A new energy distribution network resilience grading method under multi-disaster coupling

Abstract

The application provides a new energy power distribution network resilience grading method under multi-disaster coupling, comprising: obtaining a current disaster intensity time sequence; constructing a power distribution network disaster-bearing equipment failure probability calculation model considering risk amplification effect; inputting the current disaster intensity time sequence into the power distribution network disaster-bearing equipment failure probability calculation model for calculation to obtain the failure probability of each equipment at the current time; obtaining a trained conditional flow matching model; inputting the failure probability of each equipment at the current time into the trained conditional flow matching model to generate a predicted power distribution network load loss curve; and determining the resilience grade of the new energy power distribution network based on the power distribution network load loss curve. The method quantitatively characterizes the equipment failure risk amplification effect under the concurrent occurrence of strong wind and icing and the lightning-mountain fire chain, improves the generalization ability of the power distribution network load loss curve prediction under small samples, realizes resilience grading and reinforcement measure matching, supports hierarchical control, differentiated disaster prevention and emergency recovery decision-making.

Description

Technical Field

[0001] This application relates to the field of distribution network resilience and risk assessment technology, and in particular to a method for rating the resilience of new energy distribution networks under the coupled effects of multiple disasters. Background Technology

[0002] With the increasing frequency of extreme weather events and the high proportion of distributed renewable energy integration, renewable energy distribution networks face complex challenges in disaster scenarios, including the coupled impact of multiple disasters, the random evolution of equipment failure states, and the dynamic changes in load loss processes. The failure risk of disaster-bearing equipment such as lines and transformers under the combined effects of strong winds and icing, and lightning and wildfire chains, is becoming increasingly prominent. Against this backdrop, accurately assessing the resilience level of renewable energy distribution networks under the coupled effects of multiple disasters has become a key issue supporting the overall defense, tiered management, and differentiated recovery of renewable energy distribution networks.

[0003] Existing disaster risk assessment methods for power distribution networks mainly rely on single-hazard vulnerability curves, fault probability statistical models, Monte Carlo sampling, and post-disaster recovery evaluation models. These methods can provide some analysis of equipment failure risk and changes in system power supply capacity under the influence of a single disaster. However, due to the complexity of the multi-hazard coupling mechanism, inconsistent equipment failure criteria, and the time-varying nature of disaster intensity, existing methods struggle to characterize the amplification effect of disaster-bearing equipment failure risk in new energy power distribution networks under the combined effects of strong wind-icing and lightning-wildfire chains. Furthermore, traditional one-time sampling or static fault probability methods are insufficient for time-series sampling of failure probabilities for various types of equipment incorporating historical disaster intensity changes. Summary of the Invention

[0004] This application provides a method for resilience rating of new energy distribution networks under the coupled effects of multiple disasters. To solve the above-mentioned technical problems, this application adopts the following technical method: This application provides a method for resilience rating of new energy distribution networks under the coupled effects of multiple disasters, including: Obtain the current disaster intensity time series; Construct a model for calculating the failure probability of disaster-bearing equipment in the distribution network that takes into account the risk amplification effect; The current disaster intensity time series is input into the failure probability calculation model of the disaster-bearing equipment in the distribution network for calculation, so as to obtain the failure probability of each equipment at the current moment; Obtain the trained conditional flow matching model; The failure probability of each device at the current moment is input into the trained conditional flow matching model to generate the predicted distribution network load shedding curve. Based on the load shedding curve of the distribution network, the resilience level of the new energy distribution network is determined.

[0005] Optionally, the construction of the failure probability calculation model for disaster-bearing equipment in the distribution network that considers the risk amplification effect includes: Construct a criterion model for amplifying the failure risk of disaster-bearing equipment in new energy distribution networks under different disaster scenarios; Based on the aforementioned failure risk amplification criterion model, a failure probability calculation model for disaster-bearing equipment in the distribution network that considers the risk amplification effect is constructed.

[0006] Optionally, the failure risk amplification criterion model for disaster-bearing equipment in new energy distribution networks under different disaster scenarios includes the following: Failure risk amplification criterion model for equivalent load of line under strong wind-icing combined disaster, failure risk amplification criterion model for flashover voltage of transformer bushing external insulation under strong wind-icing combined disaster, failure risk amplification criterion model for air gap breakdown voltage of line under lightning-wildfire chain disaster, and failure risk amplification criterion model for flashover voltage of transformer bushing external insulation under lightning-wildfire chain action.

[0007] Optionally, the step of constructing a failure probability calculation model for disaster-bearing equipment in the distribution network that considers the risk amplification effect based on the failure risk amplification criterion model includes: Based on the aforementioned failure risk amplification criterion model, the effective resistance boundary and voltage stress boundary of the disaster-bearing equipment are introduced, and a failure over-limit margin function considering the risk amplification effect is constructed. Based on the aforementioned failure risk amplification criterion model, the degree of risk amplification is determined; Based on the failure margin function and the risk amplification degree, a calculation model for the failure probability of disaster-bearing equipment in the distribution network that considers the risk amplification effect is constructed.

[0008] Optionally, the training process of the conditional flow matching model includes: Based on historical disaster records, a preset number of historical disaster intensity time series are constructed; the historical disaster intensity time series includes the historical disaster intensity time series of strong wind-ice accumulation concurrent disasters and lightning-wildfire chain disasters. Based on the historical disaster intensity time series, a training sample set of distribution network load shedding curves is generated; The training sample set of the distribution network load shedding curve is normalized to generate a normalized training sample set of the distribution network load shedding curve. Based on the normalized load shedding curves in the training sample set of the normalized load shedding curves of the distribution network, a nonlinear bridging function and a disturbance attenuation term are introduced to construct an intermediate state between the random initial curve and the real load shedding curve. The normalized conditional vectors from the training sample set of the normalized unloaded curves of the distribution network are input into the conditional encoder, and the intermediate curves and evolution time are input into the conditional velocity field network to obtain the model-predicted velocity field. A training loss function is constructed and the model parameters are updated. The objective of the loss function is to make the difference between the predicted velocity field and the target velocity field less than a preset threshold, while constraining the generated unload curve to satisfy the physical boundary. Based on the training loss function, the error between the predicted velocity field and the target velocity field is calculated, and the backpropagation algorithm is used to update the conditional encoder parameters and the conditional velocity field network parameters. After multiple rounds of iterative training, when the training loss tends to stabilize and the error in generating the unloaded curve on the validation samples meets the preset requirements, the trained conditional flow matching model is obtained.

[0009] Optionally, generating a training sample set of distribution network load shedding curves based on the historical disaster intensity time series includes: The historical disaster intensity time series is input into the distribution network disaster-bearing equipment failure probability calculation model that considers the risk amplification effect, and the failure probability of each device corresponding to the historical disaster intensity time series at the current assessment time is generated. For each device and each evaluation time, a random number following a uniform distribution of 0 to 1 is generated, and the random number is compared with the failure probability of each device at the current evaluation time to generate a dynamic failure state matrix; Based on the dynamic failure state matrix and the distribution network topology, the system load loss power at the current moment is obtained. Based on the system's load loss power at the current moment, a training sample set of distribution network load loss curves is generated.

[0010] Optionally, the step of inputting the failure probability of each device at the current moment into the trained conditional flow matching model to generate a predicted distribution network load shedding curve includes: Obtain the current disaster intensity time series, the current equipment failure probability time series, and the current equipment failure status time series; The current disaster intensity time series, the current equipment failure probability time series, the current equipment failure state time series, and the failure probability of each equipment at the current moment are used as condition vectors and input into the trained conditional flow matching model. Starting from the random initial curve, the conditional flow ordinary differential equation is solved to generate the predicted distribution network load shedding curve.

[0011] Optionally, determining the resilience level of the new energy distribution network based on the distribution network load shedding curve includes: Based on the power distribution network load loss curve, three resilience indicators are extracted: maximum load loss, total load loss, and recovery time. Based on the maximum load loss, total load loss, and recovery time, a multi-index threshold matrix hierarchical rule is constructed. Based on the multi-index threshold matrix grading rules, the resilience level of the new energy distribution network is determined.

[0012] Optionally, after determining the resilience level of the new energy distribution network based on the distribution network load loss curve, the step further includes obtaining the types of high-risk failure equipment and disaster types; Based on the aforementioned resilience level, high-risk failure equipment type, and disaster type, disaster prevention and reinforcement measures are determined.

[0013] This application has the following beneficial effects: The method proposed in this application quantifies the amplification effect of equipment failure risk under strong wind-icing and lightning-wildfire chains, improves the generalization ability of distribution network load loss curve prediction under small sample conditions, realizes resilience rating and matching of reinforcement measures, and supports hierarchical management, differentiated disaster prevention and emergency recovery decision-making. Attached Figure Description

[0014] Figure 1 A flowchart illustrating a method for resilience rating of a new energy distribution network under multi-hazard coupling effects, provided in an embodiment of this application; Figure 2 A typical IEEE 123-node distribution network topology diagram provided for embodiments of this application. Detailed Implementation

[0015] 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.

[0016] To solve the above technical problems, such as Figure 1 As shown, this application proposes a method for resilience rating of new energy distribution networks under multi-hazard coupling effects, including: Step S101: Obtain the current disaster intensity time series; Disaster intensity time series data is data on the destructive power of a disaster arranged over time, reflecting the dynamic process of an event from its generation and peak to its decline. It is a key basis for monitoring risk evolution and early warning response. It is usually presented as a curve or sequence graph to track the development trend of disasters and assist in emergency decision-making.

[0017] Step S102: Construct a calculation model for the failure probability of disaster-bearing equipment in the distribution network that considers the risk amplification effect; To characterize the degradation process of the disaster-bearing capacity of power lines and transformers under the coupled effects of multiple disasters, this application uses the equivalent load of power lines under strong wind-icing combined disasters, as well as the flashover voltage of transformer bushing external insulation, and the air gap breakdown voltage of power lines and the flashover voltage of transformer bushing external insulation under lightning-wildfire chain disasters as failure criteria for disaster-bearing equipment. Among these, the equivalent load of power lines is a load-related criterion; an increase in its value increases the risk of mechanical failure of the power lines. The air gap breakdown voltage of power lines and the flashover voltage of transformer bushing external insulation are withstand voltage-related criteria; a decrease in their values ​​increases the risk of electrical insulation failure. Based on these differences, failure risk amplification criterion models for disaster-bearing equipment in new energy distribution networks are established for different disaster scenarios.

[0018] Considering that strong winds alter the freezing efficiency of supercooled water droplets, leading to a further increase in conductor icing thickness and a simultaneous increase in conductor windward area, drag coefficient, and wind pressure load, a failure risk amplification criterion model for the equivalent load of the line under a combined strong wind and icing disaster is constructed based on wind speed and existing icing thickness. (1) The increase in ice thickness under strong winds is as follows: (2) In the formula, This indicates the criterion for amplifying the failure risk of the line under the combined disaster of strong winds and icing; This indicates a scenario involving both strong winds and icing. Indicates wind speed; This indicates the original ice thickness; Indicates the baseline load of the line; Indicates the density of the ice layer; Represents gravitational acceleration; This indicates the increase in ice thickness caused by strong winds; Indicates the diameter of the bare conductor in the circuit; Indicates air density; Indicates aerodynamic viscosity; Indicates the reference Reynolds number; , , This represents the fitting parameters for the drag coefficient; This represents the density of supercooled water droplets; Indicates the characteristic diameter of a supercooled water droplet; This represents the correction factor for water droplet collision efficiency; Indicates freezing efficiency; This indicates the amount of liquid water in the air.

[0019] Considering the bias caused by icing on the bushing surface under wind force, resulting in bridging of the bushing gap by ice and a shortened effective creepage distance, the flashover voltage of the external insulation decreases. Based on the nonlinear competitive relationship between wind force and ice adhesion force, a failure risk amplification criterion model for the flashover voltage of the transformer bushing external insulation is constructed under strong wind-icing concurrent disasters: (3) In the formula, This indicates a criterion for amplifying the failure risk of the flashover voltage of the external bushing insulation of a transformer under a combined strong wind and icing disaster. This represents the flashover voltage of the transformer bushing external insulation under reference conditions; This indicates the flashover field strength along the bushing surface under reference conditions; This indicates the reduced effective creepage distance length after the gap in the umbrella skirt is completely bridged by ice. Indicates the number of gaps in the sleeve skirt; For the first The distance between the gaps in the sleeve skirts, For the first The equivalent radius of a sleeve umbrella skirt; This is the equivalent drag coefficient of the icing sleeve; For temperature The adhesion strength between the underlying ice and the casing surface.

[0020] Considering the alteration of the local gas environment after wildfire plumes intrude into the air gap of power lines, and the further injection of lightning current into the high-temperature plume channel, the concentration of ionized particles increases, while the breakdown field strength of the air gap decreases. Therefore, a failure risk amplification criterion model for the air gap breakdown voltage under lightning-wildfire chain disasters is constructed: (4) In the formula, This indicates the criterion for amplifying the failure risk of air gap breakdown voltage under lightning-wildfire chain disasters; This indicates a lightning-wildfire chain reaction scenario; Indicates the intensity level of lightning; Indicates the intensity level of the wildfire; This indicates the line air gap breakdown voltage under reference conditions; This represents the correction factor for the degradation of the line air gap breakdown voltage; Indicates the reference air breakdown field strength; This indicates the equivalent length of the air gap along the path of wildfire plume intrusion; Indicates the air gap distance of the line; This represents the growth factor of wildfire intensity on plume intrusion length; This indicates the additional concentration of ionized particles after lightning is injected into the plume channel; Indicates the baseline ionized particle concentration; This represents the efficiency coefficient for converting lightning energy into ionized particles; Indicates the reference lightning current intensity; Indicates the duration of the reference lightning current; Indicates the reference plume temperature; and An index representing the influence of lightning intensity on current amplitude and duration; An index indicating the impact of wildfire severity on plume temperature; This represents the correction factor for the enhanced ionization due to plume temperature. Represents the elementary charge; This represents the equivalent cross-sectional area of ​​the plume channel.

[0021] Considering the enhanced charge trapping capability of bushings due to dust particle deposition, and the degradation effect of lightning-induced charge accumulation and local electric field distortion on the external insulation withstand capability, a failure risk amplification criterion model for the flashover voltage of transformer bushings under lightning-wildfire chain reaction is constructed based on dust deposition coverage and surface charge density. (5) Among them, the coverage of wildfire smoke and dust deposition and the surface charge density after lightning strike charge adhesion are respectively: (6) In the formula, This indicates the criterion for amplifying the failure risk of the external insulation flashover voltage of a transformer under a lightning-wildfire chain disaster; Indicates the mass density of reference dust deposition; This represents the critical mass density required for the formation of a continuous soot deposition layer. Indicates the dust deposition coverage on the casing surface; and An index representing the nonlinear impact of wildfire severity on the smoke and dust deposition process; This indicates the charge density adhering to the casing surface after enhanced dust deposition; This represents the equivalent charge adhesion area on the outer insulating surface of the bushing; This represents the efficiency coefficient for the adhesion of lightning charge to the soot deposition layer; This represents the enhancement coefficient of the charge retention capacity of the soot deposit layer; Indicates the electric field distortion coefficient on the casing surface; Represents the vacuum permittivity; This indicates the effective creepage distance of the bushing under reference conditions; and This represents the electric field distortion weighting parameter.

[0022] After obtaining the above four types of failure risk amplification criteria, the effective resistance boundary and voltage stress boundary of the disaster-bearing equipment are further introduced to construct a failure boundary margin function that considers the risk amplification effect. : (7) In the formula, Indicates the effective mechanical resistance boundary of the railway line; Indicates the voltage stress boundary of the air gap in the line; This indicates the voltage stress boundary of the external insulation of the transformer bushing; Based on this, a calculation model for the failure probability of disaster-bearing equipment in the distribution network considering the risk amplification effect is calculated: (8) (9) In the formula, Indicates the first Each disaster-bearing device in the corresponding disaster scenario The failure probability is as follows; This indicates the degree of risk amplification; the larger the value, the more pronounced the risk amplification effect. Indicates the probability of equipment foundation failure; This represents the sensitivity coefficient of the failure out-of-bounds margin to the failure probability. This represents the correction factor for the probability of failure based on the degree of risk amplification.

[0023] Step S103: Input the current disaster intensity time series into the failure probability calculation model of the disaster-bearing equipment in the distribution network for calculation, and obtain the failure probability of each equipment at the current moment; Substitute the current disaster intensity time sequence into formulas (1)-(9) in step S102 for calculation, thereby obtaining the failure probability of each device in the current distribution network at the current moment. .

[0024] Step S104: Obtain the trained conditional flow matching model; Conditional Flow Matching (CFM) is an efficient generative modeling method for training Continuous Normalized Flow (CNF). It smoothly transforms a simple prior distribution (such as Gaussian noise) into a target data distribution by designing conditional probability paths, and uses the regressed target vector field to train the neural network. CFM avoids the expensive ordinary differential equation simulations of traditional CNF while supporting flexible conditional generation. Compared to diffusion models, CFM training is more stable and inference is faster, demonstrating superior performance in tasks such as image generation and molecular conformation prediction.

[0025] Since the conditional flow matching model needs to be trained before use, its training process will be briefly explained below: To construct the training sample set of unloaded curves required for the conditional flow matching model, this application constructs 50 historical disaster intensity time series based on historical disaster records, including historical disaster intensity time series of strong wind-icing concurrent disasters and lightning-wildfire chain disasters. The evaluation duration for each disaster intensity time series is 24 hours, and the evaluation step size is [missing information]. Therefore, each disaster intensity time series contains 96 assessment moments. For the first... Disaster intensity time series, at each assessment time Substitute the current disaster intensity into formulas (1)-(9) in step S102 to calculate the failure probability of each device at the current assessment time corresponding to the historical disaster intensity time series. To simulate whether equipment actually fails during a disaster, a random number following a uniform distribution between 0 and 1 is generated for each piece of equipment at each assessment time. And this random number is combined with the failure probability of each device at the current evaluation time. Comparison: (10) In the formula, This indicates that the device has been determined to be faulty. This indicates that the equipment is not faulty. For equipment that has already failed, the recovery time is based on its fixed recovery period. Determine if recovery is possible; if the failure duration has not been reached. If the condition is not met, the device will remain in a disabled state; if the condition is not met, the device will remain in a disabled state If the failure occurs, the equipment will return to normal and will be re-evaluated in the next assessment.

[0026] Based on the above process, we can obtain the first... The dynamic failure state matrix of all lines and transformers under the disaster intensity time series is generated over 96 assessment times. The load nodes affected by the failed equipment are further identified by combining the distribution network topology connections. For any load node, if there is a failed line in its upstream power supply path, or its transformer is in a failed state, then the load node is determined to be out of power at the current time. The load power of all out-of-power load nodes at the current time is then superimposed to obtain the system's load loss power at the current time. (11) In the formula, Disaster scenario Next Round sampling Load power at any given time; Represents the set of load nodes in the distribution network; Indicates the first Each load node Load power at any given time. This is an indicator function that takes the value 1 if the condition is true, and 0 otherwise. This represents the set of disaster-bearing equipment in the power distribution network; Indicates device The set of downstream load nodes affected by the failure.

[0027] By continuously performing the above calculations along the time series, a complete distribution network load shedding curve is obtained: (12) The same disaster intensity time series is sampled repeatedly 1000 times. That is, different random number sequences are generated repeatedly under the same disaster intensity time series. Since the random numbers are different each time, even if the disaster intensity is the same, different equipment failure state matrices and different load loss curves will be obtained.

[0028] After constructing a training sample set for load shedding curves, a method for predicting load shedding curves in distribution networks based on a conditional flow matching model is further proposed. Conditional flow matching (CFM) is an efficient generative modeling method for training continuous normalized flow (CNF). It smoothly transforms simple prior distributions (such as Gaussian noise) into target data distributions by designing conditional probability paths, and uses the regression target vector field to train the neural network. CFM avoids the expensive ordinary differential equation simulations of traditional CNF while supporting flexible conditional generation. Compared to diffusion models, CFM training is more stable and inference speed is faster, demonstrating excellent performance in tasks such as image generation and molecular conformation prediction.

[0029] The disaster intensity time series, equipment failure probability time series, equipment failure state time series, and equipment fixed recovery time are used as conditional inputs, denoted as the condition vector. The goal of training the conditional flow matching model is to learn the velocity field that evolves from a random initial curve to a true unloaded curve under given conditional inputs. The specific training process is as follows: (1) The above-mentioned training sample set of distribution network load shedding curves is normalized to generate a normalized training sample set of distribution network load shedding curves. Disaster intensity, equipment failure probability, equipment recovery time, and load shedding power are normalized to a uniform numerical range to avoid the influence of variables with different dimensions on model training. The normalized condition vector and load shedding curve are denoted as follows: and .

[0030] (2) Based on the normalized load loss curve in the training sample set of the normalized load loss curve of the distribution network, in order to avoid the load loss recovery stage being too smooth due to direct linear evolution, a nonlinear bridging function and a disturbance attenuation term are introduced to construct an intermediate state between the random initial curve and the real load loss curve. (13) In the formula, For flow matching evolution time; This represents the intermediate state during the transition from the random initial curve to the actual unloaded curve; The random initial load curve is obtained by sampling from the prior distribution; For disturbance terms; It is a nonlinear bridging function used to enhance the nonlinear expressive power of the load reduction and recovery process after a disaster impact; Let be the disturbance attenuation function, so that the random disturbance at... and The two ends disappear to avoid the generated curve deviating from the initial prior and the actual unload boundary; This is the maximum amplitude adjustment factor for random disturbances; The target velocity field is directed from the random initial load curve to the actual unload curve.

[0031] (3) Normalize the conditional vector in the training sample set of the normalized load shedding curve of the distribution network. Input condition encoder Then the intermediate curve and evolution time Input conditional velocity field network The model predicts the velocity field. ; (4) Construct a training loss function and update the model parameters. The objective of the loss function is to make the predicted velocity field more accurate. The difference between the velocity field and the target velocity field is less than a preset threshold. Simultaneously, the generated load shedding curve is constrained to meet physical boundaries; that is, the load shedding power is not less than zero and does not exceed the pre-disaster baseline load. The training loss function is: (14) In the formula, Train the loss function for the conditional flow matching algorithm; The squared error of the L2 norm; Physical boundary constraints for load loss, used to limit the predicted load loss power. Not lower than zero and not exceeding the pre-disaster baseline load ; The number of training samples per batch; These are the corresponding constraint weights.

[0032] (5) Based on the loss function The error between the predicted velocity field and the target velocity field is calculated, and the conditional encoder parameters are updated using the backpropagation algorithm. Conditional velocity field network parameters After multiple rounds of iterative training, when the training loss tends to stabilize and the error in generating the unloaded curve on the validation samples meets the preset requirements, the trained conditional flow matching model is obtained.

[0033] Step S105: Input the failure probability of each device at the current moment into the trained conditional flow matching model to generate the predicted distribution network load shedding curve; After training, for a new disaster scenario, the current disaster intensity time series, the current device failure probability time series, the current device failure state time series, and the failure probability of each device at the current moment are used as the corresponding condition vectors. Input the trained conditional flow matching model, starting from the random initial curve, and solve the conditional flow ordinary differential equation: (15) (16) In the formula, The sampled random initial load curve; The new condition vector is input for the prediction phase; Indicates by Points to The process of solving the conditional flow ordinary differential equation.

[0034] The final predicted distribution network load shedding curve is obtained. : (17) Step S106: Determine the resilience level of the new energy distribution network based on the load shedding curve of the distribution network.

[0035] The distribution network load shedding curve obtained in step S105 Extract three resilience indicators: maximum load loss, total load loss, and recovery time. (18) In the formula, Maximum load loss, expressed in % Total load loss, in percentages (%) Recovery time, in units of ; For the evaluation duration; The recovery ratio threshold is typically set to 0.1. This refers to the time point at which the unload curve reaches its maximum value.

[0036] After obtaining the various indicators of the load shedding curve, a multi-indicator threshold matrix hierarchical rule is constructed, as shown in Table 1: Table 1. Grading Rules for Multi-Indicator Threshold Matrix ; Based on this matrix, the toughness level can be expressed as: (19) In the formula, This represents the hierarchical rules for the multi-index threshold matrix; This indicates the resilience level of the new energy distribution network.

[0037] Finally, disaster prevention and reinforcement measures are matched according to the resilience level, the type of high-risk failure equipment and the corresponding disaster type, as shown in Table 2.

[0038] Table 2 Matching of Disaster Prevention Measures ; Simulation verification The method proposed in this application is applied to a typical IEEE 123-node distribution network system, such as... Figure 2 As shown in the figure. This system uses feeder power nodes as balancing nodes, and branch parameters and node load parameters are based on data from the IEEE 123-node standard test system. Distributed photovoltaic and energy storage devices are connected to different feeder branches to characterize the operating characteristics of the new energy distribution network under multiple disasters. To verify the applicability of the proposed method, the resilience rating results are shown in Table 3.

[0039] Table 3. Classification Results of New Energy Distribution Network under Multi-hazard Coupling ; To verify the accuracy and effectiveness of the proposed method in predicting the load shedding curve, the LSTM time series prediction model and the Transformer time series prediction model were selected as comparison methods. The mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) of each prediction method were calculated based on the load shedding curves generated by different methods. Table 4 shows the comparison of the load shedding curve errors under different prediction methods.

[0040] Table 4 Comparison of Load Loss Curve Errors under Different Prediction Methods ; As shown in Table 4, the proposed conditional flow matching model performs well in predicting the system load shedding curve, with MAE, RMSE, and MAPE values ​​of 27.50kW, 40.67kW, and 3.32%, respectively, all lower than those of the LSTM and Transformer models. This verifies that the proposed method can more accurately characterize the load shedding process under the coupling effect of multiple disasters.

[0041] In summary, the method proposed in this application, through a failure risk amplification criterion model for disaster-bearing equipment in new energy distribution networks under different disaster scenarios, a failure overshoot margin function, a distribution network load shedding curve prediction method based on conditional flow matching algorithm, and a multi-index threshold matrix classification rule, achieves a quantitative characterization of the failure risk amplification effect of disaster-bearing equipment under strong wind-icing concurrent disasters and lightning-wildfire chain disasters. It improves the scenario generalization ability of distribution network load shedding curve prediction under small sample disaster data conditions, realizes the resilience classification of new energy distribution networks and the matching of corresponding disaster prevention and reinforcement measures, and provides technical support for distribution network hierarchical management, differentiated disaster prevention and reinforcement, and emergency recovery decision-making.

[0042] 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.

[0043] 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.

[0044] 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.

[0045] 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 resilience rating of new energy distribution networks under multi-hazard coupling effects, characterized in that, include: Obtain the current disaster intensity time series; Construct a model for calculating the failure probability of disaster-bearing equipment in the distribution network that takes into account the risk amplification effect; The current disaster intensity time series is input into the failure probability calculation model of the disaster-bearing equipment in the distribution network for calculation, so as to obtain the failure probability of each equipment at the current moment; Obtain the trained conditional flow matching model; The failure probability of each device at the current moment is input into the trained conditional flow matching model to generate the predicted distribution network load shedding curve. Based on the load shedding curve of the distribution network, the resilience level of the new energy distribution network is determined.

2. The method according to claim 1, characterized in that, The construction of the failure probability calculation model for disaster-bearing equipment in the distribution network, considering the risk amplification effect, includes: Construct a criterion model for amplifying the failure risk of disaster-bearing equipment in new energy distribution networks under different disaster scenarios; Based on the aforementioned failure risk amplification criterion model, a calculation model for the failure probability of disaster-bearing equipment in the distribution network that considers the risk amplification effect is constructed.

3. The method according to claim 2, characterized in that, The failure risk amplification criterion models for disaster-bearing equipment in new energy distribution networks under different disaster scenarios include the following: Failure risk amplification criterion model for equivalent load of line under strong wind-icing combined disaster, failure risk amplification criterion model for flashover voltage of transformer bushing external insulation under strong wind-icing combined disaster, failure risk amplification criterion model for air gap breakdown voltage of line under lightning-wildfire chain disaster, and failure risk amplification criterion model for flashover voltage of transformer bushing external insulation under lightning-wildfire chain action.

4. The method according to claim 3, characterized in that, The aforementioned failure risk amplification criterion model is used to construct a failure probability calculation model for disaster-bearing equipment in the distribution network that considers the risk amplification effect; including: Based on the aforementioned failure risk amplification criterion model, the effective resistance boundary and voltage stress boundary of the disaster-bearing equipment are introduced to construct a failure boundary margin function that considers the risk amplification effect. Based on the aforementioned failure risk amplification criterion model, the degree of risk amplification is determined; Based on the failure margin function and the risk amplification degree, a calculation model for the failure probability of disaster-bearing equipment in the distribution network that considers the risk amplification effect is constructed.

5. The method according to claim 4, characterized in that, The training process of the conditional flow matching model includes: Based on historical disaster records, a preset number of historical disaster intensity time series are constructed; the historical disaster intensity time series includes the historical disaster intensity time series of strong wind-ice accumulation concurrent disasters and lightning-wildfire chain disasters. Based on the historical disaster intensity time series, a training sample set of distribution network load shedding curves is generated; The training sample set of the distribution network load shedding curve is normalized to generate a normalized training sample set of the distribution network load shedding curve. Based on the normalized load shedding curves in the training sample set of the normalized load shedding curves of the distribution network, a nonlinear bridging function and a disturbance attenuation term are introduced to construct an intermediate state between the random initial curve and the real load shedding curve. The normalized conditional vectors from the training sample set of the normalized unloaded curves of the distribution network are input into the conditional encoder, and the intermediate curves and evolution time are input into the conditional velocity field network to obtain the model-predicted velocity field. A training loss function is constructed and the model parameters are updated. The objective of the loss function is to make the difference between the predicted velocity field and the target velocity field less than a preset threshold, while constraining the generated unload curve to satisfy the physical boundary. Based on the training loss function, the error between the predicted velocity field and the target velocity field is calculated, and the backpropagation algorithm is used to update the conditional encoder parameters and the conditional velocity field network parameters. After multiple rounds of iterative training, when the training loss tends to stabilize and the error in generating the unloaded curve on the validation samples meets the preset requirements, the trained conditional flow matching model is obtained.

6. The method according to claim 5, characterized in that, The generation of a training sample set for distribution network load shedding curves based on the historical disaster intensity time series includes: The historical disaster intensity time series is input into the distribution network disaster-bearing equipment failure probability calculation model that considers the risk amplification effect, and the failure probability of each device corresponding to the historical disaster intensity time series at the current assessment time is generated. For each device and each evaluation time, a random number following a uniform distribution of 0 to 1 is generated, and the random number is compared with the failure probability of each device at the current evaluation time to generate a dynamic failure state matrix; Based on the dynamic failure state matrix and the distribution network topology, the system load loss power at the current moment is obtained. Based on the system's load loss power at the current moment, a training sample set of distribution network load loss curves is generated.

7. The method according to claim 6, characterized in that, The failure probability of each device at the current moment is input into the trained conditional flow matching model to generate a predicted distribution network load shedding curve. include: Obtain the current disaster intensity time series, the current equipment failure probability time series, and the current equipment failure status time series; The current disaster intensity time series, the current equipment failure probability time series, the current equipment failure state time series, and the failure probability of each equipment at the current moment are used as condition vectors and input into the trained conditional flow matching model. Starting from the random initial curve, the conditional flow ordinary differential equation is solved to generate the predicted distribution network load shedding curve.

8. The method according to claim 1, characterized in that, The determination of the resilience level of the new energy distribution network based on the distribution network load shedding curve includes: Based on the power distribution network load loss curve, three resilience indicators are extracted: maximum load loss, total load loss, and recovery time. Based on the maximum load loss, total load loss, and recovery time, a multi-index threshold matrix hierarchical rule is constructed. Based on the multi-index threshold matrix grading rules, the resilience level of the new energy distribution network is determined.

9. The method according to claim 8, characterized in that, The step of determining the resilience level of the new energy distribution network based on the distribution network load shedding curve also includes: Identify the types of high-risk equipment failures and the types of disasters. Based on the aforementioned resilience level, high-risk failure equipment type, and disaster type, disaster prevention and reinforcement measures are determined.

Images