Dynamic Prediction Method for Regional Power Grid Disaster Risk under Multiple Secondary Disasters Caused by Rainstorms

Abstract

This application provides a method for dynamically predicting the disaster risk of regional power grids under multiple secondary disasters caused by rainstorms, including: constructing a failure mechanism model of power grid equipment under secondary disasters caused by rainstorms based on a global evolution equation of probability density; determining a dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid based on the failure mechanism model of power grid equipment under secondary disasters caused by rainstorms; and determining the load loss of each node in the regional power grid based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid. The method proposed in this application can improve the dynamism, precision, and engineering applicability of regional power grid disaster risk prediction under secondary disasters caused by rainstorms.

Description

Technical Field

[0001] This application relates to the field of regional power grid disaster risk prediction, and in particular to a dynamic prediction method for regional power grid disaster risk under multiple secondary disasters caused by rainstorms. Background Technology

[0002] Continuous torrential rains easily trigger multiple secondary disasters such as landslides, mudslides, flash floods, and urban flooding, exhibiting significant multi-source concurrent and chain-triggered characteristics, posing a severe threat to the safe operation of power distribution network equipment. Specifically, in mountainous and hilly areas, rainwater infiltration leads to saturation and instability of the soil and rock, inducing landslides and causing deformation or even collapse of power poles. In valley and river areas, the rapid convergence of torrential rains forms sudden mudslides or flash floods, causing pole deformation failure or substation flooding, resulting in regional power outages. In low-lying urban areas, urban flooding caused by rainfall causes waterlogging and failure of power distribution equipment, which may then lead to power outages in distribution areas. The multiple secondary disasters caused by these torrential rains have heterogeneous disaster-causing mechanisms and overlapping impact ranges on the power grid equipment, making the power grid disaster risk dynamically evolve in time and spatially differentiated, and cascading from the failure of distribution network equipment to power outages at the power grid system level. Existing methods for assessing power grid risk under secondary disasters caused by rainstorms mostly focus on landslides, lacking systematic research on the risk of power grid disasters when multiple secondary disasters such as landslides, debris flows, flash floods, and urban flooding occur simultaneously due to rainstorms. Furthermore, the research objects are mostly limited to transmission networks, failing to address the risk interaction and transmission between multi-level power grids under multiple secondary disasters caused by rainstorms. In terms of assessment framework, most methods employ static analysis, that is, assessing power grid disaster risk indicators based on pre-disaster meteorological forecast data, without fully considering the temporal evolution of power grid disaster risk under dynamic changes in the intensity of secondary disasters caused by rainfall. Summary of the Invention

[0003] This application provides a method for dynamic prediction of regional power grid disaster risk under multiple secondary disasters caused by rainstorms. To solve the above-mentioned technical problems, this application adopts the following technical methods: This application provides a method for dynamic prediction of regional power grid disaster risk under multiple secondary disasters caused by rainstorms, including: Based on the global evolution equation of probability density, a failure mechanism model of power grid equipment under secondary disasters caused by rainstorms is constructed; Based on the aforementioned power grid equipment failure mechanism model under secondary rainstorm disasters, a dynamic prediction curve for the failure probability of disaster-bearing equipment in the regional power grid is determined. Based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid, the load loss of each node in the regional power grid is determined.

[0004] Optionally, the failure mechanism model of power grid equipment under secondary disasters caused by rainstorms includes: Models for the deformation and failure mechanisms of power transmission towers under rainstorm and landslide disasters, models for the deformation and failure mechanisms of power transmission towers under rainstorm and debris flow disasters, models for the water immersion failure mechanisms of substations under rainstorm and flash flood disasters, and models for the waterlogging failure mechanisms of regional power distribution equipment under rainstorm and urban flooding disasters.

[0005] Optionally, determining the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid based on the power grid equipment failure mechanism model under the secondary disaster of rainstorm includes: Based on the failure mechanism model of power grid equipment under secondary disasters caused by rainstorms, the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of power grid equipment bearing disaster are calculated. Based on the failure mechanism model of power grid equipment under the secondary disaster of rainstorm, a global evolution equation of the failure risk probability density of power grid equipment under disaster is constructed. The intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment are used as inputs to the probabilistic constrained neural network. The global evolution equation of the probability density of the failure risk of the power grid disaster-bearing equipment is used as a probabilistic constraint penalty term and embedded into the total loss function of the probabilistic constrained neural network. The probabilistic constrained neural network is trained to obtain a trained probabilistic constrained neural network. The intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the disaster-bearing equipment in the power grid are substituted into a trained probabilistic constrained neural network for solution to determine the dynamic prediction curve of the failure probability of the disaster-bearing equipment in the regional power grid.

[0006] Optionally, the step of substituting the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment into a trained probabilistic constrained neural network for solution, to determine the dynamic prediction curve of the failure probability of the regional power grid disaster-bearing equipment; includes: Substitute the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment into the trained probability constraint neural network to generate a probability prediction surface for the failure risk of the regional power grid disaster-bearing equipment under secondary rainstorm disasters. The failure probability prediction surface of the regional power grid equipment under the secondary disaster of rainstorm is integrated within the failure domain to determine the dynamic prediction curve of the failure probability of the regional power grid equipment.

[0007] Optionally, determining the load loss at each node of the regional power grid based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid includes: A series reliability model analysis was performed on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the power grid of the region to determine the failure probability of distribution network nodes under different voltage levels. Based on the failure probability of distribution network nodes under different voltage levels, the load loss of each node in the regional power grid is determined.

[0008] Optionally, determining the load shedding of each node in the regional power grid based on the failure probability of distribution network nodes under different voltage levels includes: Introduce a random variable that follows a uniform distribution; Sampling is performed from the distribution of the random variable to determine the random number obtained from the current sample; Based on the random number and the failure probability of distribution network nodes under different voltage levels, the disaster-affected and failed power grid nodes are determined. Based on the affected and failed power grid nodes, the load loss of each node in the regional power grid is determined.

[0009] Optionally, determining the load loss of each node in the regional power grid based on the affected and failed power grid nodes includes: Based on the affected and failed power grid nodes, update the distribution network topology; Based on the updated distribution network topology, the load loss of each node in the regional power grid is determined.

[0010] Optionally, determining the load shedding at each node of the regional power grid based on the updated distribution network topology includes: Based on the updated distribution network topology, the indirect load shedding curves between distribution network nodes under different voltage levels in different fault scenarios are calculated. Based on the indirect load loss curves of distribution network nodes under different voltage levels, the intrinsic drift coefficient and intrinsic diffusion coefficient of indirect load loss of regional power grid nodes are calculated. Substitute the intrinsic drift coefficient and intrinsic diffusion coefficient of the indirect load loss of the regional power grid nodes into the trained probabilistic constrained neural network for solving to obtain the probability density prediction surface of the indirect load loss of the regional power grid nodes. Based on the probability density prediction surface of indirect load loss at the nodes of the regional power grid, the load loss of each node in the regional power grid is determined.

[0011] Optionally, determining the load loss at each node of the regional power grid based on the probability density prediction surface for indirect load loss between regional power grid nodes includes: The mathematical expectation is obtained by integrating the indirect load loss probability density prediction surface of the regional power grid node along the indirect load loss direction, thus obtaining the indirect load loss of the regional power grid node at the corresponding time. Obtain the load demand of regional power grid nodes at the corresponding time. Based on the indirect load loss, the load demand, and the failure probability of distribution network nodes under different voltage levels, the load loss of each node in the regional power grid is determined.

[0012] This application has the following beneficial effects: The method proposed in this application can improve the dynamism, precision, and engineering applicability of regional power grid disaster risk prediction under secondary rainstorm disasters. Attached Figure Description

[0013] Figure 1 A flowchart illustrating the dynamic prediction method for regional power grid disaster risk under multiple secondary disasters caused by rainstorms, provided in this application embodiment; Figure 2 This is a comparison chart of regional power grid load shedding curves under different methods provided in the embodiments of this application; Figure 2 (a) Comparison of load shedding curves of 110kV distribution network under different methods; Figure 2 (b) Comparison of load shedding curves of 35kV distribution network under different methods; Figure 2 (c) is a comparison of the load shedding curves of the 10kV distribution network under different methods. Detailed Implementation

[0014] To facilitate understanding by those skilled in the art, the present application will be further described below in conjunction with embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present application.

[0015] To solve the above technical problems, such as Figure 1 As shown, this application proposes a dynamic prediction method for regional power grid disaster risk under multiple secondary disasters caused by rainstorms, including: Step S101: Based on the global evolution equation of probability density, construct a failure mechanism model of power grid equipment under secondary disasters caused by rainstorms; A model of power grid equipment failure mechanism under secondary disasters caused by rainstorms is established based on the global evolution equation of probability density, revealing the hierarchical probability propagation relationship between rainstorm uncertainty, the probability of secondary disaster occurrence, and the failure risk of power grid equipment affected by disasters. The uncertainty of rainstorm intensity and spatiotemporal distribution easily leads to significant deviations between actual and forecasted rainfall in local areas, thus causing the failure risk of power grid equipment to be affected by secondary disasters. Rainfall forecast for the day is similar to the previous one. The first day's actual rainfall, weighted by time decay, is obtained. Daily cumulative effective rainfall There is a discrepancy. (Number) Daily cumulative effective rainfall The expression is as follows: (1) In the formula, The time decay factor characterizes the first... The weight of the contribution of rainfall to the current slope stability; For the first The random variable for precipitation forecast error under short-term precipitation forecast is described by a Logistic distribution. These are the location and scale parameters of the Logistic distribution, respectively. , The first Heaven and the Di Short-term rainfall forecast. (Regarding rainfall forecast errors) Random sampling can yield multiple sets of cumulative effective rainfall data representing the spatiotemporal uncertainty of rainstorms under different rainfall scenarios, which can be further used for the analysis of failure mechanism models of power grid equipment under secondary disasters caused by rainstorms.

[0016] Power grid disaster-resistant equipment failure risk The failure risk of equipment under secondary disasters such as landslides, debris flows, flash floods, and urban flooding can be addressed by establishing the following failure mechanism models: (1) Deformation failure mechanism model of power line towers under rainstorm and landslide disasters The relationship between the probability of landslide occurrence and effective rainfall can be fitted using a Weibull distribution, and the probability of tower deformation caused by landslides can be characterized using a normal distribution. This applies to tower deformation under rainstorm-induced landslide disasters. exist Failure probability at time step The probability density function of landslide occurrence based on effective rainfall and the probability density function of tower deformation under landslide impact are combined and integrated. The expression is as follows: (2) In the formula, and The scale and shape parameters of the Weibull distribution are determined by the maximum likelihood estimation method. The deflection of the transmission tower is calculated based on cantilever beam theory. The mean value is a normal distribution, and is usually taken as the deflection at the midpoint of the elastic cantilever beam; The standard deviation of the normal distribution is usually taken as . ,in This refers to the maximum deflection limit of transmission towers. The coefficient of variation depends on the geological environment and rainfall conditions of the transmission tower.

[0017] (2) Deformation and failure mechanism model of power line towers under rainstorm and debris flow disaster The impact intensity of debris flow can be calculated using debris flow flow estimation models, Manning's formula, continuity equations, and fluid dynamics theory. The relationship between tower failure probability and debris flow impact intensity can be fitted using a modified exponential function, and its overall expression is as follows: (3) In the formula, Tower impacted by mudslide exist The probability of failure at any given time; To improve the exponential function, correction coefficients are used; The shape parameter determines the steepness of the curve; For towers Bending capacity; The width of the cross-section of the debris flow channel; The debris flow impact coefficient is given by the standard "Design Code for Debris Flow Prevention Engineering". The recommended values ​​are 1.47 for the lattice section of the iron tower and 1.33 for the circular section of the cement pole. Debris flow density; The width of the tower facing the current; The angle between the stress-bearing surface of the tower and the impact direction of the debris flow; This is the peak runoff coefficient, and its value generally ranges from 0.4 to 0.9. The drainage area; The specific weight of the mud-rock fluid; The specific gravity of water; This refers to the unit weight of solid particles in the soil. The blockage coefficient of debris flow; The duration of the confluence of the river basin.

[0018] (3) Mechanism model of substation water immersion failure under rainstorm and flash flood disaster The peak flow of flash floods can be determined based on inference formulas. The substation overtopping flow can be calculated by combining the continuity equation and the broad-crested weir formula. The substation inundation depth can be further obtained based on the water balance principle. On this basis, the relationship between the substation outage probability and the substation inundation depth can be described by a vulnerability curve, and its overall expression is as follows: (4) In the formula, For substation exist The probability of failure at any given moment; The equation for fitting the vulnerability curve related to the substation flooding depth; for Time Substation The depth of flooding within the station; For time step; For substation The site area; It is the acceleration due to gravity; The flow coefficient is generally taken as... ; The width of the top edge; The height of the flood control wall; This indicates that overtopping occurs only when the depth of the mountain flood exceeds the height of the flood control wall; This is the roughness coefficient of the river channel, with a value ranging from 0.03 to 0.07 for mountainous river channels; The width of the cross-section of a mountain river channel; The longitudinal gradient of the riverbed is as follows.

[0019] (4) Mechanism model of power distribution equipment failure due to flooding in areas affected by rainstorm and urban flooding The depth of urban flooding can be obtained by jointly solving the two-dimensional shallow water equations of the urban surface and the one-dimensional Saint-Venant equations of the underground drainage network under different rainstorm scenarios. The mapping relationship between flood-induced failure of power distribution equipment and the depth of flooding can be established using an exponential function fitting method as follows: (5) In the formula, Probability of power distribution equipment failure; For power distribution equipment exist The depth of the floodwaters at any moment; For power distribution equipment Waterproofing elevation; The height of the cable connector; These are the attenuation coefficient and damping coefficient of the fitted exponential function, respectively. The attenuation factor determines the steepness of the failure probability curve, while the damping factor affects the overall shape and horizontal position of the curve. They need to be calibrated according to the protection parameters of various power distribution equipment.

[0020] Step S102: Based on the failure mechanism model of power grid equipment under secondary disasters caused by rainstorms, determine the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid; The failure risk of power grid equipment under four types of secondary disasters caused by rainstorms can be simplified as follows: The expression is as follows: (6) In the formula, M This represents the collection of all power grid equipment susceptible to damage under four types of secondary disasters caused by rainstorms. These represent the sets of power grid equipment that are affected by secondary disasters such as landslides, mudslides, flash floods, and urban flooding.

[0021] The uncertainty of torrential rain leads to uncertainties in landslide occurrence, debris flow impact intensity, flash flood peak flow, and urban flooding depth, further resulting in uncertainties in the failure of power line towers, substations, and distribution equipment. Therefore, the regional power-water interdependent network under secondary rainstorm disasters can be regarded as a multidimensional stochastic dynamic system. Furthermore, the development process of secondary disasters and the response of affected equipment evolve continuously over time; thus, the failure risk of any power grid equipment affected by the disaster can be considered as a one-dimensional continuous stochastic response process of the system. Since the generalized probability density evolution equation based on point evolution has insufficient tail risk characterization, this application can use a global probability density evolution equation based on swarm evolution to establish a nonlinear mapping relationship between the intensity and probability of secondary disasters and the propagation of failure risk to distribution equipment. This considers the tail risk of equipment failure under secondary rainstorm disasters and accurately characterizes the dynamic evolution process of the failure risk of affected equipment over time.

[0022] Power grid disaster protection equipment Failure risk probability density Regarding time The partial derivatives satisfy the Kramers–Moyal expansion, as shown below: (7) (8) In the formula, Risk of power grid failure of Conditional derived moments; for At any moment The specific values ​​that can be taken are as follows. Existing research has proven that a one-dimensional continuous random process... If the transition probability density satisfies the Lindeberg condition, then its third-order and higher-order conditional derivative moments are all zero. Therefore, the Kramers–Moyal expansion expression only has the first two terms remaining. Let: (9) The global evolution equation for the probability density of power grid disaster-bearing equipment failure can then be obtained: (10) In the formula, These are the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of power grid disaster-bearing equipment. Based on the failure mechanism model of power grid equipment under rainstorm, landslide, debris flow, flash flood and waterlogging disasters, Monte Carlo sampling is performed to obtain the power grid disaster-bearing equipment. At various times failure probability Substituting into formula (9), the intrinsic drift coefficient of the failure risk of power grid disaster-bearing equipment is calculated. With intrinsic diffusion coefficient .

[0023] To address the failure risk of the aforementioned power grid equipment prone to disasters, this step proposes a fast algorithm for solving the failure risk of regional power grid equipment based on a probabilistic constrained neural network. The solution process is as follows: The failure probability of disaster-prone equipment in the regional power grid They are uniformly denoted as state variables. Intrinsic drift coefficient of power grid disaster-bearing equipment failure risk With intrinsic diffusion coefficient They are respectively denoted as and The input to the Probabilistic Constrained Neural Network (PCNN) is... The output is an approximate solution to the global evolution equation of the probability density of failure risk of disaster-bearing equipment in the power grid. The conditions that PCNN needs to satisfy to solve GE-PDEE are used as probabilistic constraint penalty terms. The overall loss function for embedding PCNN is expressed as follows: (11) (12) (13) (14) In the formula, The loss function representing PCNN; This represents the probability partial differential residual term of GE-PDEE; The residual term is the initial condition; For probability normalization constraints; , , These represent the weighting coefficients of the loss functions mentioned above; These represent the probability of power grid equipment failure due to disaster at the sampling points for each of the above constraints. These represent the times at the sampling points for each of the above constraints; and These represent the initial time and the initial value of the device failure probability, respectively. and These represent the number of samples for the internal collocation points and the initial condition collocation points of the GE-PDEE, respectively. The number of samples for probability conservation constraint points; the standard deviation parameter of the Gaussian distribution is set to... .

[0024] The specific process of the solution algorithm is as follows: (1) Input the intrinsic drift coefficient of power grid disaster-bearing equipment failure risk in representative scenarios. With intrinsic diffusion coefficient Initial equipment failure probability Learning rate and attenuation rate .

[0025] (2) Using the Latin hypercube sampling method in the solution domain { Generate training collocations within the range} and normalize the coordinates of the collocations to the interval [0, 1].

[0026] (3) Initialize PCNN using the Xavier method. .

[0027] (4) Input the normalized collocation samples into PCNN and use automatic differentiation to calculate the output of PCNN. For state variables and time partial derivatives and .

[0028] (5) Calculate the total loss function according to equations (11)-(14). .

[0029] (6) Calculate the gradient of the loss function with respect to the network parameters through backpropagation, update the learnable parameters of PCNN, adjust the weights of the data-driven terms according to the exponential decay strategy, and determine the trained PCNN.

[0030] Will Input the trained PCNN to obtain the prediction surface for the failure probability of power grid equipment in the region under secondary rainstorm disasters. Finally, regarding Integrating within the failure domain yields the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid. .

[0031] Step S103: Based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid, determine the load loss of each node in the regional power grid.

[0032] For power grids in regions with different voltage levels, this step constructs a multi-level vertically interconnected topology model of 110kV, 35kV, and 10kV distribution networks. Each level of the distribution network typically exhibits a radial topology, with feeders connected via normally open tie switches, providing load transfer capabilities. The node sets of the 110kV, 35kV, and 10kV distribution networks are defined as follows: , , The output busbars of the 110kV / 35kV substation and the 35kV / 10kV substation are defined as H&M interface and M&L interface, respectively. The node sets of H&M and M&L interfaces are denoted as follows: , .

[0033] A series reliability model analysis was conducted on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid. Determine the failure probability of distribution network nodes under different voltage levels : (15) In the formula, Represented as nodes A collection of electrical grid equipment that supplies power.

[0034] Secondary disasters caused by torrential rain led to the failure of power grid equipment. The faults propagated hierarchically along the 110kV, 35kV, and 10kV distribution network, ultimately affecting the normal power supply to users. For distribution network nodes At any moment State variables, nodes When it fails ,node During normal operation Introduce a random variable that follows a uniform distribution. ,from Sampling is performed from the distribution, and if the random number obtained from the current sample is... Less than the current time Node failure probability When a power grid node is deemed to have failed due to a disaster, that is... Conversely, it operates normally. Each sampled grid node failure scenario is different. To ensure reliable power supply to critical grid nodes, load transfer and distribution network topology reconfiguration are required, updating the distribution network topology and thus affecting the distribution network power flow calculation results. The sampled grid node failure scenarios obtained here will be substituted into the following multi-level load shedding interaction iterative calculation process of the grid.

[0035] Because the loss of load at various levels of the power grid leads to an imbalance in the original power supply and demand relationship, it is difficult to achieve power supply and demand balance in a multi-level power grid through a single top-down recursive solution. This application proposes an interactive iterative calculation method for multi-level power grid load loss under disasters such as landslides, mudslides, flash floods, and urban flooding caused by rainstorms, to calculate the indirect load loss curves of regional power grid nodes under different fault scenarios. The specific steps of the interactive iterative calculation method for multi-level power grid load loss are as follows: (1) Set the number of iterations k=1, and the maximum number of iterations. Initialize the expected boundary power values ​​for the H&M and M&L interfaces: (16) (17) In the formula, , These are the initial values ​​for the active power transmitted at the H&M and M&L nodes, respectively. , These are the expected values ​​for the active power transmitted at the H&M and M&L nodes, respectively. , These are the initial values ​​for the reactive power transmitted at the H&M and M&L nodes, respectively. , These are the expected values ​​for reactive power transmitted at the H&M and M&L nodes, respectively.

[0036] (2) Based on the active and reactive power of the H&M interface , With the goal of minimizing the load shedding of the 110kV distribution network, considering load transfer and topology reconfiguration of the 110kV distribution network, the power flow of the 110kV distribution network is solved to obtain the operating boundary of the transmission from the H&M interface to the downstream 35kV distribution network: (18) (3) Based on the active and reactive power of the M&L interface , With the goal of minimizing the load shedding of the 35kV distribution network, considering load transfer and topology reconfiguration of the 35kV distribution network, the power flow of the 35kV distribution network is solved to obtain the power flow of each 35kV distribution network area. Actual active and reactive power requirements obtained from the H&M interface. and And the operational boundary of M&L interface transmission to the lower-level 10kV distribution network: (19) (4) With the goal of minimizing the load loss of the 10kV distribution network, considering load transfer and topology reconfiguration of the 10kV distribution network, the power flow of the 10kV distribution network is solved to obtain the power flow of each 10kV distribution network area. Actual active and reactive power requirements obtained from the M&L interface. and .

[0037] (5) Check the convergence conditions. Set the iterative convergence to meet the following conditions: Equation (20) is the supply and demand imbalance of the H&M interface, and Equation (21) is the supply and demand imbalance of the M&L interface. If all conditions are met, the iterative convergence is achieved, and proceed to step 7; otherwise, proceed to step 6.

[0038] (20) (twenty one) In the formula, , The first The difference between the supply and demand of active power at H&M and M&L nodes in the next iteration; , The first The difference between the supply and demand of reactive power at H&M and M&L nodes in the next iteration; , These represent the convergence thresholds for active power and reactive power, respectively.

[0039] (6) Update the boundary power of the H&M and M&L interfaces, and update the expected value of the boundary power for the next iteration based on the supply and demand deviation of this iteration: (twenty two) (twenty three) In the formula, This represents the number of iterations. , These are the update rates of the active and reactive power boundaries of the H&M and M&L nodes, respectively, used to control the step size of the boundary update, and their values ​​range from [0, 1].

[0040] make .like If the iteration terminates, proceed to step (7); otherwise, return to step (2).

[0041] (7) Obtain the indirect load shedding curves of distribution network nodes at different voltage levels (110kV, 35kV, 10kV) under different fault scenarios, i.e., distribution network node At different times The indirect load loss and the corresponding initial node load loss.

[0042] Referring to the contents of formulas (7) to (9), the regional power grid nodes Indirect loss of load probability density Regarding time The partial derivatives satisfy the Kramers–Moyal expansion. Previous studies have shown that the transition probability density of a one-dimensional continuous stochastic process satisfies the Lindeberg condition, meaning its third-order and higher-order conditional derivative moments are all zero. Therefore, the Kramers–Moyal expansion expression only contains the first two terms, which are the intrinsic drift coefficients of load shedding between nodes in the regional power grid. With intrinsic diffusion coefficient : (twenty four) The load shedding curves between nodes in 110kV, 35kV, and 10kV distribution networks under different fault scenarios, i.e., distribution network nodes At different times Substituting the indirect load shedding into formula (24), the intrinsic drift coefficient of indirect load shedding at the regional power grid node can be obtained. With intrinsic diffusion coefficient .

[0043] Intrinsic drift coefficient of indirect load shedding in regional power grid nodes With intrinsic diffusion coefficient The global evolution equation for the probability density of load shedding in the regional power grid can be obtained as follows: (25) in, and These are the intrinsic drift coefficient and intrinsic diffusion coefficient of the regional power grid node load shedding, respectively. For regional power grid nodes Unload evolution process The probability density; for At any moment The specific value to be taken.

[0044] Substituting the intrinsic drift coefficient and intrinsic diffusion coefficient of the inter-node load shedding in the regional power grid into the aforementioned trained probabilistic constrained neural network, the global evolution equation of the regional power grid load shedding probability density is solved, i.e., the state variables are obtained. The probability of indirect load shedding at regional power grid nodes The input to PCNN is... Become After solving, the probability density prediction surface for indirect load shedding between regional power grid nodes is obtained. ,right Indirect loss of load The expected value of the directional integral is used to obtain the regional power grid node. exist Indirect loss of load at any moment: (26) Set up regional power grid nodes exist The load demand at any given time is Combining the failure probability of distribution network nodes under different voltage levels and indirect loss of load The load shedding at each node of the regional power grid can be obtained. as follows: (27) Simulation Experiment To verify the effectiveness of the proposed method, based on the topologies of the IEEE-13 node distribution network, the IEEE-34 node distribution network, and the IEEE-33 node distribution network, the reference voltages were equivalently set to 110kV, 35kV, and 10kV, respectively, forming a simulation example system for a multi-level distribution network including 110kV, 35kV, and 10kV. To verify the accuracy and effectiveness of the proposed method in the dynamic prediction of regional power grid disaster risk under multiple secondary disasters caused by rainstorms, three comparative methods were introduced. Method 1 is the method proposed in this application; Method 2 is a dynamic prediction method for power grid disaster risk based on SMC, which calculates the failure probability of all disaster-bearing equipment in the power grid through a mechanistic model, and further samples and evaluates the multi-level load shedding of the power grid as an accurate result of the power grid disaster risk to evaluate the accuracy of the proposed method; Method 3, based on the method proposed in this application, only performs static risk assessment based on pre-disaster rainfall forecasts and the initial operating state of the power grid, ignoring the real-time rolling updates of actual measurements and rainfall forecasts. The regional power grid load shedding curves under secondary disasters caused by rainstorms obtained by the three methods are shown below. Figure 2 As shown in Table 1, the expected errors in grid load loss, power shortage, and calculation time at the end of the rainstorm predicted by different methods are compared.

[0045] Table 1. Comparison of errors and calculation times in dynamic prediction of regional power grid risks using different methods. ; Depend on Figure 2 As shown in Table 1, the power grid load loss curves of Method 1 and Method 2 basically overlap. The relative deviations of the load loss in the 110kV, 35kV, and 10kV distribution networks are 2.45%, 2.71%, and 3.05%, respectively, and the relative deviations of the expected power shortage are 3.05%, 3.37%, and 3.32%, respectively. All errors are controlled within 4%, verifying that the proposed method can consider equipment failure and the tail risk of regional power grid node disasters, and accurately predict the regional power grid load loss under multiple secondary disasters caused by rainstorms. In contrast, Method 3, because it only relies on the pre-disaster deterministic rainfall forecast and the initial operating state of the power grid for static assessment and does not consider the real-time rolling updates of measured and forecasted rainfall, has a significantly larger risk assessment error.

[0046] In summary, the method proposed in this application can achieve dynamic prediction of power grid disaster risk under multiple secondary disasters caused by rainstorms, fill the gap in existing methods that do not consider the power grid disaster risk when secondary disasters occur concurrently with rainstorms, solve the problem that existing power grid disaster risk assessments rely heavily on static meteorological forecast data and are difficult to describe the dynamic changes in the intensity of secondary disasters caused by rainstorms, and can improve the dynamism, precision and engineering applicability of regional power grid disaster risk prediction under secondary disasters caused by rainstorms.

[0047] In some embodiments, this application also provides a computer system including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0048] This application also provides a computer-readable storage medium for storing a computer program. This computer-readable storage medium can be applied to a computer device, and the computer program causes the computer device to execute the corresponding processes in the methods described above in the embodiments of this application; for brevity, further details are omitted here.

[0049] The above embodiments are preferred implementations of this application. In addition, this application can be implemented in other ways. Any obvious substitutions without departing from the concept of this technical solution are within the protection scope of this application.

[0050] To facilitate understanding by those skilled in the art of the improvements made by this application compared to the prior art, some of the accompanying drawings and descriptions have been simplified, and for clarity, some other elements have been omitted from this application. Those skilled in the art should realize that these omitted elements may also constitute the content of this application.

Claims

1. A method for dynamically predicting the catastrophic risk of an area power grid under heavy rain multiple secondary disasters, characterized in that, include: Based on the global evolution equation of probability density, a failure mechanism model of power grid equipment under secondary disasters caused by rainstorms is constructed; Based on the aforementioned power grid equipment failure mechanism model under secondary rainstorm disasters, a dynamic prediction curve for the failure probability of disaster-bearing equipment in the regional power grid is determined. Based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid, the load loss of each node in the regional power grid is determined.

2. The method according to claim 1, characterized in that, The failure mechanism model of power grid equipment under secondary disasters caused by rainstorms includes: Models for the deformation and failure mechanisms of power transmission towers under rainstorm and landslide disasters, models for the deformation and failure mechanisms of power transmission towers under rainstorm and debris flow disasters, models for the water immersion failure mechanisms of substations under rainstorm and flash flood disasters, and models for the waterlogging failure mechanisms of regional power distribution equipment under rainstorm and urban flooding disasters.

3. The method according to claim 1, characterized in that, The method for determining the dynamic prediction curve of the failure probability of power grid equipment under secondary rainstorm disasters based on the power grid equipment failure mechanism model includes: Based on the failure mechanism model of power grid equipment under secondary disasters caused by rainstorms, the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of power grid equipment bearing disaster are calculated. Based on the failure mechanism model of power grid equipment under the secondary disaster of rainstorm, a global evolution equation of the failure risk probability density of power grid equipment under disaster is constructed. The intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment are used as inputs to the probabilistic constrained neural network. The global evolution equation of the probability density of the failure risk of the power grid disaster-bearing equipment is used as a probabilistic constraint penalty term and embedded into the total loss function of the probabilistic constrained neural network. The probabilistic constrained neural network is trained to obtain a trained probabilistic constrained neural network. The intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment are substituted into a trained probabilistic constrained neural network for solution to determine the dynamic prediction curve of the failure probability of the regional power grid disaster-bearing equipment.

4. The method according to claim 3, characterized in that, The step of substituting the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment into a trained probabilistic constrained neural network for solution, and determining the dynamic prediction curve of the failure probability of the regional power grid disaster-bearing equipment; includes: Substituting the intrinsic drift coefficient and intrinsic diffusion coefficient of the failure risk of the power grid disaster-bearing equipment into the trained probability constraint neural network, a probability prediction surface for the failure risk of the regional power grid disaster-bearing equipment under secondary rainstorm disasters is generated. The failure probability prediction surface of the regional power grid equipment under the secondary disaster of rainstorm is integrated within the failure domain to determine the dynamic prediction curve of the failure probability of the regional power grid equipment.

5. The method according to claim 4, characterized in that, The determination of load loss at each node of the regional power grid based on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the regional power grid includes: A series reliability model analysis was performed on the dynamic prediction curve of the failure probability of disaster-bearing equipment in the power grid of the region to determine the failure probability of distribution network nodes under different voltage levels. Based on the failure probability of distribution network nodes under different voltage levels, the load loss of each node in the regional power grid is determined.

6. The method according to claim 5, characterized in that, The determination of the load shedding at each node of the regional power grid based on the failure probability of distribution network nodes under different voltage levels includes: Introduce a random variable that follows a uniform distribution; Sampling is performed from the distribution of the random variable to determine the random number obtained from the current sample; Based on the random number and the failure probability of distribution network nodes under different voltage levels, the disaster-affected and failed power grid nodes are determined. Based on the affected and failed power grid nodes, the load loss of each node in the regional power grid is determined.

7. The method according to claim 6, characterized in that, The determination of the load loss at each node of the regional power grid based on the affected and failed power grid nodes includes: Based on the affected and failed power grid nodes, update the distribution network topology; Based on the updated distribution network topology, the load loss of each node in the regional power grid is determined.

8. The method according to claim 7, characterized in that, The determination of load shedding at each node of the regional power grid based on the updated distribution network topology includes: Based on the updated distribution network topology, the indirect load shedding curves between distribution network nodes under different voltage levels in different fault scenarios are calculated. Based on the indirect load loss curves of distribution network nodes under different voltage levels, the intrinsic drift coefficient and intrinsic diffusion coefficient of indirect load loss of regional power grid nodes are calculated. Substitute the intrinsic drift coefficient and intrinsic diffusion coefficient of the indirect load loss of the regional power grid nodes into the trained probabilistic constrained neural network for solving to obtain the probability density prediction surface of the indirect load loss of the regional power grid nodes. Based on the probability density prediction surface of indirect load loss at the nodes of the regional power grid, the load loss of each node in the regional power grid is determined.

9. The method according to claim 8, characterized in that, The determination of the load loss at each node of the regional power grid based on the probability density prediction surface for indirect load loss between regional power grid nodes includes: The mathematical expectation is obtained by integrating the indirect load loss probability density prediction surface of the regional power grid node along the indirect load loss direction, thus obtaining the indirect load loss of the regional power grid node at the corresponding time. Obtain the load demand of regional power grid nodes at the corresponding time. Based on the indirect load loss, the load demand, and the failure probability of distribution network nodes under different voltage levels, the load loss of each node in the regional power grid is determined.

Images