Power distribution network sub-region comprehensive load modeling method and device

Abstract

The application provides a power distribution network sub-region comprehensive load modeling method and device, and relates to the technical field of power distribution. The method comprises the following steps: in the constructed to-be-simulated power distribution network model comprising multiple operation scenarios, different disturbance events are applied; the characteristic parameters of each node are extracted; clustering partition is performed based on the characteristic parameters, and multiple sub-power distribution network regions are obtained; a comprehensive load model is constructed for each sub-power distribution network region, wherein the comprehensive load model comprises a static load ZIP model, an induction motor model, a distributed photovoltaic model and an electric vehicle charging station model; according to a preset parameter stability index, the values of the to-be-identified parameters of the comprehensive load model corresponding to each sub-power distribution network region are determined; and based on the values of the identified parameters of the comprehensive load model corresponding to each sub-power distribution network region, a dynamic simulation component is output. The identification result obtained by the modeling method provided by the application has good stability.

Description

Technical Field

[0001] This invention relates to the field of power distribution technology, and in particular to a method and equipment for regional integrated load modeling of power distribution networks. Background Technology

[0002] With the continuous development of technology, my country's power distribution network is gradually transforming into a new type of active power distribution network. On the one hand, the proportion of distributed power sources, represented by photovoltaics, is increasing rapidly, and on the other hand, new loads, including electric vehicles, are constantly increasing.

[0003] In related technologies, the method of simplifying the entire network load into a single integrated load model is difficult to accurately describe the dynamic behavior of the distribution network under fault disturbances. Currently, the integrated load models used are mostly based on the ZIP static load model and the induction motor model. The ZIP static load model is a linear combination of impedance, constant current, and constant power. By reasonably setting the proportions of these three, it can simulate the voltage and power characteristics of different types of loads in steady state relatively accurately. The induction motor model can be used to simulate asynchronous motor loads, including electromechanical transients and electromagnetic transient processes.

[0004] However, the inventors discovered that current integrated load modeling methods typically perform uniform identification on the entire network or a single node. For different operating scenarios, all parameters need to be re-identified, resulting in a large computational load. Furthermore, current parameter identification results usually use the error of a single identification as the evaluation criterion, rarely conducting systematic analysis of the statistical stability of the identified parameters. Consequently, the identification results obtained by current modeling methods have poor stability. Summary of the Invention

[0005] This invention provides a method and equipment for regional integrated load modeling of distribution networks to solve the problem of poor stability of identification results obtained by current modeling methods.

[0006] In a first aspect, embodiments of the present invention provide a method for regional integrated load modeling of a distribution network, including: Different disturbance events are applied to the constructed power distribution network model to be simulated, which includes multiple operating scenarios; Extract the feature parameters of each node; Clustering and partitioning are performed based on the aforementioned feature parameters to obtain multiple sub-distribution network areas; A comprehensive load model is constructed for each sub-distribution network area. The comprehensive load model includes a static load ZIP model, an induction motor model, a distributed photovoltaic model, and an electric vehicle charging station model. Based on the preset parameter stability index, the values ​​of the parameters to be identified for the integrated load model corresponding to each sub-distribution network area are determined respectively. Based on the numerical values ​​of the integrated load model with identification parameters corresponding to each sub-distribution network area, a dynamic simulation component is output.

[0007] In one possible implementation, the feature parameters include dynamic response features, resource distribution features, and a continuous dynamic response sequence; The clustering and partitioning based on the feature parameters yields multiple sub-distribution network areas, including: Based on the dynamic response characteristics and the resource distribution characteristics, the scalar characteristic distance of each node is determined; Based on the continuous dynamic response sequence, the dynamic trajectory distance of each node is determined; Based on the scalar feature distance of each node and the dynamic trajectory distance, a comprehensive distance matrix is ​​constructed; Based on agglomerative hierarchical clustering, the comprehensive distance matrix is ​​partitioned to obtain the multiple sub-distribution network areas.

[0008] In one possible implementation, the continuous dynamic response sequence includes a voltage response time sequence, an active power response time sequence, and a reactive power response time sequence. The determination of the scalar feature distance of each node based on the dynamic response characteristics and the resource distribution characteristics includes: The dynamic response characteristics and the resource distribution characteristics are normalized to obtain a comprehensive feature vector; Based on the Euclidean distance and the comprehensive feature vector, the scalar feature distance of each node is determined; Determining the dynamic trajectory distance of each node based on the continuous dynamic response sequence includes: Based on the dynamic time adjustment algorithm, the nonlinear aligned shortest path distances of the voltage response time series, active power response time series and reactive power response time series are calculated respectively. The dynamic trajectory distance is determined based on the mean of the nonlinear aligned shortest path distances of the voltage response time series, active power response time series, and reactive power response time series.

[0009] In one possible implementation, the partitioning of the comprehensive distance matrix based on agglomerative hierarchical clustering to obtain the multiple sub-distribution network areas includes: Calculate the values ​​of each evaluation index for each number of candidate regions; The optimal number of zones for the distribution network is determined by summing the values ​​of all evaluation indicators corresponding to each candidate zone number. Based on the optimal number of partitions, the multiple sub-distribution network areas are obtained.

[0010] In one possible implementation, the multiple operating scenarios include a baseline scenario and non-baseline scenarios; The step of determining the values ​​of the parameters to be identified for the integrated load model corresponding to each sub-distribution network area based on preset parameter stability indices includes: Determine the parameters to be identified in the integrated load model; wherein, the parameters to be identified include multiple parameters; Under the baseline scenario, the parameters to be identified in the integrated load model corresponding to each sub-distribution network area are identified to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area; wherein, the parameters to be identified include structural parameters and scenario parameters. In the non-benchmark scenario, the values ​​of the scenario parameters are updated until the parameter stability index meets the set conditions; wherein, the non-benchmark scenario includes a high photovoltaic penetration rate scenario, a high electric vehicle charging station ratio scenario, and a hybrid dual high penetration rate scenario, and the values ​​of the structural parameters are fixed in the non-benchmark scenario.

[0011] In one possible implementation, the step of identifying the parameters to be identified in the integrated load model corresponding to each sub-distribution network area under the baseline scenario, to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area, includes: In the baseline scenario, the sum of the root mean square errors of the tested active power waveform and reactive power waveform with the output waveform of each integrated load model is used as the fitness function. Based on the improved particle swarm optimization algorithm, the numerical values ​​of the structural parameters and scenario parameters of each integrated load model are obtained.

[0012] In one possible implementation, updating the values ​​of the scenario parameters in the non-baseline scenario until the parameter stability index meets a set condition includes: Based on changing the value of the trust region in the trust region algorithm, the parameter stability index of all scene parameters under the target non-benchmark scenario is calculated. If the calculated stability index does not meet the preset condition, then all scene parameters under the target non-baseline scenario are updated until the calculated parameter stability index meets the preset condition. Then, the value of the scene parameter that meets the preset condition is determined as the value of all scene parameters under the target non-baseline scenario; wherein, the target non-baseline scenario is any one of the non-baseline scenarios.

[0013] In one possible implementation, the parameter stability index includes the coefficient of variation of the scene parameters and / or the confidence interval width of the coefficient of variation of the scene parameters within a set value.

[0014] In one possible implementation, the multiple operating scenarios include a scenario where photovoltaic output accounts for a greater than or equal to a first set ratio of the total load, a scenario where electric vehicle charging load accounts for a greater than or equal to a second set ratio of the total load, a scenario where photovoltaic output accounts for a greater than or equal to the first set ratio of the total load and electric vehicle charging load accounts for a greater than or equal to the second set ratio of the total load, and a baseline scenario.

[0015] Secondly, embodiments of the present invention provide a distribution network regional integrated load modeling device, comprising: The disturbance module is used to apply different disturbance events to the constructed power distribution network model to be simulated, which includes multiple operating scenarios; The extraction module is used to extract the feature parameters of each node; The partitioning module is used to perform clustering and partitioning based on the feature parameters to obtain multiple sub-distribution network areas; The modeling module is used to construct integrated load models for each sub-distribution network area. The integrated load models include static load ZIP models, induction motor models, distributed photovoltaic models, and electric vehicle charging station models. The parameter determination module is used to determine the values ​​of the parameters to be identified in the integrated load model corresponding to each sub-distribution network area based on the preset parameter stability index. The output module is used to output dynamic simulation components based on the numerical values ​​of the integrated load model with identification parameters corresponding to each sub-distribution network area.

[0016] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0017] In this embodiment of the invention, after building a distribution network model covering various operating scenarios, different disturbance events can be injected to simulate distribution network faults under multiple scenarios, effectively supporting fault assessment and analysis. By extracting characteristic parameters of each node in the power grid and conducting cluster analysis, the entire distribution network is divided into multiple sub-distribution network zones, further improving the accuracy of the overall distribution network analysis. Simultaneously, a suitable comprehensive load model is built for each sub-zone, and the values ​​of the parameters to be identified in each model are determined based on preset parameter stability indicators. This significantly improves parameter identification efficiency and ensures stable and reliable identification results, ultimately outputting a distribution network dynamic simulation component that meets accuracy standards. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the implementation of the regional integrated load modeling method for distribution networks provided in this embodiment of the invention. Figure 2This is a schematic diagram of the integrated load model topology provided in an embodiment of the present invention; Figure 3 This is a flowchart of the improved particle swarm optimization algorithm for offline full parameter identification provided in this embodiment of the invention; Figure 4 This is a flowchart of the lightweight parameter update and statistical feedback optimization process of the trust region reflection algorithm provided in this embodiment of the invention; Figure 5 This is a schematic diagram of the structure of the distribution network regional integrated load modeling device provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of a module provided in an embodiment of the present invention that can be directly called by the power system simulation software PSASP; Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] As described in the background section, current integrated load models are mostly based on ZIP static load models and induction motor models, which do not adequately consider the fault ride-through control behavior of distributed photovoltaic inverters and the rapid response behavior of electric vehicle charging stations. Furthermore, existing load modeling methods typically perform uniform identification on a network-wide or single-node basis. For different operating scenarios, these methods often require re-identification of all parameters, resulting in a large computational load. Additionally, existing parameter identification results usually use single-shot identification errors as the evaluation criterion, with little systematic analysis of the statistical stability of the identified parameters.

[0021] To address the aforementioned technical problems, this invention provides a method and equipment for regional integrated load modeling of power distribution networks.

[0022] See Figure 1 The document illustrates a flowchart of the implementation of the regional integrated load modeling method for distribution networks provided in this embodiment of the invention, which is described in detail below: S110. Apply different disturbance events to the constructed power distribution network model to be simulated, which includes multiple operating scenarios.

[0023] In this embodiment, an IEEE 33-node active distribution network system with a high proportion of distributed photovoltaic power and electric vehicles can be built in the MATLAB / Simulink simulation platform. To maintain the simulation stability of the distribution network under asymmetrical faults and complex operating conditions, the transformer winding connection groups in the model are uniformly configured as Yd1 and Dy11. Static loads, induction motors, distributed photovoltaic power stations, and electric vehicle charging stations of different capacities are distributed and connected at different nodes of the distribution network.

[0024] In some embodiments, multiple operating scenarios may include a scenario where photovoltaic output accounts for a greater than or equal to a first set ratio of the total load, a scenario where electric vehicle charging load accounts for a greater than or equal to a second set ratio of the total load, a scenario where photovoltaic output accounts for a greater than or equal to the first set ratio of the total load and electric vehicle charging load accounts for a greater than or equal to the second set ratio of the total load, and a baseline scenario.

[0025] In this embodiment, various operating scenarios can include: a high photovoltaic penetration scenario, where photovoltaic output accounts for 60% of the total load; a high electric vehicle ratio scenario, where EV charging load accounts for 40% of the total load; and a hybrid high penetration scenario, where photovoltaic output accounts for 60% of the total load and EV charging load accounts for 40% of the total load. The standard benchmark scenario is one where photovoltaic output accounts for less than 60% of the total load and EV charging load accounts for less than 40% of the total load, primarily consisting of traditional rigid loads such as industrial loads, residential daily electricity consumption, and commercial regular electricity consumption, with a stable load curve, no significant fluctuations in new energy sources, and no concentrated charging and discharging impacts from EVs.

[0026] In some embodiments, a three-phase short-circuit fault disturbance event can be set on the main grid side of the system, with the fault voltage drop depth set to 0.2 pu, 0.5 pu and 0.8 pu, respectively, and the fault duration set to 0.1 s and 0.2 s.

[0027] S120. Extract the feature parameters of each node.

[0028] Among them, the characteristic parameters include dynamic response characteristics, resource distribution characteristics, and continuous dynamic response sequences.

[0029] In this embodiment, the dynamic response features of the extracted node include the active power slope, reactive power slope, total reactive power peak value during the fault, and the time required for reactive power to recover to 90% of the steady-state value after the fault is cleared.

[0030] The resource distribution characteristics of the extracted nodes include the photovoltaic penetration rate under the current operating mode, the proportion of charging station capacity to the total load of the node, and the basic value of the node's total load capacity.

[0031] Extract the continuous dynamic response sequence of the node during the fault, including the voltage U(t) response time sequence, the active power P(t) response time sequence, and the reactive power Q(t) response time sequence, to form a dataset.

[0032] S130. Clustering and partitioning based on characteristic parameters to obtain multiple sub-distribution network areas.

[0033] In some embodiments, the scalar characteristic distance of each node can be determined first based on dynamic response characteristics and resource distribution characteristics. Next, the dynamic trajectory distance of each node can be determined based on the continuous dynamic response sequence. Then, a comprehensive distance matrix is ​​constructed based on the scalar characteristic distance and dynamic trajectory distance of each node. Finally, the comprehensive distance matrix is ​​partitioned according to agglomerative hierarchical clustering to obtain multiple sub-distribution network areas.

[0034] In this embodiment, a comprehensive feature vector can be obtained by normalizing the dynamic response characteristics and resource distribution characteristics. Then, the scalar feature distance of each node is determined based on the Euclidean distance and the comprehensive feature vector.

[0035] For example, by normalizing the dynamic response characteristics and resource distribution characteristics, a comprehensive scalar feature vector is constructed for node i and node j. The scalar feature distance between the feature vectors of each node is then calculated using Euclidean distance. .

[0036] In this embodiment, a dynamic time adjustment algorithm can be used to calculate the nonlinear aligned shortest path distances for the voltage response time series, active power response time series, and reactive power response time series, respectively. Then, the dynamic trajectory distance is determined based on the mean of the nonlinear aligned shortest path distances for the voltage response time series, active power response time series, and reactive power response time series.

[0037] For example, for the continuous dynamic response sequences of nodes i and j during a fault, including the voltage U(t) response time sequence, the active power P(t) response time sequence, and the reactive power Q(t) response time sequence, the Dynamic Time Warping (DTW) algorithm is used to calculate the nonlinear alignment shortest path distance between the time sequences, and the average of the DTW distances of the three physical quantities is taken to obtain the dynamic trajectory distance. .

[0038] Next, by introducing weighting coefficients and ,satisfy The scalar feature distance and the dynamic trajectory distance are linearly weighted and fused to obtain the final weighted comprehensive distance matrix between nodes. .

[0039] In some embodiments, after obtaining the weighted composite distance matrix, the values ​​of each evaluation index under each candidate region number can be calculated first. Then, based on the sum of the values ​​of all evaluation indices corresponding to each candidate region number, the optimal number of distribution network zones is determined. Finally, based on the optimal number of zones, multiple sub-distribution network zones are obtained.

[0040] In this embodiment, the weighted comprehensive distance matrix can be first... This serves as the input matrix for the clustering algorithm. Initially, each load node is treated as an independent cluster. By recursively calculating the distance between clusters and continuously merging the closest clusters, a clustering dendrogram reflecting the dynamic characteristics of the distribution network is generated.

[0041] Then, calculate a single evaluation index: define the optimization range for the number of candidate regions K. The silhouette coefficients SC, CH index, and DB index are calculated for each K value.

[0042] Next, the range normalization method was used to process the profile coefficients SC, CH index, and DB index.

[0043] Finally, iterate through all candidate K values ​​and select the one that maximizes the comprehensive index F(K). The optimal number of partitions for the distribution network is used to output the final partitioning scheme.

[0044] Specifically, the silhouette coefficient SC characterizes the intra-cluster compactness and inter-cluster separation; a higher value is better.

[0045] The formula for calculating the profile coefficient SC is: ; Where N is the total number of distribution network load nodes participating in the clustering; The average distance from node i to all other nodes in the same cluster reflects the cluster density of the cluster to which the node belongs; is the average distance from node i to all nodes in the nearest other cluster, reflecting the inter-cluster separation of this node from other clusters; Indicates taking and The maximum value in the denominator is used for normalization.

[0046] Specifically, the Calinski-Harabasz (CH) index is the ratio of the trace of the inter-class variance matrix to the trace of the intra-class variance matrix; the larger the value, the better the clustering effect.

[0047] The formula for calculating the Calinski-Harabasz (CH) index is: ; Where N is the total number of distribution network load nodes participating in the clustering; The trace of the inter-class scatter matrix is ​​the sum of the elements on the main diagonal of the matrix, which reflects the distance of the centroid of each cluster from the global centroid. The trace of the intra-cluster discreteness matrix reflects the distance from each node in the cluster to its own cluster center; and These represent the degrees of freedom within and between classes, respectively, and are used as penalty coefficients to eliminate the natural bias in the calculation results caused by the increase in the number of clusters K, ensuring that the CH indices under different numbers of partitions are absolutely comparable.

[0048] Specifically, the Davies-Bouldin (DB) index is the average of the maximum similarity between any two clusters. The smaller the value, the better the separation between clusters.

[0049] The formula for calculating the Davies-Bouldin (DB) index is as follows: ; Where i and j are different cluster indices and ; and These are the intra-class average dispersion of the i-th and j-th clusters, respectively, which are the average distances from all nodes in each cluster to their cluster center, reflecting the degree of internal looseness of the cluster; The distance between the i-th cluster center and the j-th cluster center is the Euclidean distance, which reflects the degree of separation between the two clusters.

[0050] To eliminate differences in dimensions and numerical magnitudes among different clustering evaluation indicators and ensure the effectiveness of weight allocation, the range normalization method can be used to process each indicator. The specific processing procedure is as follows: We can define the optimization range set of the candidate region number K as follows: Iterate through all the index values ​​obtained from the set, find their maximum and minimum values ​​respectively, and perform the following normalization calculation: The normalization formula for the profile coefficient SC is: In the formula, The contour coefficient when the current number of partitions is K; and These are the maximum and minimum profile coefficients corresponding to all candidate K values ​​within the optimization range, respectively.

[0051] The normalization formula for the Calinski-Harabasz (CH) index is: In the formula, The contour coefficient when the current number of partitions is K; and These are the maximum and minimum profile coefficients corresponding to all candidate K values ​​within the optimization range, respectively.

[0052] The normalization formula for the Davies-Bouldin (DB) index is: In the formula, The contour coefficient when the current number of partitions is K; and These are the maximum and minimum profile coefficients corresponding to all candidate K values ​​within the optimization range, respectively.

[0053] The optimal number of partitions is determined by constructing a weighted comprehensive evaluation index based on the silhouette coefficient SC, CH index, and DB index; the silhouette coefficient SC, CH index, and DB index obtained in step 3.3 are normalized and then input into the comprehensive evaluation function F(K). in, , , These are the normalized index values; , , The set weighting coefficients satisfy... .

[0054] S140: Construct a comprehensive load model for each sub-distribution network area.

[0055] The integrated load model includes a static load ZIP model, an induction motor model, a distributed photovoltaic model, and an electric vehicle charging station model.

[0056] like Figure 2 As shown, the ZIP section characterizes the voltage dependence of static loads; the induction motor section characterizes the electromechanical transient dynamic behavior; the distributed photovoltaic (DGPV) section characterizes the active power output variation under voltage disturbances and the reactive power support characteristics of low-voltage ride-through; and the electric vehicle (EV) section characterizes the first-order constant power charging response characteristics with time delay. PCC represents the point of common coupling, which is the interface between the distribution network and the upper-level power grid. POC is the virtual bus, which is the common connection bus for all power sources and loads within the microgrid. RD+jXD is the line impedance from PCC to the transformer, where RD is the line resistance and XD is the line reactance. RT+jXT represents the distribution transformer between the high-voltage side of PCC and the low-voltage side of POC, where RT is the transformer resistance and XT is the transformer leakage reactance.

[0057] In this embodiment, the static load ZIP model is as follows: ; in, , , These represent the active and reactive power of the load and the amplitude of the load bus voltage under the operating conditions at the reference point. , , , , , These represent the proportions of the three components: constant impedance, constant current, and constant power. , .

[0058] In this embodiment, the third-order induction motor model is as follows: ; ; ; ; in, and The transient electromotive force of the induction motor consists of d-axis and q-axis components. and These are the d-axis and q-axis components of the stator current; This refers to the motor speed; This refers to the rotor speed; For rotor mechanical torque, is the inertial time constant; A, B, and C represent the rotational speed proportionality coefficients, and A+B+C=1; Stator resistance; For the rotor steady-state reactance, For rotor transient reactance, The rotor winding time constant; For stator reactance; For magnetizing reactance; For rotor reactance; This represents the rotor resistance.

[0059] In this embodiment, the distributed photovoltaic model is as follows: ; ; ; ; in, The active power of the load is under the reference point operating conditions. Under the reference point operating conditions, the photovoltaic power generation system operates with a unity power factor and the reactive power is 0. For the damping ratio, For natural frequency, U is the proportional control coefficient, and U is the per-unit voltage value at the grid connection point. This represents the steady-state value corresponding to U; This represents the capacity limit of the photovoltaic model. This represents the dynamic reactive power support factor during a fault.

[0060] In this embodiment, the electric vehicle charging station model is as follows: ; ; ; ; in, U represents the active power of the load under reference operating conditions, where U is the per-unit voltage value at the grid connection point. This is the constant power current command value. This is the equivalent first-order response time constant of the charging station.

[0061] Finally, the static load ZIP model, induction motor model, distributed photovoltaic model, and electric vehicle charging station load model are connected in parallel at the 10kV load bus to form a new generalized integrated load model, which aggregates the total injected power. and for: ; When the new generalized integrated load model is applied to the high-voltage side, considering the very small impedance of the distribution transformer, its power loss can be ignored; considering the losses from the distribution lines, the impedance of the distribution lines is... The power converted to the high-voltage side is: ; ; ; ; ; in, and These are the per-unit values ​​of resistance and reactance of the equivalent impedance of the power distribution line, respectively. This is the per-unit voltage value measured at the grid connection point. This represents the per-unit voltage value at the load busbar. The equivalent resistance of the distribution line. and equivalent reactance These are treated as known network topology constants, directly calculated from the physical parameters of the distribution network lines, and do not participate in subsequent identification.

[0062] S150. Based on the preset parameter stability index, determine the values ​​of the parameters to be identified for the integrated load model corresponding to each sub-distribution network area.

[0063] Among them, the parameter stability index includes the coefficient of variation of the scene parameters and / or the confidence interval width of the coefficient of variation of the scene parameters within a set value.

[0064] In some embodiments, the parameters to be identified for the integrated load model need to be determined before identification can be performed.

[0065] In this embodiment, the static load ZIP model and the induction motor model have a total of 12 independent parameters. , , , , , , , , , , , The distributed photovoltaic model has four independent parameters. , , , The load of an electric vehicle charging station has one independent parameter. In addition, four parameters need to be added, namely... Used to define the ratio of the system reference value to the induction motor reference value. Used to define the proportion of active power occupied by an induction motor. Used to characterize the proportion of distributed photovoltaic output power to total active power. Used to characterize the proportion of power absorbed by an electric vehicle charging station to the total active power.

[0066] In some embodiments, the parameters to be identified include structural parameters and scenario parameters. First, under a baseline scenario, the parameters to be identified in the integrated load model corresponding to each sub-distribution network area are identified to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area. Then, under non-baseline scenarios, the values ​​of the scenario parameters are updated until the parameter stability index meets the set conditions.

[0067] Among them, non-benchmark scenarios include high photovoltaic penetration scenarios, high electric vehicle charging station ratio scenarios, and mixed dual high penetration scenarios. The values ​​of structural parameters are fixed in non-benchmark scenarios.

[0068] In this embodiment, the parameters to be identified can be divided into structural parameters and scene parameters. Structural parameters characterize the inherent mechanisms of the region model. Scene parameters characterize the resource composition and dynamic response characteristics under changing operating scenarios. Structural parameters include... , , , , , , , , , , , , , Scene parameters include , , , , , .

[0069] In this embodiment, under the baseline scenario, the sum of the root mean square errors of the tested active power waveform and reactive power waveform with the output waveform of each integrated load model can be used as the fitness function. Based on the improved particle swarm optimization algorithm, the values ​​of the structural parameters and scenario parameters of each integrated load model can be obtained.

[0070] Specifically, in the baseline scenario, full-parameter offline identification is performed on the models of each region. The fitness function is the sum of the RMSE of the measured active and reactive power waveforms and the model output waveform. An improved particle swarm optimization (IPSO) algorithm is used for global search to obtain the initial values ​​of the structural parameters and scene parameters of the region models.

[0071] IPSO enhances its ability to escape local optima by adjusting the particle position and velocity update formulas, thereby obtaining structural and scene parameters, such as... Figure 3 As shown, the specific steps of the improved particle swarm optimization (IPSO) algorithm are as follows: First, initialize the population. Determine the parameters that need to be identified for the integrated load model. Within the physical boundary, randomly initialize the position vector Xi and velocity vector Vi of N particles. The position vector Xi represents a set of potential model parameter solutions.

[0072] Then, the fitness value is calculated. The measured low-voltage side load bus voltage is substituted into the generalized integrated load model corresponding to each particle, and the differential algebraic equation is solved by the Runge-Kutta method to obtain the active power and reactive power output by the model. The sum of the root mean square error (RMSE) of the measured power and the simulated power is used as the fitness function.

[0073] ; Next, update the optimal solution. Compare the current particle's fitness value to the previous best position and update the individual's historical best position. and the global optimal position of the entire population .

[0074] Next, adaptive parameters are calculated. To balance global search and local exploitation capabilities, inertia weights that decrease non-linearly with the number of iterations are calculated. and the acceleration factor for asynchronous changes and The updated formula is: ; ; ; in, This represents the maximum number of iterations.

[0075] Then, position and velocity are updated. The particle's velocity and position are updated according to the IPSO core operator. After the update, if the particle's position is found to exceed the set physical boundary, it is forcibly confined to the boundary and its velocity is reversed. The update formula is: ; ; in, , for A random number between [a certain number of points].

[0076] Finally, iteration and output. Continuously update the individual optimal value. and global optimal This continues until the maximum number of iterations is reached or the error meets the preset precision. The output... These are the structural parameters for that region, which are fixed in subsequent scenes. Otherwise, the adaptive parameters are recalculated and the loop continues.

[0077] In some embodiments, in a non-baseline scenario, the parameter stability index of all scene parameters in the target non-baseline scenario is calculated by changing the value of the trust region in the trust region algorithm. If the calculated stability index does not meet the preset conditions, all scene parameters in the target non-baseline scenario are updated until the calculated parameter stability index meets the preset conditions. Then, the value of the scene parameter that meets the preset conditions is determined as the value of all scene parameters in the target non-baseline scenario.

[0078] The target non-benchmark scenario is any scenario among the non-benchmark scenarios.

[0079] In this embodiment, a trust region reflection algorithm can be used. This algorithm defines a trust region in each iteration and searches for the optimal solution of a quadratic approximation model of the objective function within it, such as... Figure 4 As shown, it is particularly suitable for identifying load model parameters with physical boundary constraints, ensuring the physical consistency of equal parameters.

[0080] When switching to a high photovoltaic penetration operation scenario, the structural parameters obtained in the above steps are fixed. A set of scenario-related feature parameters is extracted as the variables to be identified, and the identification results obtained in the above steps are used as the initial values ​​for iteration. The steps for other operation scenarios are the same as those for the high photovoltaic penetration operation scenario.

[0081] Furthermore, multiple repetitive identifications were performed for the same scenario, and parameter stability indices were calculated. First, for the regional model in a high photovoltaic penetration scenario, the initial values ​​of the trust region algorithm were changed, and the lightweight parameter update identification was repeated 50 times. The scenario parameters obtained after each identification converged were recorded, and a statistical sample set for each scenario parameter was constructed. ,in Let be the parameter value obtained from the m-th identification. Then, calculate the scene parameters. , , , , , Statistical indicators included: mean, standard deviation, coefficient of variation (CV), and upper and lower bounds of the 95% confidence interval. Finally, statistical methods were used to determine parameter stability. If the equivalent response time constant of the electric vehicle charging station is found... If the coefficient of variation is greater than the set threshold of 15% or its 95% confidence interval is too wide, it indicates that the parameter is unstable in the current scenario and is marked as an unstable parameter.

[0082] Furthermore, when the parameter stability index does not meet the preset conditions, feedback optimization is performed and the scene parameter update is re-executed. Feedback optimization is performed on unstable parameters, and calculations are performed. Regarding the trajectory sensitivity of the output power, if the sensitivity is below a threshold, it should be decoupled from the model or set to a typical value. If local optimal oscillations exist, the high-frequency distribution area should be narrowed based on previous data. Find the search boundary. Repeat the lightweight parameter update steps above until the stability metrics of all scene parameters meet the requirements.

[0083] S160. Based on the numerical values ​​of the integrated load model with identification parameters corresponding to each sub-distribution network area, output dynamic simulation components.

[0084] After obtaining the values ​​of the identified parameters of the integrated load model corresponding to each sub-distribution network area, the integrated load models corresponding to all sub-distribution network areas can be integrated to form the overall equivalent model of the entire distribution network.

[0085] The overall equivalent model includes the integrated load model and the parameter-weighted overall equivalent model corresponding to each sub-distribution network area, and outputs... The stable parameter set of each sub-region will The regional models are connected and aggregated in parallel at the main grid interface to form a parallel aggregated model of the regional sub-models of the entire distribution network. Output The stable parameter set of each sub-region will The parameters of each regional model are weighted and averaged to form a weighted overall equivalent model of the entire distribution network.

[0086] ; ; in, Identify the model parameters for sub-region i. This refers to the weighted weight corresponding to this sub-region. This represents the total load capacity of the sub-region.

[0087] To verify the self-descriptive and generalization capabilities of the modeling method provided in this invention, the following models can be constructed for comparative verification, with evaluation metrics including RMSEP, RMSEQ, and parameter variation coefficient.

[0088] Model 1: A single integrated load model for the entire distribution network, without zoning.

[0089] Model 2: Regional model, but full parameter identification is performed for each scene; Model 3: Regional model with lightweight parameter updates, but without statistical feedback; Model 4: Regional model with lightweight parameter updates and statistical feedback optimization. Model 4 is a model built based on the modeling method provided in this invention.

[0090] By using the above three evaluation indicators for verification, it can be determined that the modeling method provided by the present invention has good stability.

[0091] In some embodiments, structural parameters and scenario parameters are serialized and encapsulated according to the model data structure commonly used in simulation software, and labeled according to scenario characteristics. The module input interface is configured to receive real-time voltage per-unit vectors issued by the simulation software, and the module output interface is configured to return the calculated equivalent node complex power values. Through the underlying compilation engine, the novel generalized integrated load differential algebraic equation system, which includes an internal voltage reconstruction algorithm, is compiled into a user-defined model to enable direct calls with mainstream power system simulation platforms.

[0092] In this embodiment of the invention, after building a distribution network model covering various operating scenarios, different disturbance events can be injected to simulate distribution network faults under multiple scenarios, effectively supporting fault assessment and analysis. By extracting characteristic parameters of each node in the power grid and conducting cluster analysis, the entire distribution network is divided into multiple sub-distribution network zones, further improving the accuracy of the overall distribution network analysis. Simultaneously, a suitable comprehensive load model is built for each sub-zone, and the values ​​of the parameters to be identified in each model are determined based on preset parameter stability indicators. This significantly improves parameter identification efficiency and ensures stable and reliable identification results, ultimately outputting a distribution network dynamic simulation component that meets accuracy standards.

[0093] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0094] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0095] Figure 5 A schematic diagram of the distribution network regional integrated load modeling device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below: like Figure 5 As shown, the distribution network regional integrated load modeling device includes: The disturbance module 510 is used to apply different disturbance events to the constructed distribution network model to be simulated, which includes multiple operating scenarios; Extraction module 520 is used to extract feature parameters of each node; The partitioning module 530 is used to perform clustering partitioning based on the feature parameters to obtain multiple sub-distribution network areas; Modeling module 540 is used to construct integrated load models for each sub-distribution network area. The integrated load models include static load ZIP models, induction motor models, distributed photovoltaic models, and electric vehicle charging station models. The parameter determination module 550 is used to determine the values ​​of the parameters to be identified in the integrated load model corresponding to each sub-distribution network area according to the preset parameter stability index. Output module 560 is used to output dynamic simulation components based on the numerical values ​​of the integrated load model with identification parameters corresponding to each sub-distribution network area.

[0096] In one possible implementation, the characteristic parameters include dynamic response characteristics, resource distribution characteristics, and continuous dynamic response sequences; Partitioning module 530 is used to determine the scalar feature distance of each node based on dynamic response characteristics and resource distribution characteristics; Based on the continuous dynamic response sequence, the dynamic trajectory distance of each node is determined; A comprehensive distance matrix is ​​constructed based on the scalar feature distance and dynamic trajectory distance of each node; Based on agglomerative hierarchical clustering, the comprehensive distance matrix is ​​partitioned to obtain multiple sub-distribution network areas.

[0097] In one possible implementation, the continuous dynamic response sequence includes a voltage response time series, an active power response time series, and a reactive power response time series. The partitioning module 530 is used to normalize the dynamic response characteristics and resource distribution characteristics to obtain a comprehensive feature vector. Based on Euclidean distance and comprehensive feature vector, the scalar feature distance of each node is determined; Based on the dynamic time adjustment algorithm, the nonlinear aligned shortest path distances of the voltage response time series, active power response time series, and reactive power response time series are calculated respectively. The dynamic trajectory distance is determined by the mean of the nonlinear aligned shortest path distances of the voltage response time series, active power response time series, and reactive power response time series.

[0098] In one possible implementation, the partitioning module 530 is used to calculate the values ​​of each evaluation index under each number of candidate regions. The optimal number of zones for the distribution network is determined by summing the values ​​of all evaluation indicators corresponding to each candidate zone number. Based on the optimal number of partitions, multiple sub-distribution network areas are obtained.

[0099] In one possible implementation, multiple operating scenarios include a baseline scenario and non-baseline scenarios; The parameter determination module 550 is used to determine the parameters to be identified in the comprehensive load model; the parameters to be identified include multiple parameters. In the baseline scenario, the parameters to be identified in the integrated load model corresponding to each sub-distribution network area are identified to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area; among which, the parameters to be identified include structural parameters and scenario parameters. In non-benchmark scenarios, the values ​​of scenario parameters are updated until the parameter stability index meets the set conditions; among them, the values ​​of structural parameters are fixed in non-benchmark scenarios.

[0100] In one possible implementation, the parameter determination module 550 is used to, under a reference scenario, use the sum of the root mean square errors of the tested active power waveform and reactive power waveform with the output waveform of each integrated load model as a fitness function, and based on an improved particle swarm optimization algorithm, obtain the values ​​of the structural parameters and scenario parameters of each integrated load model.

[0101] In one possible implementation, the parameter determination module 550 is used to calculate the parameter stability index of all scene parameters under the target non-baseline scene based on changing the value of the trust region in the trust region algorithm. If the calculated stability index does not meet the preset conditions, all scene parameters under the target non-baseline scenario will be updated until the calculated parameter stability index meets the preset conditions. Then, the value of the scene parameter that meets the preset conditions will be determined as the value of all scene parameters under the target non-baseline scenario. The target non-baseline scenario is any scenario among the non-baseline scenarios.

[0102] In one possible implementation, the parameter stability metric includes the coefficient of variation of the scene parameters and / or the confidence interval width of the coefficient of variation of the scene parameters within a set value.

[0103] In one possible implementation, multiple operating scenarios include a scenario where photovoltaic output accounts for a greater than or equal to a first set percentage of the total load, a scenario where electric vehicle charging load accounts for a greater than or equal to a second set percentage of the total load, a scenario where photovoltaic output accounts for a greater than or equal to the first set percentage of the total load and electric vehicle charging load accounts for a greater than or equal to the second set percentage of the total load, and a baseline scenario.

[0104] Furthermore, this invention also provides a module that can be directly called by the power system simulation software PSASP, such as... Figure 6 As shown, it includes a data and identification layer, a model encapsulation and output layer, and a third-party simulation application layer.

[0105] The data and identification layer, including the parameter determination module in the distribution network regional integrated load modeling device, integrates an improved particle swarm optimization (IPSO) algorithm and a trust region reflection algorithm solution engine written in C++. As the core operator of the system, this module is responsible for receiving the measured current and voltage sequences transmitted by the extraction module, automatically performing parameter optimization calculations under multiple scenarios in the background, and outputting the stable parameter set for each sub-region.

[0106] The model encapsulation and output layer, including the modeling module, translates the differential-algebraic equations of a novel generalized integrated load model, encompassing ZIP static loads, induction motors, distributed photovoltaics, and electric vehicles, into UDM model description statements recognizable by PSASP. This module utilizes a compiler to couple the aforementioned model logic with the identified set of stable parameters, generating a dynamic link library file. This file encapsulates an internal voltage reconstruction algorithm, ensuring that during PSASP simulation, the module can correct the internal model drive voltage in real-time based on the PCC point voltage.

[0107] In the third-party simulation application layer, users import the generated model package into the PSASP user-defined model editor. Within the PSASP simulation environment, the input interface of this component is associated with the bus voltage per-unit vector of the target substation. The output interface is associated with the active and reactive power injection of this node. During transient stability analysis calculations, PSASP calls this encapsulated component at each time step. Internally, the component adjusts its parameters based on real-time data. It automatically runs the built-in physical equations and identifies parameters, returning accurate results. and .

[0108] Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 7 As shown, the electronic device 7 of this embodiment includes a processor 70 and a memory 71. The memory 71 stores a computer program 72. When the processor 70 executes the computer program 72, it implements the steps in the various method embodiments described above. Alternatively, when the processor 70 executes the computer program 72, it implements the functions of each module / unit in the various device embodiments described above.

[0109] For example, computer program 72 may be divided into one or more modules / units, which are stored in memory 71 and executed by processor 70 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 72 in electronic device 7.

[0110] Electronic device 7 may include, but is not limited to, processor 70 and memory 71. Those skilled in the art will understand that... Figure 7 This is merely an example of electronic device 7 and does not constitute a limitation on electronic device 7. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 7 may also include input / output devices, network access devices, buses, etc.

[0111] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0112] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0113] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for regional integrated load modeling of a distribution network, characterized in that, include: Different disturbance events are applied to the constructed power distribution network model to be simulated, which includes multiple operating scenarios; Extract the feature parameters of each node; Clustering and partitioning are performed based on the aforementioned feature parameters to obtain multiple sub-distribution network areas; A comprehensive load model is constructed for each sub-distribution network area. The comprehensive load model includes a static load ZIP model, an induction motor model, a distributed photovoltaic model, and an electric vehicle charging station model. Based on the preset parameter stability index, the values ​​of the parameters to be identified for the integrated load model corresponding to each sub-distribution network area are determined respectively. Based on the numerical values ​​of the integrated load model with identification parameters corresponding to each sub-distribution network area, a dynamic simulation component is output.

2. The method for regional integrated load modeling of distribution networks according to claim 1, characterized in that, The characteristic parameters include dynamic response characteristics, resource distribution characteristics, and continuous dynamic response sequences; The clustering and partitioning based on the feature parameters yields multiple sub-distribution network areas, including: Based on the dynamic response characteristics and the resource distribution characteristics, the scalar characteristic distance of each node is determined; Based on the continuous dynamic response sequence, the dynamic trajectory distance of each node is determined; Based on the scalar feature distance of each node and the dynamic trajectory distance, a comprehensive distance matrix is ​​constructed; Based on agglomerative hierarchical clustering, the comprehensive distance matrix is ​​partitioned to obtain the multiple sub-distribution network areas.

3. The method for regional integrated load modeling of distribution networks according to claim 2, characterized in that, The continuous dynamic response sequence includes a voltage response time sequence, an active power response time sequence, and a reactive power response time sequence. The determination of the scalar feature distance of each node based on the dynamic response characteristics and the resource distribution characteristics includes: The dynamic response characteristics and the resource distribution characteristics are normalized to obtain a comprehensive feature vector; Based on the Euclidean distance and the comprehensive feature vector, the scalar feature distance of each node is determined; Determining the dynamic trajectory distance of each node based on the continuous dynamic response sequence includes: Based on the dynamic time adjustment algorithm, the nonlinear aligned shortest path distances of the voltage response time series, active power response time series and reactive power response time series are calculated respectively. The dynamic trajectory distance is determined based on the mean of the nonlinear aligned shortest path distances of the voltage response time series, active power response time series, and reactive power response time series.

4. The method for regional integrated load modeling of distribution networks according to claim 2, characterized in that, The method of partitioning the comprehensive distance matrix based on agglomerative hierarchical clustering to obtain the multiple sub-distribution network regions includes: Calculate the values ​​of each evaluation index for each number of candidate regions; The optimal number of zones for the distribution network is determined by summing the values ​​of all evaluation indicators corresponding to each candidate zone number. Based on the optimal number of partitions, the multiple sub-distribution network areas are obtained.

5. The method for regional integrated load modeling of distribution networks according to claim 1, characterized in that, The various operating scenarios include benchmark scenarios and non-benchmark scenarios; The step of determining the values ​​of the parameters to be identified for the integrated load model corresponding to each sub-distribution network area based on preset parameter stability indices includes: Determine the parameters to be identified in the integrated load model; wherein, the parameters to be identified include multiple parameters; Under the baseline scenario, the parameters to be identified in the integrated load model corresponding to each sub-distribution network area are identified to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area; wherein, the parameters to be identified include structural parameters and scenario parameters. In the non-benchmark scenario, the values ​​of the scenario parameters are updated until the parameter stability index meets the set conditions; wherein, the values ​​of the structural parameters are fixed in the non-benchmark scenario.

6. The method for regional integrated load modeling of distribution networks according to claim 5, characterized in that, In the baseline scenario, the parameters to be identified in the integrated load model corresponding to each sub-distribution network area are identified to obtain the values ​​of the structural parameters and scenario parameters of the integrated load model corresponding to each sub-distribution network area, including: In the baseline scenario, the sum of the root mean square errors of the tested active power waveform and reactive power waveform with the output waveform of each integrated load model is used as the fitness function. Based on the improved particle swarm optimization algorithm, the numerical values ​​of the structural parameters and scenario parameters of each integrated load model are obtained.

7. The method for regional integrated load modeling of distribution networks according to claim 5, characterized in that, In the non-baseline scenario, updating the values ​​of the scenario parameters until the parameter stability index meets the set conditions includes: Based on changing the value of the trust region in the trust region algorithm, the parameter stability index of all scene parameters under the target non-benchmark scenario is calculated. If the calculated stability index does not meet the preset condition, then all scene parameters under the target non-baseline scenario are updated until the calculated parameter stability index meets the preset condition. Then, the value of the scene parameter that meets the preset condition is determined as the value of all scene parameters under the target non-baseline scenario; wherein, the target non-baseline scenario is any one of the non-baseline scenarios.

8. The method for regional integrated load modeling of distribution networks according to claim 5, characterized in that, The parameter stability index includes the coefficient of variation of the scene parameters and / or the confidence interval width of the coefficient of variation of the scene parameters within a set value.

9. The method for regional integrated load modeling of distribution networks according to any one of claims 1-8, characterized in that, The various operating scenarios include a scenario where photovoltaic output accounts for a greater than or equal to a first set ratio of the total load, a scenario where electric vehicle charging load accounts for a greater than or equal to a second set ratio of the total load, a scenario where photovoltaic output accounts for a greater than or equal to the first set ratio of the total load and electric vehicle charging load accounts for a greater than or equal to the second set ratio of the total load, and a baseline scenario.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images