Comprehensive elasticity evaluation method for power distribution network

Abstract

A comprehensive elasticity evaluation method for a power distribution network relates to the technical field of power systems, and comprises the following steps: S1, constructing a multi-source space-time cube model of the power distribution network, and dividing a geographic area into hexagonal grid units; s2, establishing a multi-energy dynamic game model, distributing energy weights based on an improved Shapley value, and realizing multi-agent collaborative optimization through a weight attenuation mechanism driven by output deviation; s3, deploying an edge-cloud collaborative digital twin system, and compressing the equipment model by adopting a model pruning technology; s4, calculating a spatial heterogeneity Gini coefficient, a time persistence toughness index and an energy alternative green elastic equivalent; s5, outputting a dynamic decision support scheme based on a B / S architecture visualization platform; according to the method, the space-time evolution characteristics of the disaster events are combined with the dynamic matching of the contribution degree of the multi-energy main body, so that the problems of static weight distribution defect of elastic evaluation and response lag of multi-energy cooperative scheduling in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a comprehensive elasticity assessment method for a distribution network. Background Art

[0002] With the frequent occurrence of extreme natural disasters (such as typhoons and ice disasters) and cyber attacks, distribution network resilience assessment has become a key technical requirement for the safe operation of power systems. Existing technical solutions have significant deficiencies in dynamic response, spatial modeling, and coordination mechanisms:

[0003] Rigid static weight allocation mechanisms: Traditional resilience assessment methods (such as the Analytic Hierarchy Process and the Entropy Weight Method) rely on expert experience to construct fixed weight systems, making it difficult to capture the dynamic contribution of distributed energy resources during disaster evolution. For example, when photovoltaic output plummets during the initial stages of a typhoon's landfall, energy storage system scheduling continues according to preset weights, increasing the risk of power outages for critical loads. While the multi-temporal and spatiotemporal coordination model proposed in patent CN114638456A incorporates time series analysis, it fails to address the issue of dynamic weight adjustment, making it difficult to adapt to real-time changes in disaster scenarios.

[0004] Lack of spatial heterogeneity modeling: Mainstream technologies often focus on time series analysis, ignoring the impact of geospatial factors (such as terrain elevation and disaster diffusion paths) on power grid damage. Research shows that line damage rates in different geographic regions from the same typhoon can vary by up to 300%. Existing methods fail to integrate spatial grid analysis, resulting in errors exceeding 15% in load loss assessment.

[0005] Centralized simulation is not timely enough: Due to data transmission delays, strategy rehearsals for cloud-based digital twin solutions generally take more than 10 seconds, which cannot meet the prescribed 5-second response requirement. This may delay key decisions in rapidly evolving typhoon scenarios.

[0006] Inefficient coordination among multiple energy entities: Existing models treat photovoltaics, energy storage, and other energy sources as independent units, lacking incentive mechanisms for competitive advantage. In practice, resource scheduling conflicts often arise among various entities driven by economic gains, resulting in elastic support resource utilization rates of less than 40%, highlighting the need for designing coordination mechanisms. Summary of the Invention

[0007] In order to overcome the deficiencies in the background technology, the present invention discloses a comprehensive elasticity evaluation method for a distribution network.

[0008] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0009] A comprehensive elasticity assessment method for a distribution network includes the following steps:

[0010] S1. Construct a multi-source spatiotemporal cube model of the distribution network, divide the geographical area into hexagonal grid cells, integrate meteorological disaster prediction data, equipment status data, and multi-energy output data, and generate a spatiotemporal feature tensor;

[0011] S2. Establish a multi-energy dynamic game model, allocate energy weights based on the improved Shapley value, and achieve multi-agent collaborative optimization through a weight decay mechanism driven by output deviation;

[0012] S3. Deploy an edge-cloud collaborative digital twin system, use model pruning technology to compress device models, and implement millisecond-level preview of elastic recovery strategies through virtual-real data synchronization protocols.

[0013] S4. Calculate the spatial heterogeneity Gini coefficient, the temporal persistence resilience index, and the green elasticity equivalent of energy substitution to generate a multi-dimensional elasticity assessment report;

[0014] S5. Based on the B / S architecture visualization platform, real-time disaster paths, elastic heat maps, and recovery strategy recommendation information are superimposed to output dynamic decision support solutions.

[0015] Preferably, the step S1 uses Geohash coding to divide the distribution network coverage area into hexagonal grids with a side length of 500m; loads multidimensional dynamic parameters to each grid cell, including wind speed field intensity Line vulnerability index , photovoltaic output forecast deviation rate; perform spatiotemporal anomaly detection through the ST-DBSCAN algorithm and dynamically adjust the density threshold in is the set of disaster intensity values ​​within time period t.

[0016] Preferably, the multi-energy dynamic Bo Ben model construction in step S2 includes:

[0017] Defining Energy Aggregates , calculate the elastic contribution within period t , where the photovoltaic reliability coefficient , energy storage ;

[0018] Assign weights based on the modified Shapley value:

[0019]

[0020] Among them, when the actual output deviation When Decay weight, .

[0021] Preferably, the implementation of the digital bend health system in step S3 includes:

[0022] Performed structured pruning on the feeder device model, retaining key parameters and compressing the LSTM network to 32 hidden neurons, reducing the model size by 82%;

[0023] Design virtual and real data synchronization protocol, when real-time measurement value With simulation value When the deviation exceeds 5%, parameter correction is triggered:

[0024]

[0025] Where R is the noise covariance matrix, obtained through field calibration;

[0026] Three types of resilience recovery strategies are pre-generated: topology reconstruction based on Prim's algorithm, multi-energy weighted scheduling, and Dijkstra path planning for mobile power vehicles.

[0027] Preferably, the multi-dimensional elasticity assessment in step S4 includes:

[0028] Assessment of spatial heterogeneity: Calculation of load loss Gini coefficient

[0029]

[0030] in is the average load loss, is the total number of grids;

[0031] Temporal sustainability assessment: Defining a resilience index

[0032]

[0033] when When triggering an early warning;

[0034] Energy Substitutability Assessment: Calculating Green Elasticity Equivalents

[0035]

[0036] Thermal power generators , photovoltaic .

[0037] Preferably, the visualization platform implementation in step S5 includes:

[0038] Basic layer rendering: Drawing real-time typhoon paths and equipment outage status based on WebGL, and displaying red alerts in areas with wind speeds exceeding 25m / s;

[0039] Analysis layer overlay: Use heat map gradient coloring to represent elasticity score. The scoring formula is:

[0040]

[0041] in This is the equivalent value for the full thermal power supply scenario;

[0042] Forecast layer interaction: Click any grid to pop up the recovery strategy details, including estimated recovery time, recommended energy mix and carbon emission comparison data.

[0043] Due to the adoption of the above-mentioned technical solution, the present invention has the following beneficial effects:

[0044] The present invention discloses a comprehensive resilience assessment method for distribution networks. Based on a lightweight digital twin engine for edge computing, the strategy rehearsal delay is compressed from 8.2 seconds in traditional solutions to 0.6 seconds, meeting millisecond-level decision-making requirements. Combined with the Shapley value dynamic weight allocation mechanism, in typhoon cases, the critical load protection rate is increased from 71.3% to 94.7%, and the response efficiency is improved by 33%.

[0045] Using Geohash hexagonal grid division technology, a space-time cube model with a resolution of 500 meters was established. The line corridor terrain data and the disaster diffusion path prediction algorithm were integrated to control the load loss positioning error within 3%, which is 5 times more accurate than traditional GIS systems.

[0046] By building a cooperative game model and introducing blockchain smart contracts, the "free-riding" behavior of energy entities is suppressed through a weight decay penalty mechanism, which has led to a 42% increase in the fulfillment rate of photovoltaic and energy storage. At the same time, combined with a mobile power vehicle path planning algorithm that optimizes the probability of road damage, the arrival time of emergency resources is shortened by 37%. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0048] The present invention can be explained in detail through the following embodiments. The purpose of disclosing the present invention is to protect all technical improvements within the scope of the present invention. In the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "front", "back", "left", "right", etc. to indicate directions or positional relationships, they only correspond to the drawings of this application for the convenience of describing the present invention, and do not indicate or imply that the device or element referred to must have a specific direction.

[0049] Example 1, combined with the attached Figure 1 A comprehensive elasticity assessment method for a distribution network includes the following steps:

[0050] S1. Construct a multi-source spatiotemporal cube model of the distribution network, divide the geographical area into hexagonal grid units, integrate meteorological disaster prediction data, equipment status data, and multi-energy output data, and generate a spatiotemporal feature tensor.

[0051] Specifically, Geohash coding is used to divide the distribution network coverage area into 500m×500m hexagonal grids, and the time axis is superimposed to form a space-time cube. Each grid unit integrates 14 types of dynamic parameters, including meteorological disaster prediction data (typhoon path probability, rainfall intensity), equipment status data (line temperature, circuit breaker opening times), multi-energy output data (photovoltaic power prediction curve, energy storage SOC) and load data (important load proportion, historical recovery time), to construct a feature tensor with a dimension of F∈R14×T (T is the number of time slices). The spatiotemporal anomaly data is eliminated by the improved ST-DBSCAN algorithm, and its density threshold is Dynamically adjust as disasters evolve, is the set of disaster intensity values ​​of all grid cells in time period t

[0052] .

[0053] Disaster impact intensity rating:

[0054] Defining Line Vulnerability Index , where α, β, γ are material correlation coefficients, Based on the improved Jenks natural fracture algorithm, the vulnerability index is combined with the weighted value S of wind speed and rainfall. disaster =0.6⋅v wind +0.4⋅P rain Perform joint clustering to divide the warning areas into four levels: red (>85% percentile), orange, yellow, and blue, and generate a dynamic risk heat map.

[0055] S2. Establish a multi-energy dynamic game model, allocate energy weights based on the improved Shapley value, and achieve multi-agent collaborative optimization through a weight decay mechanism driven by output deviation.

[0056] Specifically, let the energy set participating in the game be E = {PV, ESS, Bio, Grid}, and define the elastic contribution of each subject in time period t as Where ηe is the energy reliability coefficient (0.92 for photovoltaics and 0.98 for energy storage). The improved Shapley value is used to assign weights:

[0057]

[0058]

[0059] In the formula The elastic contribution of alliance S in time period t, To contribute, is the reliability coefficient. Introduce the regret value constraint: if a subject actually outputs , whose weight is ,λ=0.05, forcing the subject to improve the fulfillment rate; is the output deviation of energy e in period t, .

[0060] Constructing the objective function , the constraints include power balance and energy storage charge and discharge rate limits. A distributed solution is implemented using the Alternating Direction Multiplier Method (ADMM), with the game strategy updated every 15 minutes. Each entity's output commitment is recorded via a blockchain smart contract. is the number of energy entities in the alliance S. For example, when S={PV, ESS}, |S|=2; is the total number of energy entities. For example, when E={PV, ESS, Bio, Grid}, |E|=4.

[0061] S3. Deploy an edge-cloud collaborative digital twin system, use model pruning technology to compress the device model, and implement millisecond-level preview of elastic recovery strategies through virtual-real data synchronization protocols.

[0062] Specifically, a TensorRT-based inference engine was deployed at the edge of the substation to perform structured pruning on the feeder equipment model: key parameters (such as transformer winding resistance and circuit breaker operation time) were retained, and the LSTM network hidden layer neurons were compressed from 128 to 32, reducing the model size by 82%. real and the simulation results x virtual When the deviation exceeds 5%, parameter correction is triggered:

[0063]

[0064]

[0065] in is the Kalman gain matrix, H is the observation matrix with dimension m×n, m is the number of measurement variables, n is the number of state variables; R is the noise covariance; is the state prediction covariance matrix.

[0066] For typhoon disaster scenarios, three types of strategies are pre-generated in the digital twin platform:

[0067] Topology reconstruction strategy, based on the improved Prim algorithm to find the largest connected subgraph and prioritize critical loads;

[0068] Multi-energy dispatch strategy: call the weight result of game optimization, press Control energy storage output;

[0069] Mobile power vehicle path planning, combining Dijkstra algorithm with road damage prediction to generate the minimum time path , is the probability of passing the road section.

[0070] S4. Calculate the spatial heterogeneity Gini coefficient, the temporal persistence resilience index, and the green elasticity equivalent of energy substitution to generate a multi-dimensional elasticity assessment report;

[0071] Specifically, calculate the load loss Gini coefficient , reflecting the geographical unevenness of disaster impacts; defining the resilience index , quantify the degree of fit between the recovery process and the ideal curve, where t0, t1 are the start and end times of the disaster impact, such as the typhoon landing time is t0, and the power grid is fully restored time is t1; is the actual power supply power at time τ (kW); is the theoretical power supply power (kW) under normal working conditions at time τ. Introducing green elastic equivalent ,in is the energy type penalty factor, which is 1.0 for thermal generators and 0.1 for photovoltaics; is the CO2 emission coefficient of energy source e (gCO2 / kWh), e.g., 650 for thermal power generators, 48 ​​for photovoltaics, and 0 for energy storage systems (indirect emissions are already accounted for in the grid); is the total output of energy e during the disaster (kWh).

[0072] S5. Based on the B / S architecture visualization platform, real-time disaster paths, elastic heat maps, and recovery strategy recommendation information are superimposed to output dynamic decision support solutions.

[0073] Specifically, the base layer displays real-time typhoon paths and equipment outage status; the analysis layer renders a resilience assessment heat map (red represents low-resilience areas); and the prediction layer overlays a three-hour disaster impact forecast and recommended strategies. Click any grid to view a detailed assessment report, including the contribution of each energy source and expected recovery time.

[0074] Basic layer rendering: Drawing real-time typhoon paths and equipment outage status based on WebGL, and displaying red alerts in areas with wind speeds exceeding 25m / s;

[0075] Analysis layer overlay: Use heat map gradient coloring to represent elasticity score. The scoring formula is:

[0076]

[0077] in This is the equivalent value for the full thermal power supply scenario;

[0078] Forecast layer interaction: Click any grid to pop up the recovery strategy details, including estimated recovery time, recommended energy mix and carbon emission comparison data.

[0079] The parts of the present invention that are not described in detail are prior art. It is obvious to those skilled in the art that the present invention is not limited to the details of the above-mentioned exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and it is intended that all changes that fall within the meaning and scope of equivalent elements are included in the present invention.

Claims

1. A comprehensive elasticity assessment method for a distribution network, characterized in that: The following steps are involved: S1. Construct a multi-source spatiotemporal cube model of the distribution network, divide the geographical area into hexagonal grid cells, integrate meteorological disaster prediction data, equipment status data, and multi-energy output data, and generate a spatiotemporal feature tensor; S2. Establish a multi-energy dynamic game model, allocate energy weights based on the improved Shapley value, and achieve multi-agent collaborative optimization through a weight decay mechanism driven by output deviation; S3. Deploy an edge-cloud collaborative digital twin system, use model pruning technology to compress device models, and implement millisecond-level preview of elastic recovery strategies through virtual-real data synchronization protocols. S4. Calculate the spatial heterogeneity Gini coefficient, the temporal persistence resilience index, and the green elasticity equivalent of energy substitution to generate a multi-dimensional elasticity assessment report; S5. Based on the B / S architecture visualization platform, real-time disaster paths, elastic heat maps, and recovery strategy recommendation information are superimposed to output dynamic decision support solutions.

2. The comprehensive elasticity assessment method for a distribution network according to claim 1, wherein: In step S1, the distribution network coverage area is divided into hexagonal grids with a side length of 500m using Geohash coding; multi-dimensional dynamic parameters are loaded to each grid cell, including wind speed field intensity Line vulnerability index , photovoltaic output forecast deviation rate; perform spatiotemporal anomaly detection through the ST-DBSCAN algorithm and dynamically adjust the density threshold in is the set of disaster intensity values ​​within time period t.

3. The comprehensive elasticity assessment method of the distribution network according to claim 1, characterized in that: The multi-energy dynamic BoBen model construction in step S2 includes: Defining Energy Aggregates , calculate the elastic contribution within period t , where the photovoltaic reliability coefficient , energy storage ; Assign weights based on the modified Shapley value: Among them, when the actual output deviation When Decay weight, .

4. The comprehensive elasticity assessment method of the distribution network according to claim 1, characterized in that: The implementation of the digital health system in step S3 includes: Performed structured pruning on the feeder device model, retaining key parameters and compressing the LSTM network to 32 hidden neurons, reducing the model size by 82%; Design virtual and real data synchronization protocol, when real-time measurement value With simulation value When the deviation exceeds 5%, parameter correction is triggered: Where R is the noise covariance matrix, obtained through field calibration; Three types of resilience recovery strategies are pre-generated: topology reconstruction based on Prim's algorithm, multi-energy weighted scheduling, and Dijkstra path planning for mobile power vehicles.

5. The comprehensive elasticity evaluation method of the distribution network according to claim 1, characterized in that: The multi-dimensional elasticity assessment in step S4 includes: Assessment of spatial heterogeneity: Calculation of load loss Gini coefficient in is the average load loss, is the total number of grids; Temporal sustainability assessment: Defining a resilience index when When triggering an early warning; Energy Substitutability Assessment: Calculating Green Elasticity Equivalents Thermal power generators , photovoltaic .

6. The comprehensive elasticity assessment method of the distribution network according to claim 1, characterized in that: The visualization platform implementation in step S5 includes: Basic layer rendering: Drawing real-time typhoon paths and equipment outage status based on WebGL, and displaying red alerts in areas with wind speeds exceeding 25m / s; Analysis layer overlay: Use heat map gradient coloring to represent elasticity score. The scoring formula is: in This is the equivalent value for the full thermal power supply scenario; Forecast layer interaction: Click any grid to pop up the recovery strategy details, including estimated recovery time, recommended energy mix and carbon emission comparison data.

Images