding the deterministic parameters of the distribution network and drainage network, the power outage accident Further simplified expression: (17) Then the component form of a single distribution network node is expressed as: (18) remember time The corresponding evolution region is , then the probability space and state space of the electricity-water risk dependent network are described as follows: (19) (20) In the formula, express The probability of the region corresponding to the evolution over time; , Respectively and The power outage status of the system at all times; yes time Region boundaries of state space; express The overall rate of change over time is expressed as ; is the unit vector on the boundary of the state space region; is the infinitesimal area element on the boundary of the state space; Combining the Gaussian divergence theorem and simplifying the probability space and state space of the electricity-water risk dependent network, the generalized probability density calculation equation for power outage accidents is obtained: (21) In the formula express With uncertain parameters The joint probability density function between Indicates power outage The time derivatives of the components in are denoted by ; The uncertainty parameter probability space is selected Representative discrete points Calculate system parameters in the probability density calculation model of power outage accidents ,Will The corresponding Consider it as a known quantity and directly substitute it into the relevant power outage accident The generalized probability density calculation equation of is used to obtain the probability density calculation model of power outage accidents.

7. The method for predicting the situation of rainstorm-induced disaster accidents with risk dependence of urban distribution network and drainage network according to claim 1 is characterized in that: Solve the power outage probability density calculation model to obtain the joint probability density function of each representative discrete point , and indivual The cumulative summation results in the power outage accident The specific contents of the probability density evolution expression to quantify the power outage risk include: The initial conditions of the power outage probability density calculation model are expressed as: (22) In the formula, is the Dirac function; represents the initial moment; Indicates a deterministic initial blackout event; Indicates uncertain parameters Take representative discrete values The corresponding joint probability density function value when ; Establish the research area and The spatiotemporal difference network is used to solve the probability density calculation model of power outage accidents using the finite difference method with total variation non-increasing format: (23) In the formula, the accident variable and time Respectively and The grid lines are evenly divided, express No. Grid and No. Grid The value at ; and Respectively represent division and The grid line spacing; express and The grid ratio of express Overall rate of change over time abbreviation of; ； ， ； ; The flux limiter is expressed as: (24) in, is a unit step function; is a Roe-Sweby low dissipation flux limiter with ; Satisfy the same rules; By solving the accident probability density calculation model, we can obtain the joint probability density function of each representative discrete point. Finally, the power outage risk is quantified by summing up all the situations, and the power outage accident is obtained. The probability density evolution expression of is as follows: (25)。 8. The method for predicting the situation of rainstorm-induced disaster accidents with risk dependence of urban distribution network and drainage network according to claim 1 is characterized in that: After building the power outage accident probability density calculation model, it also includes error correction of the model-driven output results, including: (1) Obtain equipment to collect multi-source historical data from the air, space, and ground, obtain the deviation between the model result and the actual probability from the accident probability density calculation model, and pre-process the data; (2) Divide the dataset into a training set for training the neural network model, a validation set for adjusting model parameters and preventing overfitting, and a test set for the final evaluation of model performance; (3) Construct a real-time correction strategy for power outage risk probability prediction based on the large language model (LLM), with multi-source data as input and risk probability prediction error as output: first, construct the prompting project of LLM; second, freeze the parameters of all layers of LLM and embed two learnable neural network modules on this basis. The first neural network module is located between the multi-source meteorological data and the LLM data input interface, and is used to align the input multi-source data with the text data received by LLM. Its output result is connected in series with the prompting project as the LLM text input; the second neural network module is located between the text output of LLM and the risk probability prediction error, and is used to map the LLM output to the time series data of the risk probability prediction error; (4) Simplify the real-time correction strategy of power outage risk probability prediction based on LLM into a deviation learning function , the probability error of power outage risk prediction driven by the model with a future A-step length is predicted by the multi-source monitoring data with a previous H-step length, which is specifically expressed as follows: (26) Where: ; ; express Nodes in Time-varying space-ground multi-source data, writing ,in Indicates Nodes in Moment Multi-source data features, written ; express Nodes in The prediction probability error at time t is written as ; (5) The relevant data of the urban distribution network and drainage network and the short-term rainfall forecast data are input into the power outage risk model to predict the power outage probability. The real-time air-space-ground multi-source monitoring data is input into the trained large model network to obtain the prediction probability error. The model risk distribution results are rolled corrected, and the model parameters are regularly fine-tuned using new data. The rolling correction formula for the power outage probability is expressed as follows: (27) In the formula, and Respectively represent the rolling correction process The power outage probability after time correction and the power outage probability predicted by the model.

Images