Matching Methods for Hybrid Distribution Network Architecture Adapted to High Proportion of New Energy Access

Abstract

This invention relates to the field of distribution network technology and discloses a method for matching hybrid distribution network architectures adapted to high-proportion renewable energy access. The method includes: constructing a probabilistic demand vector set; extracting multi-dimensional capability indicators based on the probabilistic demand vector set to construct a dynamic capability vector set and recording dynamic response parameters; obtaining the cumulative robust matching degree and dynamic response capability index of candidate architectures based on the probabilistic demand vector set and the dynamic capability vector set; obtaining a comprehensive score for the candidate architectures based on the cumulative robust matching degree and dynamic response capability index, and ranking them according to the comprehensive score to obtain the optimal architecture scheme. By calculating the comprehensive score of the candidate architectures through the cumulative matching degree of the candidate architectures, combined with the dynamic response capability index constructed based on the dynamic response parameters, the optimal architecture can be selected, ensuring that it possesses both long-term comprehensive adaptability and the ability to cope with rapid power fluctuations caused by high-proportion renewable energy access.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and more specifically, to a method for matching hybrid power distribution network architectures to accommodate a high proportion of renewable energy access. Background Technology

[0002] With the large-scale integration of distributed photovoltaic and wind power, and the rapid growth of DC loads such as electric vehicles and data centers, hybrid AC / DC distribution networks, with their flexible energy dispatch capabilities and adaptability to new energy sources, have become an important development direction for new power systems. In AC / DC distribution network planning, scientifically selecting the optimal architecture (such as SOP interconnection, AC-DC interconnection, AC-DC-AC interconnection, and "two AC-DC" interconnection) based on the load characteristics and new energy penetration rate of typical application scenarios (such as industrial parks, rural distribution networks, and urban core areas) is crucial for improving the economic efficiency, security, and low-carbon operation of the distribution network.

[0003] In recent years, scholars have proposed quantitative matching methods for distribution network architectures based on multi-dimensional indicator systems. For example, Chinese invention patent application CN121367200A discloses a method for selecting AC / DC distribution network structures. This method establishes a flexibility probability demand vector for typical scenarios and a dynamic capability vector for candidate architectures. After processing using max-min classification normalization, it calculates the matching degree using weighted cosine similarity, thereby ranking and selecting the best candidate architectures. This matching method comprehensively considers multi-dimensional indicators such as economic cost, renewable energy absorption rate, voltage deviation, and line overload rate, providing a quantitative decision-making basis for architecture selection.

[0004] However, the above method has the following shortcomings: (1) The scenario modeling is mainly based on deterministic and static typical daily operation curves, which is difficult to cope with the strong intermittency and prediction uncertainty brought about by the high proportion of new energy access. At the same time, the use of fixed probability demand vector ignores the intraday dynamic changes of net load, which may result in the selected architecture having a high matching degree in some periods and insufficient adjustment capacity in other periods. (2) Cosine similarity only measures the consistency of direction and ignores the absolute difference in the magnitude of dynamic capability vectors, which may misjudge the scheme that is "weak in capability but correct in direction"; and its implicit assumption of independent indicators cannot reflect the coupling relationship between indicators such as economic cost and absorption rate. (3) Relying on steady-state power flow calculations, it is impossible to evaluate time response characteristics such as energy storage response speed, converter regulation delay, and voltage change rate. In scenarios with a high proportion of new energy access, rapid power fluctuations place higher demands on the dynamic flexibility of the distribution network, making it difficult to distinguish the dynamic performance of different architectures under rapid power fluctuations.

[0005] Therefore, there is an urgent need for a hybrid distribution network architecture matching method that can adapt to high proportions of renewable energy access. By considering the uncertainty and time-varying characteristics of renewable energy output, and the comprehensive matching metric of consistency of integration direction and absolute approximation, a scientific selection of candidate architectures under uncertain environments can be achieved, thereby improving the foresight, adaptability and reliability of distribution network architecture solutions. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, this invention provides a hybrid distribution network architecture matching method adapted to high proportion of renewable energy access, which solves the problems of poor robustness and low matching degree of architecture selection due to ignoring the power output characteristics of renewable energy and lack of dynamic response assessment, making it unable to adapt to typical scenarios of high proportion of renewable energy access.

[0007] To achieve the above objectives, embodiments of the present invention provide a method for matching hybrid distribution network architectures adapted to high-proportion renewable energy access, comprising: constructing a probabilistic demand vector set based on the access scenario and output characteristics of high-proportion renewable energy; performing time-domain simulation calculations on candidate architectures based on the probabilistic demand vector set, extracting multi-dimensional capability indicators to construct a dynamic capability vector set and recording dynamic response parameters; obtaining the cumulative robust matching degree of candidate architectures based on the probabilistic demand vector set and the dynamic capability vector set, and constructing dynamic response capability indicators based on the dynamic response parameters; obtaining a comprehensive score for candidate architectures based on the cumulative robust matching degree and the dynamic response capability indicators, and ranking them according to the comprehensive score to obtain the optimal architecture scheme.

[0008] In a preferred embodiment, the construction of a probabilistic demand vector set based on the high proportion of new energy access scenarios and their output characteristics includes: dividing typical time periods into multiple time scales based on historical meteorological data and load data; generating probabilistic scenarios based on historical new energy output and load data of typical time periods; selecting multi-dimensional demand indicators of the architecture based on typical time periods and their probabilistic scenarios, constructing a probabilistic demand vector, and obtaining a probabilistic demand vector set.

[0009] In a preferred embodiment, the step of performing time-domain simulation calculations on the candidate architecture, extracting multi-dimensional capability indicators to construct a dynamic capability vector set, and recording dynamic response parameters includes: performing time-domain simulation calculations on the candidate architecture based on the probabilistic demand vector set to obtain simulation results of the candidate architecture; and extracting multi-dimensional capability indicators of the candidate architecture based on the simulation results to construct a dynamic capability vector set and recording dynamic response parameters of the candidate architecture.

[0010] In a preferred embodiment, obtaining the cumulative robust matching degree of the candidate architecture based on the probabilistic demand vector set and the dynamic capability vector set includes: calculating the weighted cosine similarity and the weighted Euclidean distance similarity according to the probabilistic demand vector set and the dynamic capability vector set respectively, and introducing a directional weight coefficient to obtain the comprehensive matching degree of the candidate architecture; and based on the comprehensive matching degree, considering the characteristics of new energy power output, introducing a robustness penalty factor to obtain the cumulative robust matching degree of the candidate architecture.

[0011] In a preferred embodiment, the step of calculating weighted cosine similarity and weighted Euclidean distance similarity based on the probability demand vector set and the dynamic capability vector set respectively includes: performing classification and normalization processing on each index attribute in the probability demand vector set and the dynamic capability vector set; and calculating weighted cosine similarity and weighted Euclidean distance similarity based on the classification and normalization processing results; wherein, the weighted cosine similarity represents the directional consistency between the probability demand vector and the dynamic capability vector in the probability demand vector set and the dynamic capability vector set; and the weighted Euclidean distance similarity represents the absolute difference between the probability demand vector and the dynamic capability vector in the probability demand vector set and the dynamic capability vector set.

[0012] In a preferred embodiment, the step of obtaining the cumulative robust matching degree of the candidate architecture by introducing a robustness penalty factor based on the comprehensive matching degree and considering the characteristics of new energy output includes: calculating the mean and standard deviation of the comprehensive matching degree of the candidate architecture in each time period-scenario based on the comprehensive matching degree of each time period-scenario; introducing a robustness penalty factor based on the mean and standard deviation, and summing the results by weighting the results according to the time period weights to obtain the cumulative robust matching degree of the candidate architecture.

[0013] In a preferred embodiment, constructing a dynamic response capability index based on dynamic response parameters includes: based on the maximum voltage change rate and the average converter response delay, introducing the influence coefficients of the voltage change rate and the response delay to construct the dynamic response capability index.

[0014] In a preferred embodiment, constructing a dynamic response capability index based on dynamic response parameters further includes: setting a preset dynamic performance qualification threshold and verifying the dynamic response capability of candidate architectures in conjunction with the dynamic response capability index; wherein, the dynamic response capability verification includes: comparing the dynamic response capability index of candidate architectures with the dynamic performance qualification threshold, and setting an elimination mechanism to eliminate candidate architectures that fail the dynamic response capability verification.

[0015] In a preferred embodiment, obtaining a comprehensive score for candidate architectures based on cumulative robustness matching degree and dynamic response capability index, and ranking them according to the comprehensive score to obtain the optimal architecture scheme, includes: defining the comprehensive score of candidate architectures as the product of cumulative robustness matching degree and dynamic response capability index; calculating the comprehensive score of candidate architectures and ranking them, and obtaining the optimal architecture scheme according to the ranking result; wherein, the architecture scheme includes: architecture identifier, comprehensive architecture score, and one or more items from the recommended architecture ranking table.

[0016] In a preferred embodiment, obtaining the comprehensive score of the candidate architectures and sorting them according to the comprehensive score further includes: dynamically correcting the final comprehensive score of the candidate architectures based on the comprehensive score of the candidate architectures and the dynamic response capability verification results, and sorting them according to the final comprehensive score.

[0017] The beneficial effects of this invention are: (1) By considering the characteristics of new energy output, a dynamic and probabilistic flexible probability demand vector set is established. For each probability scenario and each typical time period, time-domain simulation calculations are performed on the candidate architecture to extract capability indicators and establish dynamic capability vector sets for each candidate architecture under multiple uncertain probability scenarios and typical time periods. The established probability demand vector set and dynamic capability vector set provide a complete and comprehensive data foundation for subsequent candidate architecture matching degree calculation and dynamic response capability evaluation. (2) Based on the established probability demand vector set and dynamic capability vector set, the long-term robustness of the candidate architecture is evaluated, the cumulative matching degree of the candidate architecture is obtained, and the candidate architecture has long-term comprehensive adaptability is ensured. (3) Based on the dynamic response parameters, construct the dynamic response capability index, and combine the dynamic response capability index with the cumulative robust matching degree to calculate the comprehensive score of the candidate architecture. This can overcome the defects of the candidate architecture relying on steady-state power flow and ignoring time response characteristics, so that the optimal architecture obtained by ranking according to the comprehensive score meets the needs of the hybrid distribution network under rapid power fluctuation and has the ability to cope with rapid power fluctuations with a high proportion of new energy access. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the matching method for hybrid distribution network architectures that adapt to a high proportion of renewable energy access; Figure 2 A schematic diagram of the SOP interconnected distribution network candidate architecture selected in S21; Figure 3 A schematic diagram of the AC-DC interconnected distribution network candidate architecture selected in S21; Figure 4 A schematic diagram of the AC-DC-AC interconnected distribution network candidate architecture selected in S21; Figure 5 A schematic diagram of the candidate architecture for the "two AC and one DC" interconnected distribution network selected in S21; Figure 6 A logical block diagram of a matching method for a hybrid distribution network architecture that adapts to a high proportion of renewable energy access. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 To address the impact of uncertainties and time-varying characteristics of renewable energy output on architecture selection in hybrid distribution networks with a high proportion of renewable energy access, this invention provides a hybrid distribution network architecture matching method adapted to high-proportion renewable energy access. The method includes: constructing a probabilistic demand vector set based on the access scenario and output characteristics of high-proportion renewable energy; performing time-domain simulation calculations on candidate architectures based on the probabilistic demand vector set, extracting multi-dimensional capability indicators to construct a dynamic capability vector set, and recording dynamic response parameters; obtaining the cumulative robust matching degree of candidate architectures based on the probabilistic demand vector set and the dynamic capability vector set, and constructing dynamic response capability indicators based on the dynamic response parameters; obtaining a comprehensive score for candidate architectures based on the cumulative robust matching degree and the dynamic response capability indicators, and ranking them according to the comprehensive score to obtain the optimal architecture scheme.

[0021] For typical scenarios with a high proportion of renewable energy access, considering the characteristics of renewable energy output, a probabilistic demand vector set is constructed based on the architecture requirements of the hybrid distribution network. Candidate architectures for the hybrid distribution network are selected, and time-domain simulation calculations are performed on each candidate architecture based on the probabilistic demand vector set. Multi-dimensional capability indicators corresponding to the probabilistic demand vector set are extracted from the simulation results. A dynamic capability vector set is constructed based on these multi-dimensional capability indicators, while recording dynamic response parameters reflecting the dynamic response capabilities of each candidate architecture. The cumulative robustness matching degree of each candidate architecture is calculated by combining the probabilistic demand vector set and the dynamic capability vector set. Simultaneously, based on the recorded dynamic response parameters of the candidate architectures, a dynamic response capability indicator characterizing the dynamic response capabilities of the candidate architectures is constructed. Finally, based on the obtained cumulative robustness matching degree and dynamic response capability indicator, a comprehensive score for each candidate architecture is calculated, and the candidate architectures are ranked according to their comprehensive scores. The optimal architecture scheme is selected based on the ranking results.

[0022] This invention, under typical scenarios with a high proportion of renewable energy access, establishes a dynamic and probabilistic set of flexible probabilistic demand vectors by considering the characteristics of renewable energy output. For each probabilistic scenario and each typical time period, time-domain simulation calculations are performed on candidate architectures to extract capability indicators and establish dynamic capability vector sets for each candidate architecture under multiple uncertain probabilistic scenarios and typical time periods. The established probabilistic demand vector sets and dynamic capability vector sets provide a complete and comprehensive data foundation for subsequent candidate architecture matching degree calculations and dynamic response capability assessments. Based on the established probabilistic demand vector sets and dynamic capability vector sets, and considering the characteristics of renewable energy output, the long-term robustness of candidate architectures is evaluated from a long-term perspective, obtaining the cumulative matching degree of the candidate architectures. This ensures that the candidate architectures have good robustness under a relatively large number of probabilistic scenarios and typical time periods, avoiding overfitting to a single typical scenario and resulting in insufficient long-term adaptability. Meanwhile, by recording the dynamic response parameters during the simulation calculation of the candidate architecture, constructing a dynamic response capability index based on the dynamic response parameters, and combining the dynamic response capability index with the cumulative robustness matching degree to calculate the comprehensive score of the candidate architecture, the defects of the candidate architecture relying on steady-state power flow and ignoring time response characteristics can be overcome. This allows the optimal architecture scheme obtained by ranking based on the comprehensive score to adapt to the needs of the hybrid distribution network under rapid power fluctuations, making the optimal architecture scheme more in line with the actual needs of dynamic flexibility in high-proportion renewable energy access scenarios.

[0023] To more clearly illustrate the matching method for hybrid distribution network architectures adapted to high proportions of renewable energy integration, please refer to [link to relevant documentation]. Figure 6 The present invention is described in detail below: S1. Based on typical scenarios with a high proportion of new energy access, and considering the characteristics of new energy output, construct a probability demand vector set with multiple time scales and probability scenario sets. For typical scenarios with a high proportion of renewable energy connected to the grid, considering the seasonality, intraday volatility, and forecasting uncertainty of renewable energy output, a dynamic and probabilistic set of flexible probability demand vectors is constructed. Specifically: S11. Based on historical meteorological data and load data, divide typical time periods into multiple time scales; Based on historical meteorological data and load characteristics, the year is divided into several typical periods.

[0024] In one specific embodiment, the year is divided into four seasons: spring, summer, autumn, and winter. For each season, three typical day types are selected: sunny days (high photovoltaic output), cloudy days (low photovoltaic output), and extreme days (such as consecutive rainy days or extreme high-temperature load days). This yields a total of [number missing] days. A typical time period is denoted as .

[0025] S12. Based on historical renewable energy output and load data for typical periods, generate probabilistic scenarios that characterize the uncertainty of renewable energy forecasts. For each typical time period Historical renewable energy output and load data were collected during this typical period, and the Monte Carlo sampling method was used to generate [data / information]. Several probability scenarios are used to characterize the predictive uncertainty of photovoltaic and wind power output. The specific steps are as follows: In a specific embodiment, take ,statistics The probability distribution of photovoltaic power output prediction error within a time period (such as a normal distribution or an empirical distribution based on historical residuals) yields the mean error as... and standard deviation For each scenario The prediction error correction coefficient is obtained by randomly sampling according to this distribution. Then the probability scenario The photovoltaic output curve below is:

[0026] In the formula, Index of intraday times; For time period The benchmark typical daily photovoltaic power output curve.

[0027] Similarly, wind power output and load can also be generated using a similar method to photovoltaic output, generating probabilistic scenarios and assigning equal probability or weights to all scenarios based on historical frequency. Specifically, in the probabilistic scenarios... The wind power output curve below is The time-varying load curve is In the formula, For intraday time index, For probabilistic scenarios The wind power output curve below, For probabilistic scenarios The load output curve under the condition.

[0028] S13. Based on typical time periods and their probability scenarios, construct a probability demand vector and obtain a probability demand vector set according to multi-dimensional demand indicators. S131. Considering that rapid power fluctuations affect the dynamic flexibility of the distribution network when a high proportion of renewable energy is integrated, and that the rational application of renewable energy can improve the overall economic efficiency of the distribution network, the multi-dimensional demand indicators are determined to include economic cost demand indicators. New energy consumption rate demand indicators Voltage deviation requirements Line load rate requirements and net load fluctuation mitigation capability requirements The multi-dimensional demand indicators, among which, the net load fluctuation mitigation capability demand indicator This is the minimum permissible ratio of the net load fluctuation regulation power that the system can provide per unit time (e.g., 15 minutes) to the net load fluctuation amplitude.

[0029] In a specific implementation of a rural power distribution network, an economic cost requirement indicator (i.e., annualized comprehensive cost) is set. Yuan, new energy consumption rate demand indicators Voltage deviation requirement (i.e., average voltage deviation) Line load rate requirements and net load fluctuation mitigation capability requirements Among them, the requirement for net load fluctuation smoothing capability This means the architecture must be able to smooth out at least 80% of net load fluctuations. It should be noted that those skilled in the art can adjust the economic cost requirements based on the specific application scenario. New energy consumption rate demand indicators Voltage deviation requirements Line load rate requirements and net load fluctuation mitigation capability requirements The demand indicators are multi-dimensional. Generally speaking, for industrial park scenarios, the economic demand cost can be appropriately increased and the renewable energy consumption rate can be reduced; for urban core area application scenarios, the voltage deviation demand and line heavy load rate can be increased.

[0030] S132. Based on multi-dimensional demand indicators, construct each time period and each probability scenario The probability demand vector is as follows :

[0031] Since economic cost, voltage deviation, and line overload rate are cost-based indicators, lower values ​​are better; while absorption rate and fluctuation mitigation capability are benefit-based indicators, higher values ​​are better. Therefore, their original numerical forms will be maintained before subsequent classification and normalization. The probability demand vectors constitute a probability demand vector set, denoted as . .

[0032] It should be noted that the probabilistic scenario generation method can also adopt Latin hypercube sampling or scenario generation technology based on generative adversarial networks (GANs), so that the established multi-scenario, multi-time period probabilistic demand vector set can reflect the uncertainty of new energy output, seasonal time-varying characteristics and dynamic adjustment needs.

[0033] In summary, by fully considering the uncertainty and time-varying characteristics of new energy output, a probabilistic demand vector set combining multiple time scales and probabilistic scenario sets is constructed. The whole year is divided into multiple typical periods, and multiple probabilistic scenarios are generated for each typical period to characterize the prediction error and fluctuation range of new energy output such as photovoltaic and wind power, thus accurately depicting the hybrid distribution network architecture requirements in the actual operating environment.

[0034] S2. Based on the probability demand vector set, perform time-domain simulation calculations on the candidate architecture, extract multi-dimensional capability indicators to construct a dynamic capability vector set and record dynamic response parameters. Based on each probability scenario constructed in step S1 and each typical time period For each candidate architecture, time-domain simulation calculations are performed, and its multi-dimensional capability indicators in actual operation are extracted to form a dynamic capability vector set. Specifically: S21. Select typical networking configurations as candidate architectures; Please see Figures 2-5 In one specific embodiment, the following four typical medium- and low-voltage distribution network configurations in hybrid distribution networks are selected as candidate architectures and denoted as: ;in, This indicates a SOP interconnected distribution network (i.e., flexible interconnection switch). This indicates an AC-DC interconnected distribution network (single-ended converter connected to DC line). This indicates an AC-DC-AC interconnected distribution network (i.e., point-to-point DC transmission). This indicates a "two AC and one DC" interconnected distribution network (i.e., multi-terminal DC bus / backbone network).

[0035] S22. Based on each candidate architecture, establish its corresponding time-domain simulation model; For each candidate architecture Electromagnetic transient or quasi-steady-state time-domain simulation models corresponding to each candidate architecture are established in a power system simulation platform. The power system simulation platform can be PSCAD / EMTDC, MATLAB / Simulink, or the open-source tool OpenDSS. The time-domain simulation model includes: AC system, DC system, distributed generation, load, and control strategy. The AC system can adopt the approach described in the invention patent application with publication number CN121367200A. Figure 2 The 33-node standard distribution network topology includes line impedance, transformers, and load nodes; the DC system is connected to voltage source converters (VSC), DC lines, DC buses, and DC loads / energy storage according to the candidate architecture type; distributed generation is connected to photovoltaic (PV) power plants and wind farms (WT) at designated nodes, and its output time-series data comes from the scenario generated by S1. The curve below , The time-varying load curve is Similarly, disturbances can be applied according to the scenario. This can be achieved by using converters with typical droop control or constant power control, and by coordinating with energy storage controlled by constant power or virtual synchronous machines.

[0036] The preset simulation time step is The simulation step size should be 1 second or less to ensure the accuracy of the dynamic process. Meanwhile, the total simulation duration is 24 hours, corresponding to this typical period. Therefore, a simulation step size of 1 second is used. Second.

[0037] It should be noted that for large-scale scenarios (e.g.) (A total of 4800 time-domain simulations) can be performed, and the following acceleration strategies can be adopted: (a) using parallel computing to run multiple simulations simultaneously on a cluster; (b) for quasi-steady-state change scenarios, simplified time-series power flow simulations can be used instead of full electromagnetic transient simulations, with detailed models used only during critical dynamic periods. All of the above acceleration methods fall within the protection scope of this invention.

[0038] S23. Based on the probability demand vector set, simulation calculations are performed using the time-domain simulation models corresponding to each candidate architecture to obtain simulation results; multi-dimensional capability indicators of the candidate architecture are extracted from the simulation results to construct dynamic capability vectors, and a dynamic capability vector set is formed based on the dynamic capability vectors. For each typical time period Each probability scenario and each candidate architecture Simulation results are obtained using time-domain simulation models corresponding to each candidate architecture. Multi-dimensional capability indicators, corresponding to each demand index in the probabilistic demand vector, are then extracted from the simulation results. These multi-dimensional capability indicators represent the simulated actual values ​​of their respective indicators, including economic cost capability indicators, renewable energy absorption rate capability indicators, voltage deviation capability indicators, line overload rate capability indicators, and net load fluctuation mitigation capability indicators. Based on these extracted multi-dimensional capability indicators, a dynamic capability vector is constructed. Furthermore, when extracting each capability indicator, specific requirements are set for each candidate architecture. Record its dynamic response parameters during key periods (such as the period of rapid increase in photovoltaic output within 1 hour after sunrise and the period of rapid decrease in output within 1 hour before sunset) in order to verify the dynamic response capability of the candidate architecture in the future.

[0039] S231. Extract multi-dimensional capability indicators from the simulation results and construct a dynamic capability vector based on these indicators; among them, the multi-dimensional capability indicators include: economic cost capability indicators. New energy consumption rate indicators Voltage deviation capability index Line load capacity indicators and net load fluctuation smoothing index Specifically: S2311, Extraction of Economic Cost Capability Indicators This includes electricity purchase costs, grid loss costs, and annualized equipment investment costs: , , , In the formula, This represents the annual electricity purchase cost (i.e., the total cost of purchasing electricity from the upper-level power grid). This refers to the annual network loss cost (i.e., the expenses incurred due to losses of lines, equipment, etc.). The average annual investment cost of equipment such as converters and DC lines, calculated based on their service life; For a moment The price of electricity purchased from the upper-level power grid (which may be a time-of-use price); For a moment The active power purchased by the system from the upper-level power grid; This is the simulation time step; The penalty electricity price corresponding to the network loss cost (usually the average electricity purchase price or a special penalty price); For a moment The total network power loss of the system (including losses from AC lines, converters, DC lines, etc.). This represents all time segments throughout the year (each time step). (Accumulate)

[0040] S2312, Extracting indicators of new energy consumption rate capacity (%) represents the proportion of actual renewable energy power generation consumed to available renewable energy power generation. , In the formula, The actual renewable power received; Available renewable power (which is limited by equipment capacity and grid regulation capabilities). This is the simulation time step.

[0041] S2313, Capability to extract voltage deviation index (%) represents the average voltage deviation of all nodes in the entire network at all times: , In the formula, For nodes At any moment The voltage amplitude; Rated voltage (per unit value is 1.0); The total number of nodes; This represents the total number of time sections. Indexed by the number of nodes; Index for the number of time segments (i.e., time points).

[0042] S2314, Extracting Line Load Rate Capability Indicators (%)for: , In the formula, For the line At any moment The power of the current flow; For the thermal stability limit of the line, This represents the number of lines.

[0043] S2315, Extracting indicators of net load fluctuation mitigation capability It is the minimum ratio of the net load fluctuation regulation power that the system can provide per unit time to the actual fluctuation amplitude of the net load.

[0044] Calculate the net load curve In the formula, For a moment Net load power; For a moment Total load active power; For a moment Photovoltaics always contribute power; For a moment Wind power always contributes to the effort.

[0045] Based on this net load curve, the entire time period Divided into There are 3 consecutive time windows, each with a length of 1. (e.g., 15 minutes); Calculate the length of all time windows. Actual fluctuation range of net load within (e.g., 15 minutes) for: ; In the formula, For the first The fluctuation range of net load within a time window; This represents the maximum net load within the time window. This represents the minimum net load within the time window. For the first All moments within a time window A set of.

[0046] Based on the actual fluctuation range of net load within all time windows With adjustable power capacity The net load fluctuation mitigation capability index was obtained. for: , In the formula, The index of the time window (i.e., representing the first...) (a time window) .

[0047] Among them, adjustable power capacity The dynamic response capability is determined by factors such as the available capacity of the converter and the charge / discharge power of the energy storage in the simulation.

[0048] Economic cost capability indicators extracted from S2311-S2315 New energy consumption rate indicators Voltage deviation capability index Line load capacity indicators and net load fluctuation smoothing index Construct dynamic capability vectors as .

[0049] S232. Based on the constructed dynamic capability vectors, a dynamic capability vector set is formed; For each time period Each probability scenario and each candidate architecture (For clarity, four typical candidate architectures selected in a specific embodiment of S21 are used as examples.) After performing simulation and capability index extraction, a dynamic capability vector is obtained. and candidate architectures Dynamic response parameters during critical periods , .

[0050] Based on all dynamic capability vectors, the dynamic capability vector set is formed as follows: And record each candidate architecture Dynamic response parameters during critical periods , For use in subsequent verification.

[0051] S24. Record the dynamic response parameters of the candidate architecture during the simulation process; wherein, the dynamic response parameters include the maximum voltage change rate and the average converter response delay; Based on the probability requirement vector set, when performing time-domain simulations for each candidate architecture, the data for each candidate architecture is recorded. The dynamic response parameters during key time periods in the time-domain simulation include the maximum voltage change rate and the average converter response delay. Specifically: The maximum absolute value of the voltage change rate between adjacent time points is calculated for all nodes during the rapid ramp-up period, and this value is taken as the maximum voltage change rate. : , In the formula, Candidate architecture The maximum rate of change of voltage; For nodes At any moment The voltage amplitude.

[0052] The average time required for all voltage source converters (VSCs) to receive a power command change and for the actual output power to reach 90% of the target value is calculated as the system average converter response delay. : , In the formula, Candidate architecture The system average converter response delay; The number of all voltage source converters (VSCs); For the first Response delay of the converter; Number the converter index.

[0053] In summary, based on each probability scenario constructed in step S1 and each typical time period Based on the constructed probability demand vector set, time-domain simulation calculations were performed on each candidate architecture, and the simulated actual values ​​of each candidate architecture were extracted as multi-dimensional capability indicators. A dynamic capability vector set was then constructed based on this, forming a dynamic capability vector set. This dynamic capability vector set provides a complete data foundation for subsequent classification normalization, matching degree calculation, and robustness evaluation of the obtained candidate architectures under multiple uncertain probability scenarios and typical time periods. Simultaneously, based on the simulation results, two dynamic response parameters—the maximum voltage change rate and the average converter response delay—were recorded for each candidate architecture during critical time periods, providing a parameter basis for subsequent dynamic response capability verification.

[0054] S3. Based on the probability demand vector set and the dynamic capability vector set, the cumulative robustness matching degree of the candidate architecture is obtained, and the dynamic response capability index is constructed according to the dynamic response parameters. Based on the probability demand vector set obtained in S1 and the dynamic capability vector set obtained in S2, the vector indicators within each vector set are classified and normalized according to their indicator attributes. Based on the classification and normalization results, weighted cosine similarity is calculated to represent the directional consistency between the probability demand vector and the dynamic capability vector in the multi-dimensional indicator space; weighted Euclidean distance similarity is calculated to represent the absolute difference between the probability demand vector and the dynamic capability vector. Then, a directional weight coefficient is introduced, and the weighted cosine similarity and weighted Euclidean distance similarity are weighted and fused to obtain the comprehensive matching degree of each candidate architecture in each time period and scenario. Based on the comprehensive matching degree and considering the characteristics of new energy power output, to ensure that the candidate architecture can adapt in the long term and maintain good robustness, a robustness penalty factor is introduced. The cumulative robust matching degree of the candidate architecture is obtained by weighted summation over time periods.

[0055] Meanwhile, to ensure that the final selected architecture has good dynamic response capabilities in practical applications, overcome dependence on steady-state power flow, and eliminate the deficiency of ignoring time response characteristics, a response capability index is constructed based on the dynamic response parameters of the architecture during key periods recorded in the time-domain simulation. Furthermore, in some embodiments, the dynamic response capabilities of candidate architectures can be verified based on the dynamic response parameters recorded during simulation, and an elimination mechanism can be set to preemptively eliminate candidate architectures that do not meet the requirements. Specifically: S31. Based on the probability demand vector set and the dynamic capability vector set, perform classification and normalization processing; To eliminate the differences in the dimensions and attributes of different vector indicators between the probabilistic demand vector set and the dynamic capability vector set, the probabilistic demand vector obtained from S1 is... The dynamic capability vector obtained with S2 The classification and normalization process is as follows: For each time period Each probability scenario and each candidate architecture The requirement indicators of each probability requirement vector in the probability requirement vector set and the capability indicators of each dynamic capability vector in the dynamic capability vector set are subjected to max-min classification normalization. Specifically: Let the original matrix be... From all probability demand vectors With dynamic capability vectors In the same index space The composition of the various indicator values. For each indicator... ( ), and count all the probability demand vectors it contains. With dynamic capability vectors The original minimum value and maximum value .

[0056] Based on the indicator attributes of demand and capability indicators in each vector, each indicator is classified into benefit-type indicators and cost-type indicators. With the goal of representing better performance by larger values ​​after classification and normalization, different classification and normalization methods are used to process the indicators according to the classification results.

[0057] Taking the probability demand vector as an example, its new energy consumption rate demand index Requirements for Net Load Fluctuation Smoothing Capacity As this is a benefit-oriented indicator, its value should be as high as possible after classification and normalization. Therefore, the classification and normalization process is as follows: ; Its economic cost demand index Voltage deviation requirements and line load rate requirements As this is a cost-related indicator, its value should be as small as possible after classification and normalization. Therefore, the classification and normalization process is as follows: , In the above classification and normalization formula, These are the original index values ​​of each demand index in the probability demand vector; and Each in the same scene During the same period Within, the maximum and minimum values ​​of all candidate architectures and probability requirement vectors on the corresponding requirement indicators; All values ​​after classification and normalization , All are mapped to intervals Furthermore, a larger value indicates better performance. The probability requirement vector after classification normalization is denoted as: In the formula, This represents the normalized economic cost demand value. This represents the normalized demand value for renewable energy consumption rate. This is the normalized voltage deviation requirement value; This represents the normalized line overload rate demand value. This represents the normalized net load fluctuation mitigation capability requirement.

[0058] Similarly, for dynamic capacity vectors, their renewable energy absorption rate is an important indicator. Net load fluctuation mitigation index These are efficiency-type indicators, specifically economic cost-capacity indicators. Voltage deviation capability index and line load capacity indicators As cost-related indicators, after classification and normalization, the classified and normalized dynamic capability vector is denoted as: In the formula, This is the normalized economic cost capability value; This is the normalized value for the renewable energy absorption rate. This is the normalized voltage deviation capability value; This is the normalized line overload capacity value; This is the normalized value for net load fluctuation mitigation capability.

[0059] S32. Based on the classification and normalization results, calculate the weighted cosine similarity and weighted Euclidean distance similarity respectively, and introduce a directional weight coefficient to obtain the comprehensive matching degree of each candidate architecture in each time period and scenario; specifically: S321. Weighted cosine similarity is used to measure the directional consistency between the probability demand vector and the dynamic capability vector after classification normalization in the multidimensional index space. Its calculation formula is as follows: , In the formula, For time period Scene Next candidate architecture The weighted cosine similarity, with values ​​ranging from 1 to 10. Furthermore, the larger the value, the higher the directional consistency. For the indicator weight vector, ,in, The total number of indicators is represented by the value 'n'. In this embodiment, both the constructed probability demand vector and dynamic capacity vector include five indicators: economic cost, renewable energy absorption rate, voltage deviation, line overload rate, and net load fluctuation shifting capability. Each weight component And generally ; This is the element-wise multiplication (Hadamard product), which means multiplying the corresponding components of two vectors. Let L2 be the L2 norm (i.e., Euclidean norm) of the vector.

[0060] After classifying and normalizing the probability demand vector, when all its components approach 1, the weighted cosine similarity reflects whether the "direction" of the dynamic capability vector is consistent with the ideal direction.

[0061] S322, Weighted Euclidean distance similarity is used to measure the absolute difference between the dynamic capability vector and the probability requirement vector after classification normalization.

[0062] To avoid the influence of dimensions and to ensure comparability with the weighted cosine similarity in S321, it is converted into a similarity form (the larger the value, the smaller the difference). Specifically: The weighted Euclidean distance is calculated as follows: , In the formula, The first dynamic capability vector Each category is a normalized component; The first probability demand vector Each category is a normalized component.

[0063] The maximum possible weighted Euclidean distance occurs when the dynamic capability vector is all 0s (worst case) and the probability demand vector is all 1s (optimal case), i.e.: , Based on this, the weighted Euclidean distance similarity is defined as: , In the formula, For weighted Euclidean distance similarity, its value is in Within the range, the closer the value is to 1, the closer the dynamic capability vector is to the probability demand vector in absolute value (i.e., the better the overall performance of each indicator).

[0064] S323. Based on weighted cosine similarity and weighted Euclidean distance similarity, a directional weight coefficient is introduced for weighted fusion to obtain the comprehensive matching degree of the candidate architecture; Based on weighted cosine similarity and weighted Euclidean distance similarity, a direction weight coefficient is introduced to simultaneously consider directional consistency and absolute approximation. The weighted cosine similarity and weighted Euclidean distance similarity are weighted and fused as follows: , In the formula, For time period Scene Next candidate architecture The overall matching degree, with a value range of 100%. The larger the value, the better the architecture matches the rain requirements in that time period and scenario; For directional weighting coefficients, when When the overall matching degree degenerates into a pure weighted cosine similarity, when... When the overall matching degree depends only on the absolute approximation degree (i.e., weighted Euclidean distance similarity), when At that time, the overall matching degree takes into account both weighted cosine similarity and weighted Euclidean distance similarity.

[0065] Among them, the direction weight coefficient The weighting can be adjusted according to the characteristics of the scenario. For example, in high-penetration scenarios, more attention is paid to absolute absorption capacity (i.e., renewable energy absorption rate), so the directional weighting coefficient can be appropriately reduced. .

[0066] It should be noted that the indicator weight vector in S321 With the directional weighting coefficient in S323 The value of can be adjusted by those skilled in the art according to actual engineering needs, such as by entrusting experts to assign values ​​using the Analytic Hierarchy Process (AHP), or by adaptive calculation based on scenario characteristics (such as the penetration rate of new energy sources). Such adjustments are all within the protection scope of this invention. Furthermore, when When there is a strong correlation among the indicators, principal component analysis can be performed first to reduce the dimensionality, and then the method in step S32 of this invention can be applied. This is still considered an equivalent substitution of the present invention.

[0067] Based on the classification and normalization results, weighted cosine similarity and weighted Euclidean distance similarity are calculated respectively, and a directional weight coefficient is introduced to obtain the comprehensive matching degree of each candidate architecture in each time period and scenario. This comprehensive matching degree calculation employs both weighted cosine similarity (measuring directional consistency) and weighted Euclidean distance similarity (measuring absolute closeness), and uses directional weighting coefficients... By employing weighted fusion, compared to matching degree calculation methods that rely solely on directional consistency, this invention can effectively distinguish between candidate architectures with "correct capability vector directions but generally low absolute values" and those with "correct directions and excellent absolute values." This makes the matching degree evaluation of candidate architectures more comprehensive and accurate, avoiding misjudgments caused by the blind spot of module length. At the same time, the comprehensive matching degree of candidate architectures calculated based on weighted cosine similarity and weighted Euclidean distance similarity also lays the foundation for subsequent robustness evaluation and cumulative matching degree calculation.

[0068] S33. Based on the comprehensive matching degree and considering the characteristics of new energy output, a robustness penalty factor is introduced to obtain the cumulative robust matching degree of the candidate architecture. Based on the comprehensive matching degree of each candidate architecture obtained from S32 in each time period and scenario Considering the uncertainty of renewable energy output, the matching degree may fluctuate under different probability scenarios within the same time period. To evaluate the overall performance of the candidate architecture in long-term operation and its robustness to the uncertainty of renewable energy output, the comprehensive matching degree under different probability scenarios within the same time period is analyzed. Statistical analysis was performed, and a robustness penalty factor was introduced to calculate the cumulative robustness matching degree of the candidate architecture throughout the year. Specifically: S331. Based on the comprehensive matching degree of the candidate architecture in each time period and scenario, calculate the mean and standard deviation of the comprehensive matching degree; For each candidate architecture ( ,in (corresponding to the four typical networking forms in S21) and each typical time period ( , (representing 12 months in a year), based on the comprehensive matching degree of each candidate architecture in each time period and scenario obtained from S32. ( ), calculate the mean with standard deviation Two statistics.

[0069] Typical period Inside The average overall matching degree under each probability scenario Used to reflect the architecture during typical periods The average matching performance within the range is calculated as follows: ; Typical period Inside Standard deviation of overall matching degree under various probability scenarios Used to reflect the architecture during typical periods The degree of fluctuation in the matching degree under different probability scenarios (i.e., uncertainties in the output of different new energy sources) is measured. The smaller the standard deviation, the better the robustness. The calculation is as follows: .

[0070] S332. Based on the mean and standard deviation, a robustness penalty factor is introduced, and the cumulative robustness matching degree of the candidate architecture is obtained by weighting the sum by time period. Based on the calculated mean with standard deviation Two statistics are used to comprehensively evaluate candidate architectures. Throughout the year, a robust penalty factor is introduced, rewarding both high average matching accuracy and low inter-scenario volatility based on overall performance. and weighted by time period Perform a weighted summation to obtain the cumulative robust matching score. : , In the formula, Candidate architecture The cumulative robust matching score throughout the year is used, and the larger the value, the better the overall performance. The best matching is selected based on this score. Typical period The weighting coefficients satisfy ; As a robust penalty factor, The strength of the penalty for matching fluctuations can be controlled by the robustness penalty factor. When, it means no punishment; when At that time, the larger the standard deviation, the more points are deducted.

[0071] Among them, typical period Weighting coefficients The weighting can be based on the actual operating time or importance of typical time periods; for example, if the peak photovoltaic output period at midday on sunny summer days in rural power distribution networks has the highest requirement for grid absorption capacity, then this period can be given a larger weight; if the importance of each period is similar, then equal weighting can be applied. In practical applications, the parameters can be set by planners based on local load characteristics and the penetration rate of new energy sources.

[0072] Robust punishment factor This determines the decision-maker's tolerance for performance fluctuations; a generally recommended range is [range missing]. If the planning objective is to achieve the lowest possible performance under the worst-case scenario, a larger value can be chosen. (like If average performance is a greater concern, a smaller value can be chosen. (like In practical applications, It can also be adaptively adjusted through cross-validation or by matching with the risk preferences of power grid operation.

[0073] It should be noted that the form of punishment in this step... This is a commonly used robustness optimization objective. Those skilled in the art can also use other robustness metrics, such as maximizing the worst-case scenario matching degree. Or adopt conditional value of risk (VaR) The methods described herein are equivalent substitutions under the concept of this invention and still fall within the protection scope of this invention.

[0074] In summary, by extending the comprehensive matching evaluation under a single scenario to a robustness assessment that considers the characteristics of new energy output (mainly the uncertainty of new energy output), the mean and standard deviation of the matching degree of each candidate architecture under different scenarios are calculated, and a robustness penalty factor is introduced. The cumulative robust matching degree of the candidate architecture is obtained by weighting and summing according to the time period weight. This ensures that the final selected architecture not only performs well in terms of average performance, but also improves the long-term comprehensive adaptability of the architecture under all-weather, multi-season, and different weather conditions, avoiding the problem of "high matching in some time periods and serious insufficiency in other time periods" that is easily caused by single-scenario selection.

[0075] S34. Based on the recorded dynamic response parameters, construct dynamic response capability indicators and perform dynamic response capability verification. For each candidate architecture ( (corresponding to the four typical networking forms in S21), and each candidate architecture is recorded in S22-S23. The corresponding time-domain simulation model performs a time-domain simulation of the dynamic response parameter, specifically the maximum voltage change rate. and system average converter response delay Construct dynamic response capability index And the dynamic response capability of the candidate architecture is verified. Specifically: S341. Based on the maximum voltage change rate and the average converter response delay, the influence coefficients of the voltage change rate and the response delay are introduced to construct a dynamic response capability index. To comprehensively evaluate the dynamic performance of candidate architectures, the maximum voltage change rate, a recorded dynamic response parameter, was used. and system average converter response delay Construct dynamic response capability index as follows: , In the formula, Candidate architecture Dynamic response capability indicators; The influence coefficient of voltage change rate; This is the impact coefficient of response delay.

[0076] Among them, the maximum voltage change rate The smaller the value, the stronger the architecture's ability to maintain voltage stability during periods of drastic fluctuations in renewable energy power. System average converter response delay The smaller the value, the faster the power regulation response of the architecture.

[0077] S342. Preset dynamic performance qualification threshold and verify the dynamic response capability of candidate architectures in combination with dynamic response capability indicators. To prevent candidate architectures with poor dynamic response capabilities from being selected as the best architecture, a pre-set dynamic performance threshold is set. (Pick Dynamic response capability verification includes comparing the dynamic response capability metrics of candidate architectures with dynamic performance qualification thresholds, and setting an elimination mechanism to eliminate candidate architectures that fail the dynamic response capability verification. That is, if a candidate architecture's dynamic response capability is not met... If the candidate architecture is deemed not to meet the requirements of rapid power fluctuations in high-proportion new energy scenarios, an elimination mechanism can be set to directly eliminate it and remove it from the subsequent ranking.

[0078] In one specific embodiment, a preset This value can be calculated based on the actual operating standards of the power grid (such as voltage change rate not exceeding 1kV / s, response delay not exceeding 50 ms). For example: if the requirement is... kV / s, ms, get , ,but Therefore, take This ensures the elimination of schemes that exceed limits.

[0079] It should be noted that the influence coefficient of the voltage change rate The influence coefficient of response delay It can be adjusted according to the actual power grid's sensitivity to voltage fluctuations and response speed. For example, for data center power supply scenarios with extremely high power quality requirements, the [adjustment / adjustment] can be increased. This is to strengthen the penalty for the rate of voltage change.

[0080] Similarly, in practical applications, planners can flexibly set the dynamic performance qualification threshold according to local operating procedures. The preset dynamic performance qualification threshold is... The values ​​can also be set to be relatively lenient or stringent based on the planner's risk preference. In the dynamic response capability verification process, setting an elimination mechanism is a preferred implementation. Specifically, candidate architectures that fail the dynamic response capability verification are eliminated, while candidate architectures that pass the verification are retained. The candidate architectures that fail the dynamic response capability verification are those that do not possess the ability to meet the rapid power fluctuation requirements of high-proportion renewable energy scenarios.

[0081] In summary, for typical scenarios with a high proportion of renewable energy integration, to meet the requirements for rapid adjustment capabilities of the hybrid distribution network architecture, a dynamic response capability index for candidate architectures is constructed based on the dynamic response parameters obtained during the simulation process. Furthermore, the dynamic response capability of candidate architectures is verified beforehand, eliminating those that do not meet the requirements. This fills the gap in existing architecture selection methods that only focus on the stability performance of the distribution network architecture while ignoring the time response characteristics of adjustment resources. This ensures that the optimal architecture ultimately selected by this invention does not rely on a steady-state power flow model and can better adapt to the actual needs of scenarios with a high proportion of renewable energy integration.

[0082] S4. Based on the cumulative robustness matching degree and dynamic response capability index, obtain the comprehensive score of the candidate architectures, and rank them according to the comprehensive score to obtain the optimal architecture solution. S41. Based on the cumulative robustness matching degree and dynamic response capability index of the candidate architecture obtained in S3, the comprehensive score of the candidate architecture is defined as the product of the cumulative robustness matching degree and the dynamic response capability index. Specifically: The cumulative robust matching degree calculated by S332 With dynamic response capability indicators Multiply by the product to obtain the comprehensive score of the candidate architecture that passes the dynamic response capability test. for: , By accumulating robust matching degree With dynamic response capability indicators Multiply them to get the combined score of the product. This reflects that candidate architectures with stronger dynamic response capabilities will receive higher scores under the same static-robust performance; candidate architectures whose dynamic capabilities only reach the passing grade will have their scores appropriately reduced, but will not be eliminated.

[0083] It should be noted that the overall score of the candidate architecture In addition to using cumulative robust matching degree With dynamic response capability indicators In addition to multiplication, weighted summation or other nonlinear mapping methods can also be used, as long as they can reflect the positive impact of dynamic performance on the overall score of candidate architectures.

[0084] By incorporating dynamic response capability into the evaluation system for candidate architecture matching, and combining this dynamic response capability index with the cumulative robustness matching degree, a comprehensive score for the candidate architecture is calculated. This ensures that the final selected architecture possesses both long-term operational robustness and good dynamic response capability, thus solving the problem of relying solely on steady-state power flow and being unable to distinguish the actual performance of different candidate architectures under rapid power fluctuations. The candidate architecture obtained by combining dynamic response capability and cumulative robustness matching degree is more in line with the actual needs of dynamic flexibility in scenarios with a high proportion of new energy access.

[0085] S42. Based on the comprehensive score of the candidate architecture and the dynamic response capability verification results, dynamically adjust the final comprehensive score of the candidate architecture. Based on the comprehensive score of the candidate architectures and combined with the dynamic response capability verification results, the final comprehensive score of the candidate architectures is dynamically adjusted to determine the preferred embodiment. Specifically: For candidate architectures that pass the dynamic response capability verification (i.e.) The final comprehensive score of its candidate architecture after modification. for: For candidate architectures that fail the dynamic response capability verification (i.e. When an elimination mechanism is in place, it will be eliminated early. In this case, based on the cumulative robust matching degree... With dynamic response capability indicators By combining the dynamic response capability verification results, the final comprehensive score of each candidate architecture is obtained through dynamic adjustments. for: , In a specific embodiment, the candidate architecture in S21 (For example, the "two-way interconnection" type) is recorded through S23. kV / s, ms; get , ,calculate Preset dynamic performance qualification threshold ,because Therefore, this candidate architecture is retained. Based on the cumulative robust matching degree... Formula, calculated (Example value), then through comprehensive scoring Formula calculation Then, let's look at the candidate architectures in S21. (For example, SOP interconnection type) is obtained through S23 recording. kV / s, ms; then ,because Therefore, this candidate architecture was directly eliminated.

[0086] S43. Calculate the comprehensive score or final comprehensive score of the candidate architectures and sort them. Obtain the optimal architecture scheme based on the sorting results. The architecture scheme includes one or more of the following: architecture identifier, comprehensive architecture score, and recommended architecture sorting table.

[0087] S431. When ranking candidate architectures based on their comprehensive scores, without considering the dynamic response capability verification results, there is no need to consider the elimination mechanism set during the dynamic response capability verification. All candidate architectures can be ranked according to their directly calculated comprehensive scores. The optimal architecture is obtained by sorting the data. Specifically: Calculate each candidate architecture Overall rating The overall score obtained. The system ranks the candidate architectures and selects the one with the highest comprehensive score as the optimal hybrid distribution network architecture for adapting to a high proportion of renewable energy access. The ranking results can be output as needed for comparison of multiple schemes in practical applications. For example, the final output of the optimal architecture scheme includes: optimal architecture identifier. and its corresponding architecture name (e.g., "two AC and one DC" interconnected distribution network), and comprehensive architecture score. and at least one item from the recommended architecture ranking table, wherein the recommended architecture ranking table lists the rankings of all qualified candidate architectures and their... , , The indicators are broken down into smaller parts so that planners can understand the scoring structure or make manual adjustments based on additional constraints (such as investment budgets).

[0088] S432. Based on the comprehensive score of the candidate architectures and the dynamic response capability verification results of the candidate architectures, a revised final comprehensive score is obtained for the candidate architectures. These final comprehensive scores are then used to rank the candidate architectures and determine the optimal architecture solution. Specifically: During the dynamic response capability verification process, if an elimination mechanism is set, candidate architectures eliminated according to the set elimination mechanism will no longer participate in the final comprehensive score ranking of candidate architectures; if no elimination mechanism is set, the final comprehensive score of all candidate architectures will be calculated. The system sorts all candidate architectures according to their final comprehensive scores, selects the optimal architecture, and outputs the optimal architecture solution.

[0089] In one specific embodiment, the final comprehensive scores of all candidate architectures are sorted from largest to smallest, resulting in the following ranking: ,in, The number of candidate architectures that passed the verification ( The candidate architecture ranked first is selected as the recommended architecture. The optimal architecture scheme is output, including: the optimal architecture is identified as... And its corresponding architecture name (e.g., "two AC and one DC" interconnected distribution network), the final comprehensive score of the optimal architecture is and at least one item from the recommended architecture ranking table, wherein the recommended architecture ranking table lists the rankings of all qualified candidate architectures and their... , , The indicators are broken down into smaller parts so that planners can understand the scoring structure or make manual adjustments based on additional constraints (such as investment budgets).

[0090] In a specific embodiment, taking a rural power distribution network scenario as an example, assuming that the four candidate architectures selected are those in S21, and that none of the four architectures are eliminated after the dynamic response capability verification in S34, their final comprehensive scores are shown in Table 1 below: Table 1: Cumulative robustness matching degree, dynamic response capability index, and final comprehensive score statistics of four candidate architectures in rural power distribution network scenarios.

[0091] Based on Table 1 above, the sorting result is as follows: Therefore The "two AC and one DC" interconnected distribution network was selected as the optimal architecture. This selection result is consistent with the AC / DC distribution network structure selection method disclosed in Chinese invention patent application CN121367200A, which adopts a matching conclusion based on a single cosine similarity. However, this invention further distinguishes the comprehensive advantages and disadvantages among candidate architectures by combining cosine similarity and Euclidean distance similarity, and by introducing penalty volatility and setting dynamic response capability verification correction. This provides a more comprehensive, accurate, robust, and suitable selection basis for high-proportion renewable energy access scenarios.

[0092] It should be noted that when the comprehensive scores or final comprehensive scores of two candidate architectures are very close (difference less than 1%), both can be output side by side, along with other engineering indicators (such as total investment, land area, and technology maturity) for decision-makers to weigh comprehensively. Furthermore, the output optimal architecture can be displayed using radar charts or parallel coordinate graphs to show the scores of each architecture across various dimensions (directional consistency, absolute approximation, robustness, and dynamic capabilities), intuitively presenting the differences in strengths and weaknesses.

[0093] In summary, for typical scenarios with a high proportion of renewable energy access, this study fully considers the uncertainty and time-varying characteristics of renewable energy output. It constructs a probabilistic demand vector set combining multiple time scales and a set of probabilistic scenarios. Based on this probabilistic demand vector set, time-domain simulations are performed on candidate architectures. The simulation results yield multi-dimensional capability indicators and dynamic response parameters reflecting the dynamic response capabilities of the candidate architectures. Then, a dynamic capability vector is constructed based on these multi-dimensional capability indicators, forming a dynamic capability vector set. Finally, based on the probabilistic demand vector set and the dynamic capability vector set, weighted cosine similarity and weighted Euclidean distance similarity are calculated simultaneously to match the demand indicators in the probabilistic vector set with the dynamic capability vector set in terms of both directional consistency and absolute approximation. The system extracts various capability indicators from the capability vector set and introduces directional weight coefficients to obtain a more comprehensive and accurate overall matching degree for candidate architectures. It then statistically analyzes the overall matching degree of each candidate architecture in each classic time period-scenario, calculating the mean and standard deviation of the overall matching degree. A robustness penalty factor is introduced to penalize low average overall matching degrees and high inter-scenario fluctuations. The cumulative robust matching degree is obtained through weighted summation across all time periods. Simultaneously, based on recorded dynamic response parameters, a dynamic response capability index is constructed, and combined with the cumulative robust matching degree to calculate the overall score of the candidate architectures. This ensures that the optimal architecture, ranked by its overall score, possesses both long-term comprehensive adaptability and the ability to cope with rapid power fluctuations under high-proportion renewable energy access. Furthermore, by combining the overall score of the candidate architectures with the dynamic response capability verification results, the final overall score of the candidate architectures is dynamically adjusted. This further distinguishes the actual performance of different candidate architectures during rapid fluctuations in renewable energy output (such as photovoltaic ramp-up), compensating for the shortcomings of relying solely on steady-state power flow, which cannot reflect dynamic characteristics such as energy storage speed and converter delay. This further ensures that the finally selected optimal architecture possesses the ability to cope with rapid power fluctuations under high-proportion renewable energy access.

[0094] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0095] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0096] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0097] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0098] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0099] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for matching the architecture of a hybrid distribution network adapted to a high proportion of renewable energy access, characterized in that, include: Based on the high proportion of new energy access scenarios and their output characteristics, a probabilistic demand vector set is constructed; Based on the probability demand vector set, time-domain simulation calculations are performed on the candidate architecture to extract multi-dimensional capability indicators to construct a dynamic capability vector set and record dynamic response parameters; Based on the probability demand vector set and the dynamic capability vector set, weighted cosine similarity and weighted Euclidean distance similarity are calculated respectively, and directional weight coefficients are introduced to obtain the comprehensive matching degree of the candidate architecture. Based on the comprehensive matching degree, considering the characteristics of new energy output, a robustness penalty factor is introduced to obtain the cumulative robust matching degree of the candidate architecture. Based on the dynamic response parameters, a dynamic response capability index is constructed. The comprehensive score of candidate architectures is defined as the product of cumulative robustness matching degree and dynamic response capability index. The architectures are then ranked according to the comprehensive score to obtain the optimal architecture scheme.

2. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 1, characterized in that, The probabilistic demand vector set is constructed based on the high proportion of new energy access scenarios and their output characteristics; including: Based on historical meteorological and load data, typical time periods at multiple time scales are divided; Based on historical renewable energy output and load data for typical periods, probabilistic scenarios are generated; Based on typical time periods and their probabilistic scenarios, multi-dimensional demand indicators of the architecture are selected to construct a probabilistic demand vector and obtain a probabilistic demand vector set.

3. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access as described in claim 1, characterized in that, The step of performing time-domain simulation calculations on candidate architectures, extracting multi-dimensional capability indicators to construct a dynamic capability vector set, and recording dynamic response parameters includes: Based on the probability demand vector set, time-domain simulation calculations are performed on the candidate architecture to obtain the simulation results of the candidate architecture. Based on simulation results, multi-dimensional capability indicators of candidate architectures are extracted to construct a dynamic capability vector set, and the dynamic response parameters of candidate architectures are recorded.

4. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 1, characterized in that, The calculation of weighted cosine similarity and weighted Euclidean distance similarity based on the probability demand vector set and the dynamic capability vector set includes: Based on the attributes of each indicator in the probability demand vector set and the dynamic capability vector set, classification and normalization processing is performed. Based on the classification and normalization results, weighted cosine similarity and weighted Euclidean distance similarity are calculated respectively. The weighted cosine similarity represents the directional consistency between the probability demand vectors and dynamic capability vectors in the probability demand vector set and the dynamic capability vector set; the weighted Euclidean distance similarity represents the absolute difference between the probability demand vectors and dynamic capability vectors in the probability demand vector set and the dynamic capability vector set.

5. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 3 or 4, characterized in that, Based on the comprehensive matching degree, considering the characteristics of new energy output, a robustness penalty factor is introduced to obtain the cumulative robust matching degree of the candidate architecture, including: Based on the overall matching degree of the candidate architecture in each time period and scenario, the mean and standard deviation of the overall matching degree are calculated. Based on the mean and standard deviation, a robustness penalty factor is introduced, and the cumulative robustness matching degree of the candidate architecture is obtained by weighting the sum by time period.

6. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 1, characterized in that, The construction of dynamic response capability indicators based on dynamic response parameters includes: Based on the maximum voltage change rate and the average converter response delay, the influence coefficients of the voltage change rate and the response delay are introduced to construct a dynamic response capability index.

7. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 6, characterized in that, The construction of dynamic response capability indices based on dynamic response parameters also includes: A preset dynamic performance qualification threshold is set, and the dynamic response capability of candidate architectures is verified by combining dynamic response capability indicators. The dynamic response capability verification includes: comparing the dynamic response capability index of the candidate architecture with the dynamic performance qualification threshold, and setting an elimination mechanism to eliminate candidate architectures that fail the dynamic response capability verification.

8. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 1 or 7, characterized in that, The process of ranking based on comprehensive scores to obtain the optimal architecture scheme includes: Calculate the comprehensive score of the candidate architectures and rank them, and obtain the optimal architecture scheme based on the ranking results; The architecture scheme includes one or more of the following: architecture identifier, comprehensive architecture score, and recommended architecture ranking table.

9. The method for matching a hybrid distribution network architecture adapted to a high proportion of new energy access according to claim 8, characterized in that, The process of obtaining a comprehensive score for candidate architectures and ranking them based on the comprehensive score also includes: Based on the comprehensive score of the candidate architectures and the results of dynamic response capability verification, the final comprehensive score of the candidate architectures is dynamically adjusted, and they are then ranked according to the final comprehensive score.

Images