A dynamic reactive power partitioning method based on transient voltage constraints

Abstract

The present invention discloses a dynamic reactive power zoning method based on transient voltage constraints, which relates to the technical field of reactive power zoning in power grids. The method comprises: constructing a power grid voltage stability evaluation system; determining a transient voltage safety boundary threshold value of a regional power grid transient voltage stability evaluation index; screening a primary critical fault set; iteratively optimizing and outputting clustering results of the primary critical fault set; determining key buses corresponding to several primary critical faults based on the clustering results of the primary critical fault set and the regional power grid transient voltage stability evaluation index; selecting a partition center from the key bus set; cyclically calculating the voltage electrical distance between the partition center and the bus to be observed using a distance metric algorithm to obtain an electrical distance matrix; performing initial partitioning of the power grid based on the electrical distance matrix; merging and adjusting the initial partitions to determine the optimal reactive power partition. The present invention achieves effective allocation and scheduling of reactive power resources.

Description

Technical Field

[0001] The present invention relates to the technical field of reactive power partitioning of power grids, and in particular to a dynamic reactive power partitioning method based on transient voltage constraints. Background Art

[0002] The voltage stability of a power system refers to its ability to maintain voltage stability after a fault occurs and is an important indicator for evaluating the operational status of the power grid. The integration of renewable energy makes voltage-reactive power control more challenging, so system operators need to manage reactive power resources to maintain voltage stability. Grid zoning is key to improving voltage stability and reactive power control capabilities. Zoning methods are primarily based on electrical distance, including node impedance, transient voltage information, reactive voltage sensitivity, and modularity indicators. These methods perform cluster analysis by quantifying the electrical connection strength or coupling between grid nodes to achieve more accurate and reasonable grid zoning schemes.

[0003] For example, Chinese patent CN105790279B discloses a method for reactive power and voltage partitioning of power systems based on a spectral clustering algorithm. This method simplifies the power grid model by constructing a weighted topological matrix and a Laplace matrix. Eigenvector clustering is performed using an improved K-means algorithm, and the optimal initial partition is selected using modularity Q as an indicator. The partitions are adjusted through connectivity and reactive power balance verification until all conditions are met. However, this method lacks consideration for the dynamic changes in reactive power reserves, and partitioning that does not meet the conditions may affect the adjustment results.

[0004] For example, Chinese patent CN108899898A discloses a reactive voltage zoning method. This method calculates the LQ sensitivity matrix under critical grid conditions, partitions load nodes using hierarchical clustering, constructs a weighted reactive topology based on reactive power flows, further partitions the topology using community discovery, and overlays layers based on reactive reserve indicators to achieve optimized reactive voltage zoning results. However, this invention does not adequately consider the dynamic adaptability of the grid system to real-time changes, and its methods for transient voltage stability and fault identification and classification remain to be improved.

[0005] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0006] In response to the problems in the related art, the present invention proposes a dynamic reactive power zoning method based on transient voltage constraints to overcome the above technical problems existing in the existing related art.

[0007] To this end, the specific technical solutions adopted in the present invention are as follows:

[0008] A dynamic reactive power partitioning method based on transient voltage constraints, comprising:

[0009] S1. Based on transient voltage stability related parameters, a grid voltage stability evaluation system is constructed, and the grid voltage stability evaluation system includes a grid bus transient voltage stability index, a weight coefficient of the importance of bus transient voltage overshoot, and a regional grid transient voltage stability evaluation index;

[0010] S2. Based on the transient voltage stability index of the power grid busbar, the transient voltage safety boundary threshold of the regional power grid transient voltage stability evaluation index is determined by using the fault dynamic simulation technology combined with the heuristic search algorithm;

[0011] S3. Based on the transient voltage safety boundary threshold, determine the fault degree of the power grid and screen the primary critical fault set; use the clustering algorithm to perform cluster analysis on the primary critical fault set, and output the clustering results of the primary critical fault set through iterative optimization;

[0012] S4. Determine the key bus corresponding to the primary key fault based on the clustering results of the primary key fault set and the transient voltage stability evaluation index of the regional power grid; summarize the key buses to obtain a key bus set, and select a partition center from the key bus set;

[0013] S5. Using a distance metric algorithm, cyclically calculate the voltage electrical distance between the partition center and the bus to be observed to obtain an electrical distance matrix; and divide the initial partitions of the power grid based on the electrical distance matrix.

[0014] S6. Based on the pre-built modularity function and the reactive reserve coefficient of the initial partition, the initial partitions are merged and adjusted to determine the optimal reactive partition.

[0015] Furthermore, based on the transient voltage stability related parameters, a grid voltage stability evaluation system is constructed, including:

[0016] The grid bus transient voltage stability index is calculated using the grid bus transient voltage drop index, the grid bus transient voltage overshoot index, and the weight coefficient of the relative importance of adjusting the voltage drop and voltage overshoot;

[0017] Determine the weight coefficient of the importance of bus transient voltage overshoot based on the relationship between the voltage safety setting value of bus transient voltage overshoot, the voltage safety setting value of bus transient voltage sag, and the average bus voltage in the last simulation time period when the fault is cleared and the bus returns to a steady state.

[0018] The maximum value is selected from all grid bus transient voltage stability indices as the regional grid transient voltage stability evaluation index.

[0019] Furthermore, based on the transient voltage stability index of the power grid busbar, the transient voltage safety boundary threshold of the regional power grid transient voltage stability evaluation index is determined by using the fault dynamic simulation technology combined with the heuristic search algorithm, including:

[0020] Establish a power system fault simulation model, perform fault simulation in the power system fault simulation model, and obtain the transient voltage response curve of each bus;

[0021] Calculate the transient voltage stability index of each busbar according to the transient voltage response curve;

[0022] Selecting a simulated annealing algorithm and initializing its parameters; in each iteration of the simulated annealing algorithm, randomly selecting two buses from the current bus set, calculating the difference between the transient voltage stability evaluation index of the power grid bus of the two selected buses, and determining whether to accept the new bus;

[0023] When the iteration ends, the grid bus transient voltage stability evaluation index of the current bus is output as the transient voltage safety boundary threshold.

[0024] Furthermore, based on the transient voltage safety boundary threshold, the fault severity of the power grid is determined, and the primary critical fault set is screened, including:

[0025] According to the pre-configured power system fault number sequence, the fault simulation is performed on each bus in the current bus set to obtain the transient voltage response curve of each bus and calculate the transient voltage stability index of each bus;

[0026] Determine whether the transient voltage stability index of each bus corresponding to any power system fault is greater than the transient voltage safety boundary threshold. If it is greater, the corresponding power system fault is added to the primary critical fault set.

[0027] Furthermore, clustering analysis is performed on the primary key fault set using a clustering algorithm, and the clustering results of the primary key fault set are output through iterative optimization, including:

[0028] Obtain the fault feature group of each sample in the primary key fault set, and establish a fault feature quantity matrix based on the index value of each feature quantity in the fault feature group. The fault feature group includes the transient voltage stability index of the regional power grid, the fault type, the fault location, the load reactive power characteristic index, and the reactive power fluctuation of the new energy source;

[0029] Initialize the network structure and training parameters of the self-organizing feature map neural network based on the primary key faults and fault feature groups;

[0030] Training a self-organizing feature map neural network based on the feature vectors of primary critical faults and updating the training parameters;

[0031] After iterative training of the self-organizing feature map neural network, the clustering results of the primary key fault set are output.

[0032] Furthermore, based on the clustering results of the primary critical fault set and the regional power grid transient voltage stability evaluation index, the key buses corresponding to the primary critical faults are determined to include:

[0033] Calculate the average eigenvector of the clustering results of each primary critical fault set and the Euclidean distance between each fault sample in the clustering results and the average eigenvector, and take the fault sample with the smallest Euclidean distance as the core critical fault;

[0034] According to the core critical faults in the clustering results of each primary critical fault set and the regional power grid transient voltage stability evaluation index under the corresponding core critical faults, the critical bus of the corresponding clustering result is determined.

[0035] Furthermore, the distance metric algorithm is used to cyclically calculate the voltage electrical distance between the partition center and the bus to be observed, and the obtained electrical distance matrix includes:

[0036] The transient voltage electrical distance between each partition center and the bus to be observed is calculated based on the Euclidean distance algorithm, and the cyclic calculation of the transient voltage electrical distance is implemented to obtain the electrical distance matrix.

[0037] Furthermore, the initial partitioning of the power grid based on the electrical distance matrix includes:

[0038] Based on the transient voltage electrical distance, the columns in the electrical distance matrix are arranged in descending order to generate a sorting matrix;

[0039] The partition center and the busbar to be observed corresponding to each row in the sorting matrix are merged to obtain several initial partitions.

[0040] Furthermore, based on the pre-built modularity function and the reactive reserve coefficient of the initial partition, the initial partitions are merged and adjusted to determine the optimal reactive partitions, including:

[0041] Calculate the modularity function value of the initial partition based on the modularity function; obtain the reactive power reserve coefficient of the initial partition;

[0042] Determine whether the reactive power reserve coefficient of any initial partition meets the constraint conditions. If not, merge the initial partition with the adjacent initial partition and determine whether the merged initial partition meets the constraint conditions. Repeat the determination until the constraint conditions are met.

[0043] The modularity function values ​​of all initial partitions that meet the constraints are calculated, and the number of partitions and the partition adjustment results corresponding to the maximum modularity function value are taken as the optimal reactive partition.

[0044] Furthermore, obtaining the reactive reserve coefficient of the initial partition includes:

[0045] Determine the effective dynamic reactive power reserve value in the initial partition and obtain the total reactive power demand value of all loads in the initial partition;

[0046] The reactive power reserve coefficient of the initial partition is obtained by dividing the effective dynamic reactive power reserve value by the total reactive power demand of all loads.

[0047] The beneficial effects of the present invention are:

[0048] (1) The present invention constructs a series of transient voltage stability indicators and applies a simulated annealing optimization algorithm to effectively define the stability boundary of the power system when facing large disturbances, thereby ensuring the dynamic safety and reliability of the power grid.

[0049] (2) The present invention uses the SOM algorithm to perform cluster analysis on multiple features of severe faults, effectively identifying the critical busbars under severe fault conditions, reducing the analysis complexity, and improving the accuracy of power grid fault diagnosis.

[0050] (3) In terms of the implementation of the dynamic reactive power zoning strategy, the present invention proposes a two-stage voltage-reactive power zoning method that takes into account the dynamic reactive power reserve and the electrical distance of the critical bus transient voltage. In practical applications, it can reduce reactive network losses and improve the overall operating efficiency and economy of the power grid.

[0051] (4) This paper considers a dynamic reactive power zoning strategy based on transient voltage constraints under fault conditions. Through detailed simulation analysis and case verification, a new dynamic reactive power zoning method is proposed. This method addresses the transient voltage stability problem of the power system under large disturbances and achieves effective allocation and scheduling of the system's reactive power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 is a flow chart of a dynamic reactive power partitioning method based on transient voltage constraints according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of an AC / DC system equipped with a phase regulator according to an embodiment of the present invention;

[0055] Figure 3is a grid-connected busbar circuit diagram of a synchronous condenser according to an embodiment of the present invention;

[0056] Figure 4 is V according to an embodiment of the present invention load -Q sc curve chart;

[0057] Figure 5 is a preliminary zoning diagram based on transient voltage electrical distance according to an embodiment of the present invention;

[0058] Figure 6 is a flow chart of a dynamic reactive power partitioning strategy according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0060] According to an embodiment of the present invention, a method for dynamic reactive power zoning based on transient voltage constraints is provided. This method, based on transient voltage information, fault clustering, and dynamic reactive power reserve, aims to achieve dynamic grid zoning and fault identification, thereby improving grid safety and reliability.

[0061] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, the dynamic reactive power partitioning method based on transient voltage constraints according to an embodiment of the present invention includes:

[0062] S1. Based on transient voltage stability related parameters, a grid voltage stability evaluation system is constructed, and the grid voltage stability evaluation system includes a grid bus transient voltage stability index, a weight coefficient of the importance of bus transient voltage overshoot, and a regional grid transient voltage stability evaluation index;

[0063] S2. Based on the transient voltage stability index of the power grid busbar, the transient voltage safety boundary threshold of the regional power grid transient voltage stability evaluation index is determined by using the fault dynamic simulation technology combined with the heuristic search algorithm;

[0064] S3. Based on the transient voltage safety boundary threshold, determine the fault degree of the power grid and screen the primary critical fault set; use the clustering algorithm to perform cluster analysis on the primary critical fault set, and output the clustering results of the primary critical fault set through iterative optimization;

[0065] S4. Determine the key bus corresponding to the primary key fault based on the clustering results of the primary key fault set and the transient voltage stability evaluation index of the regional power grid; summarize the key buses to obtain a key bus set, and select a partition center from the key bus set;

[0066] S5. Using a distance metric algorithm, cyclically calculate the voltage electrical distance between the partition center and the bus to be observed to obtain an electrical distance matrix; and divide the initial partitions of the power grid based on the electrical distance matrix.

[0067] S6. Based on the pre-built modularity function and the reactive reserve coefficient of the initial partition, the initial partitions are merged and adjusted to determine the optimal reactive partition.

[0068] In order to facilitate understanding of the above technical solutions of the present invention, the above technical solutions of the present invention are further explained below from the perspectives of architecture and principle.

[0069] 1. Based on the transient voltage stability criterion, a continuous quantitative index L for the transient voltage stability of the power grid bus is constructed. TVSI , the weight coefficient k of the importance of busbar transient voltage overshoot TVSI , regional power grid transient voltage stability evaluation index S TVSI The details are as follows:

[0070] According to the transient voltage stability criterion, a continuous quantitative index L for the transient voltage stability of the power grid bus is constructed. TVSI , the calculation formula is as follows:

[0071] L TVSI =L TVSI.dip +k TVSI L TVSI.over (8)

[0072] In formula (8), L TVSI.dip Indicates the transient voltage drop index of the power grid bus; L TVSI.over Indicates the transient voltage overshoot index of the power grid bus; k TVSI k is the weight coefficient for adjusting the relative importance of voltage drop and voltage overshoot, with a value range of [0, 1]. TVSI The larger the value, the greater the impact of voltage overshoot, and vice versa.

[0073] Weight coefficient k of the importance of busbar transient voltage overshoot TVSI ,consider and The values ​​of the weight coefficients in these three cases are calculated as follows:

[0074]

[0075] Among them, L TVSI.dip and L TVSI.over The construction method of is as shown in formula (10) and formula (11).

[0076]

[0077] In formula (9), formula (10) and formula (11), V0 and V min They represent the initial value of bus voltage and the minimum value of voltage after fault removal respectively; V max represents the maximum bus voltage after the fault is cleared; V(t) represents the bus voltage at time t; V th.over and T th.over Respectively represent the safety setting values ​​of voltage and time of busbar transient voltage overshoot, and the present invention selects 1.1pu and 0.4s; T span.over Indicates that the busbar transient voltage is higher than V th.over The maximum duration of V th.dip and T th.dip Respectively represent the safety setting values ​​of the voltage and time of the bus transient voltage drop, and the present invention selects 0.8pu and 0.4s; T span.dip Indicates that the busbar transient voltage drops below V th.dip The maximum duration of T tole.dip Indicates the maximum duration of the busbar transient voltage drop below 0.98V0; T tole.over Indicates the maximum duration of transient voltage overshoot exceeding 1.02V0. is the average bus voltage during the last simulation period when the fault is cleared and the system returns to steady state; t c0 It indicates the initial moment when the busbar transient voltage drops below 0.98V0 after the fault is eliminated; t c1 Indicates the moment when the busbar transient voltage drops and recovers to a value higher than 0.98V0 after the fault is eliminated; t c2 It indicates the initial moment when the busbar transient voltage is higher than 1.02V0 during overshoot after the fault is eliminated; t c3 It indicates the moment when the busbar transient voltage gradually recovers to a steady state below 1.02V0 after a short dip and overshoot after the fault is eliminated.

[0078] k TVSI The construction of the weight coefficient mainly considers the "drop area" of the busbar transient voltage, which is applied not only to the busbar voltage transient drop process but also to the transient overshoot process. The larger the "drop area", the greater the amplitude of the busbar voltage fluctuation, the longer the voltage drop or overshoot time, and the worse the transient stability of the system. Consider constructing the regional power grid transient voltage stability evaluation index S TVSI , take the worst evaluation index of transient voltage stability of a single bus in the regional power grid, that is, The maximum value of is calculated as follows:

[0079]

[0080] In formula (12), represents the transient voltage stability evaluation index of the regional power grid under fault j; represents the transient voltage stability index of bus i under fault j, i = 1, 2, ..., N, where N is the total number of evaluated buses in the regional power grid.

[0081] 2. Combine fault dynamic simulation and simulated annealing algorithm to search for index S TVSI The transient voltage safety boundary threshold θ TVSI The process is as follows. Figure 2 As shown in the figure, under a three-phase short-circuit grounding fault, by randomly selecting two buses a and b, the transient voltage stability index L of the bus set to be evaluated is calculated based on equations (9), (10) and (11). TVSI.dip 、L TVSI.over , and k TVSI , and then we can get the transient voltage stability index L of the two buses. TVSI , based on the difference Δf between the two busbar indices, determine whether to accept the new busbar, and continuously select the new busbar until the algorithm iteration is completed, and finally output the index L of the current busbar accepted TVSI As the optimal threshold value for judging system transient stability / instability, θ TVSI .

[0082] The details are as follows:

[0083] By combining fault dynamic simulation and simulated annealing algorithm, the search index S TVSI The transient voltage safety boundary threshold θ TVSI The steps for determining the voltage safety threshold based on the simulated annealing optimization algorithm are as follows:

[0084] Step 21: Build a simulation model and set a severe fault on the busbar set to be evaluated.

[0085] Step 22: Set a fault in the simulation model and perform a short-time-scale fault simulation to obtain the transient voltage response curve of each bus.

[0086] Step 23: Calculate the transient voltage stability index L of each bus based on the transient voltage response curve obtained from the fault simulation. TVSI .

[0087] Step 24: Initialize the parameters of the simulated annealing algorithm, including the initial temperature T0, the termination temperature T f , cooling coefficient α and threshold θ TVSI.

[0088] Step 25: Perform iterative solution of simulated annealing algorithm. In each iteration, two buses a and b are randomly selected from the current bus set, and the two buses L are calculated according to formula (8). TVSI The difference Δf=gf=L TVSI,a -L TVSI,b , judge whether to accept the new busbar, if the difference judges Δf<0, then accept L TVSI The larger busbar b is used as the new current busbar, otherwise the busbar F = e -Δf / T Accept bus b as the new current bus.

[0089] Step 26: Reduce the current temperature T by the cooling coefficient α and proceed to the next iteration.

[0090] Step 27: When the set number of iterations is reached, the iteration is terminated and the transient voltage stability evaluation index L of the current bus is output. TVSI As the voltage safety threshold θ TVSI .

[0091] The process of an AC / DC system with a phase regulator installed, such as Figure 2 As shown, Figure 2 The process includes:

[0092] Initialization temperature T, cooling coefficient α, number of iterations L, threshold θ TVSI . Build system simulation model and fault simulation.

[0093] Randomly select two buses a and b from the bus set to be calculated.

[0094] Calculate the transient voltage drop index of the power grid busbar a Grid bus transient voltage overshoot index Adjust the weight coefficient of the relative importance of voltage sag and voltage overshoot and busbar b

[0095] when Less than 2 and If it is less than 2, then calculate judge Is it less than the voltage safety threshold θ TVSI .when Not less than θ TVSI , then accept bus b as the new current bus, otherwise calculate the difference between the voltage safety threshold of bus a and the quantified grid bus transient voltage stability index The difference between the bth busbar And calculate Δf=gf, and judge whether Δf<0 is true. If it is true, then accept bus b as the new current bus. Otherwise, according to a certain probability F=e -Δf / T Accept bus b as the new current bus.

[0096] when Less than 2 and If it is less than 2, then calculate judge Is it less than θ TVSI .when Not less than θ TVSI , then according to a certain probability F=e -Δf / T Accept busbar b as the new current busbar. Otherwise, calculate Δf=gf, and judge whether Δf<0 is established. If it is established, then accept bus b as the new current bus. Otherwise, according to a certain probability F=e -Δf / T Accept bus b as the new current bus.

[0097] Cool down according to the cooling coefficient and determine whether the number of iterations has been reached. When it has been reached, output the threshold.

[0098] Third, determine severe faults (i.e., primary critical faults) using the transient voltage stability threshold of the power grid busbars, and preliminarily screen the severe fault set. Use self-organizing maps to perform cluster analysis on the screened severe faults, and iteratively output the clustering results. Based on the clustering results and the transient voltage stability evaluation index of the regional power grid, determine the critical busbars for severe faults. Aggregate the critical busbars for each fault set to obtain the critical busbar set for severe faults. The details are as follows:

[0099] Calculate the transient voltage stability evaluation index S of the regional power grid under various types of faults TVSI , according to θ TVSI Finally, determine whether the fault is a serious fault and obtain the serious fault set. The detailed steps are as follows:

[0100] Step 31: Construct a regional power grid simulation model in the electromechanical transient simulation software and determine a busbar set, which includes a total of m buses.

[0101] Step 32: Number the expected faults such as AC bus interphase short circuit, ground short circuit under different ground resistance conditions, generator fault, etc. to obtain Fault1, Fault2, Fault3, ..., Fault n .

[0102] Step 33: Perform short-time-scale fault simulation in the order of fault numbers to obtain the current fault Fault u The transient response curves of each bus in the busbar set are calculated, and the transient voltage stability index value of each bus is obtained. Then the transient voltage stability evaluation index of the regional power grid is obtained

[0103] Step 34: Determination Is it greater than the threshold θ TVSI ,like Then collect statistics and summarize faults u Enter the serious fault set, if Let u=u+1 and return to step 3 to recalculate the transient voltage stability evaluation index of the regional power grid under the next fault. When the fault search calculation reaches Fault n Next End the initial screening of serious faults.

[0104] The steps for clustering severe faults based on the self-organizing feature map (SOM) neural network are as follows:

[0105] 1) Data preparation and preprocessing

[0106] Input sample Fault1, Fault2, Fault3, ..., Fault n And the five characteristic quantities corresponding to each sample are: regional power grid transient voltage stability index Fault type, fault location, load reactive power characteristic indicators, and renewable energy reactive power fluctuation. Fault type is a non-numeric feature, so it is encoded using label coding, for example: three-phase ground fault: 0; two-phase interphase fault: 1; two-phase ground fault: 2; single-phase ground fault: 3; DC line disconnection: 4.

[0107] The fault location is represented by the impedance calculation method. The balance node in the simulation model is selected as the reference point. The impedance modulus of the line from each fault node to the reference node is calculated. The characteristic quantity is expressed as To express it, the calculation formula is:

[0108]

[0109] Where, is the impedance modulus of the line from the fault node to the reference node, is the resistance of the line between the fault node and the reference node; is the reactance of the line between the fault node and the reference node; j is the imaginary unit.

[0110] The load reactive power characteristic index refers to the reactive power of the load and the voltage change at the load endpoint. The present invention uses a power relationship to describe the relationship between the two, which can be expressed by the following formula:

[0111]

[0112] Where Q Pload.SOM Represents the reactive power characteristic index of the load near the node; Q Pload0.SOM Indicates the reference reactive power of the load; V Pload.SOM Indicates the voltage at the load terminal; V Pload0.SOM Represents the base voltage of the load. The index n is fitted and estimated based on measured and empirical data. Different load types generally have different values. Static loads have a smaller reactive power index, while dynamic loads have a larger reactive power index. The load index for induction motors is generally between 2.0 and 3.0.

[0113] The reactive power fluctuation of renewable energy in the power system may affect the reactive power balance and stability of the system. Therefore, it is necessary to construct the reactive power fluctuation index of renewable energy. std.SOM It is expressed as follows:

[0114]

[0115] Where N std.SOM Indicates the number of photovoltaic and wind turbine generators near the node; Q stdi.SOM Indicates the maximum fluctuation of reactive power of each photovoltaic and wind turbine generator; Q std0.SOM Indicates the benchmark reactive power of photovoltaic and wind turbine generators.

[0116] The serious faults after the above screening As the input sample, according to the index value of each feature quantity constructed above, the fault feature quantity matrix FAULT is obtained:

[0117]

[0118] In order to better train the model, the five characteristic index values ​​constructed above are standardized to keep them on the same scale. The present invention adopts maximum and minimum value normalization, and the calculation formula is:

[0119]

[0120] Where, Fault a ' ,b Indicates the value of the bth feature of the ath sample after standardization; Fault a,b represents the bth feature value of the ath sample; FAULT represents the fault feature value matrix; max(·) represents the maximum value in the matrix; min(·) represents the minimum value in the matrix.

[0121] 2) Create SOM training network

[0122] Initialize the SOM network. The SOM network consists of an input layer and a competition layer (output layer). The number of neurons in the input layer is determined by the feature dimension M of the input sample data. SOM The present invention determines the serious faults after the above screening As the input sample, set the feature quantity corresponding to each fault, and the number of feature quantities is the dimension M SOM =5.

[0123] The number of neurons in the competition layer is set manually. According to the empirical method, the present invention selects the number of neurons in the competition layer k SOM The calculation formula is:

[0124]

[0125] Where n SOM is the total number of input sample data. In order to make the output space two-dimensional, select a close The square of k SOM value.

[0126] 3) Initialize weights, design learning rate and neighborhood radius

[0127] Initialize the weight vector of the competitive layer neurons:

[0128]

[0129] Initialize the weight vector of the competition layer neurons to a random value in the range [0,1]. Design the initial learning rate α 0.SOM and the initial neighborhood radius radius 0.SOM .

[0130] 4) Training SOM

[0131] Randomly select a severely faulty feature vector and calculate its distance to each neuron on the competition layer. The neuron with the smallest distance is the winning node. SOM (FAULT,W i,j ) to express distance:

[0132]

[0133] Find the neuron with the smallest distance The calculation formula is:

[0134]

[0135] Where, Dis SOM (FAULT,W i,j ) is the distance between the current node and each neuron on the competition layer; W i,j,kRepresents the weight vector of the kth sample neuron. n SOM is the total number of sample data input.

[0136] 5) Update weights, learning rate, and neighborhood radius

[0137] Update the winning node. For each neuron (i, j) in the competitive layer, use the following formula to update the weight vector W of each neuron i,j :

[0138] W i,j (t+1)=W i,j (t)+α(t)·h i,j (t)·(FAULT k -W i,j (t))(21)

[0139] Where t represents the current iteration number; α(t) represents the learning rate at the current iteration number, which usually decreases gradually over time; h i,j (t) represents the intensity of the neighborhood weight update of neuron (i, j) in the competition layer at the current number of iterations; FAULT k Represents the feature vector of the kth sample of the input, where k∈[1,n SOM ]; W i,j (t) represents the weight vector of neuron (i, j) at the current iteration number.

[0140] During the training iteration of SOM, the learning rate α(t) and the neighborhood radius radius(t) are usually gradually reduced, which can be calculated by exponential decay:

[0141]

[0142] Where λ learning and λ neighborhood Represent the time constants of learning rate and neighborhood radius respectively. According to the empirical method, λ learning and λ neighborhood The value range of α is [100,1000], 0.SOM The value range is [0.01, 0.1], radius 0.SOM The value is

[0143] Neighborhood function h i,j (t) is used to calculate the intensity of the neighborhood weight update of each neuron during training, and is represented by a Gaussian function:

[0144]

[0145] Where, Indicates that Euclidean distance is used for calculation, specifically:

[0146]

[0147] Where, radius 2 ( t ) represents the square of the neighborhood radius, and its calculation method can refer to formula (21); (i, j) is the current neuron, (i * ,j * ) is the neuron with the smallest distance obtained through training.

[0148] 6) Repeat the training of SOM and update the weights, learning rate and neighborhood radius until the preset number of iterations is reached and the clustering results are output.

[0149] Using the clustering results of the SOM training network and combining it with the transient voltage stability evaluation index of the regional power grid, the critical busbar with severe faults is determined. The process is as follows:

[0150] 1) Calculate the average feature vector of each fault set

[0151] Let a fault set be f SOM , which contains N fSOM fault samples, each fault sample contains 5 feature quantities, calculate the average value of each feature quantity, the average value of the i-th feature quantity, i∈[1,5] for:

[0152]

[0153] Where, Denotes the fault set f SOM The average value of the i-th feature; Denotes the fault set f SOM The value of the i-th feature quantity of the j-th fault sample in .

[0154] 2) Determine the critical faults of each fault set (the core critical faults obtained from the primary critical faults, hereinafter referred to as critical faults)

[0155] According to the average feature vector of each fault set, calculate the Euclidean distance between each fault sample in the fault set and the average feature vector, and find the fault sample with the smallest distance as the key fault of the fault set. The formula is:

[0156]

[0157] Where, represents the feature vector of the jth sample in the fault set, Denotes the fault set f SOMThe specific calculation of the Euclidean distance in formula (27) is:

[0158]

[0159] 3) Determine the critical busbar for each fault set and output

[0160] According to the key faults of each fault set and the transient voltage stability index of the regional power grid under the fault Sure The corresponding fault bus is the key bus of the fault set. Summarize the key buses of each fault set to get the critical bus set Bus of severe faults. fault .

[0161] 4. Taking the key busbar set as the partition center, the electrical distance matrix D is obtained by cyclic calculation using the Euclidean distance method. center.ij The large power grid is initially divided into zones using transient voltage electrical distance as an indicator, and the dynamic reactive power reserves of the zones are combined and adjusted through the modularity function to determine the optimal zone. The details are as follows:

[0162] By calculating the difference between the reactive power at the nose point of the VQ curve of each dynamic reactive compensation device and the current reactive power output, the effective reactive power reserve of each dynamic reactive compensation device can be obtained. The voltage V at the load bus can be obtained through multiple power flow simulations and fitting. load And the output reactive power Q of the dynamic reactive compensation equipment dynamic The VQ relationship curve between them. The simplified circuit of dynamic reactive power compensation equipment such as synchronous condenser connected to the grid is as follows Figure 3 As shown, Figure 3 Source represents the infinite power source equivalent to the large power grid, SC represents the synchronous phase regulator connected to the grid, in order to construct V load -Q sc curve, maintaining the power factor cosθ load and synchronous condenser active power P sc Constant. The following formula is derived:

[0163]

[0164] Substituting formula (29) into formula (30), we can get V load -Q sc The relationship formula is:

[0165]

[0166] In the above formula, P load Indicates the active power on the load bus; E indicates the electromotive force of the load bus; X line Indicates the inductive reactance on the load bus; V loadRepresents the voltage at the load bus; θ load is the power factor angle; Q load is the reactive power on the load bus; Q sc Indicates the reactive power of synchronous condenser.

[0167] By solving multiple power flow problems and collecting multiple sets of data, the V at a certain load bus node can be obtained. load -Q sc Curves, such as Figure 4 As shown, where the nose point of the curve is V collapse Represents the voltage collapse point, at which point Q sc is the maximum effective reactive power output Q of the synchronous condenser sc.effective , that is, the maximum effective reactive output of each reactive source.

[0168] Define the effective reactive power reserve of each dynamic reactive power source as V load -Q sc The difference between the reactive output of the curve and the current reactive output. The effective reactive reserve of the entire regional power grid is the sum of the effective reactive reserves of all dynamic reactive sources, as shown in the following formula:

[0169]

[0170] In formula (32), Q grid.res is the total effective dynamic reactive power reserve in the regional power grid; Q sc.effective.i is the ith dynamic reactive power source V load -Q sc Reactive power at the nose of the curve; Q sc.i is the reactive power currently generated by the i-th dynamic reactive source; n dynamic is the total number of dynamic reactive power sources in the regional power grid.

[0171] Assume that the critical busbar set with serious fault is:

[0172]

[0173] Each zone center is C fault.i :

[0174]

[0175] Where m center-bus is the initial number of partitions.

[0176] Calculate the center C of each partition fault.i The electrical distance of transient voltage between the busbars and other buses is calculated using the Euclidean distance method, and the formula is:

[0177]

[0178] Where Dcenter.ij Indicates the central busbar C of the i-th partition fault.i and the j-th observed bus Bus j Electrical distance between simulation Indicates the number of time steps in the simulation process; V ik and V jk They are the critical bus with serious fault i and Bus fault.i and the jth critical bus with serious fault fault.j The voltage value at the kth time point. D center.ij The smaller the value, the smaller the distance is, the more similar the transient voltage characteristics of the two buses are, and the more suitable they are to be divided into one partition.

[0179] Cyclic calculation of partition center bus C fault.i and all other buses that need to be observed j The electrical distance between them can be obtained by center.ij , providing a basis for subsequent precise partitioning.

[0180] According to the calculated electrical distance matrix D center.ij , taking into account the two key factors of transient voltage electrical distance and dynamic reactive power reserve, the partition is gradually optimized through two independent stages. The steps are as follows:

[0181] 1) First, the electrical distance matrix D center.ij Sort by column from large to small to get the sorted matrix D sort.mn , assuming that in addition to the partition center bus C fault.i , In addition, the total number of other buses that need to be observed is n study-bus , the busbar set is:

[0182]

[0183] Sorted matrix D sort.mn Any nth in n∈[1,n study-bus ]The values ​​of the column elements follow:

[0184]

[0185] The matrix D sort.mn The first row corresponds to the central busbar C of the partition fault.i With the corresponding other busbar B study.i Divide into the same partition, ensuring:

[0186] All buses are partitioned and not repartitioned.

[0187] The specific process is as follows Figure 5 As shown, we finally get m center-bus Initial partition Figure 5 In, D center.ij is the electrical distance matrix; B study.i is the i-th busbar under study; D sort.mn is the sorted electrical distance matrix; n study-bus is the total number of observed buses; m center-bus Indicates the number of initial partitions; C fault.i Represents the central busbar of the i-th partition.

[0188] 2) After initially partitioning the large power grid using transient voltage electrical distance as an indicator, the dynamic reactive power reserve of the partitions is considered for merging or adjustment, and a modularity function is introduced to determine the optimal partition. The modularity function is as follows:

[0189]

[0190] Where R ij Indicates the i-th partition Area i and the jth partition Area j The weight between i Indicates the i-th partition Area i The sum of the weights between other partitions and within the region, κ j Indicates the jth partition Area j The sum of the weights between other partitions and within the region; M represents the sum of the weights of all partitions; Φ(i,j) is the Kronecker function, when the partition Area i and Area j When they are adjacent, Φ(i,j)=1, otherwise Φ(i,j)=0.

[0191] The effective dynamic reactive power reserve in the region constructed based on the weight index in the modularity function is used to determine the optimal partition based on the dynamic reactive power reserve. The calculation formula is shown in formula (5):

[0192]

[0193] Where Q grid.res.i Indicates area i The effective dynamic reactive power reserve in Q grid.res.j Indicates area j The effective dynamic reactive power reserve in .

[0194] The objective function of partition merging or adjustment is constructed as:

[0195] max f=F dynamic (6)

[0196] Constructing the partition reactive power reserve coefficient as the constraint condition of the partition:

[0197]

[0198] Where C rs.i Indicates the area i Reactive power reserve coefficient; NL.i represents the area i The number of medium loads, Q L.k is the reactive power demand of load k. C rs.i The larger the value, the smaller the partition area. i The stronger the anti-interference ability is, the more sufficient the dynamic reactive power reserve is. rs The value is 0.2.

[0199] The zoning method considering voltage, electrical distance and dynamic reactive power reserve is as follows: Figure 6 The specific steps are as follows:

[0200] 1) Based on the critical bus set with severe faults, the initial partition center is obtained According to formula (32), the voltage electrical distance between each partition center and other buses to be observed is further calculated to obtain the initial matrix. The initial distance matrix is ​​sorted in ascending order to obtain the sorting matrix D sort.mn , merge the partition center and the bus to be observed to obtain the initial partition based on voltage electrical distance:

[0201]

[0202] 2) According to formula (2), calculate the modularity function value F of the initial partition dynamic .

[0203] 3) For each partition Area i , calculate its dynamic reactive power reserve coefficient C according to formula (32) and formula (7) rs.i , to determine whether it satisfies the constraint C rs.i ≥γ rs , determine whether partition adjustment is necessary.

[0204] 4) If the partition Area i Does not meet the constraints, find the area i Adjacent areas with sufficient reactive power reserves h , Area i Merge into the Area partition h Get a new partition.

[0205] 5) If the partition Area iSatisfy the constraints and calculate the modularity function value F after repartitioning. dynamic , continue to traverse the next partition Area i+1 Repeat the process of judging whether the dynamic reactive power reserve coefficient meets the constraint conditions until i>m center-bus .

[0206] 6) Compare the modularity function value F calculated each time dynamic value, so that F dynamic The number of partitions and the partition adjustment result corresponding to the maximum value are the final voltage-reactive power optimal partition results considering the dynamic reactive power reserve under fault conditions.

[0207] The present invention can quantitatively assess the effectiveness of condensers in improving the transient stability of AC and DC power grids within the timeframe of transient processes. This feature enables timely intervention by the dispatch center to ensure system stability recovery, thereby reducing potential economic losses and failure risks.

[0208] 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 in the scope of protection of the present invention.

Claims

1. A dynamic reactive power partitioning method based on transient voltage constraints, characterized in that: include: S1. Based on transient voltage stability related parameters, a grid voltage stability evaluation system is constructed, and the grid voltage stability evaluation system includes a grid bus transient voltage stability index, a weight coefficient of the importance of bus transient voltage overshoot, and a regional grid transient voltage stability evaluation index; S2. Based on the transient voltage stability index of the power grid busbar, the transient voltage safety boundary threshold of the regional power grid transient voltage stability evaluation index is determined by using the fault dynamic simulation technology combined with the heuristic search algorithm; S3. Based on the transient voltage safety boundary threshold, determine the fault degree of the power grid and screen the primary critical fault set; use the clustering algorithm to perform cluster analysis on the primary critical fault set, and output the clustering results of the primary critical fault set through iterative optimization; S4. Determine the key bus corresponding to the primary key fault based on the clustering results of the primary key fault set and the transient voltage stability evaluation index of the regional power grid; summarize the key buses to obtain a key bus set, and select a partition center from the key bus set; S5. Using a distance metric algorithm, cyclically calculate the voltage electrical distance between the partition center and the bus to be observed to obtain an electrical distance matrix; and divide the initial partitions of the power grid based on the electrical distance matrix. S6. Based on the pre-built modularity function and the reactive reserve coefficient of the initial partition, the initial partitions are merged and adjusted to determine the optimal reactive partition.

2. A method for dynamic reactive power partitioning based on transient voltage constraints according to claim 1, characterized in that: The construction of a grid voltage stability evaluation system based on transient voltage stability related parameters includes: The grid bus transient voltage stability index is calculated using the grid bus transient voltage drop index, the grid bus transient voltage overshoot index, and the weight coefficient of the relative importance of adjusting the voltage drop and voltage overshoot; Determine the weight coefficient of the importance of bus transient voltage overshoot based on the relationship between the voltage safety setting value of bus transient voltage overshoot, the voltage safety setting value of bus transient voltage sag, and the average bus voltage in the last simulation time period when the fault is cleared and the bus returns to a steady state. The maximum value is selected from all grid bus transient voltage stability indices as the regional grid transient voltage stability evaluation index.

3. The method for dynamic reactive power partitioning based on transient voltage constraint according to claim 1, characterized in that: The method of determining the transient voltage safety boundary threshold of the regional power grid transient voltage stability evaluation index based on the power grid bus transient voltage stability index by using the fault dynamic simulation technology combined with the heuristic search algorithm includes: Establish a power system fault simulation model, perform fault simulation in the power system fault simulation model, and obtain the transient voltage response curve of each bus; Calculate the transient voltage stability index of each busbar according to the transient voltage response curve; Selecting a simulated annealing algorithm and initializing its parameters; in each iteration of the simulated annealing algorithm, randomly selecting two buses from the current bus set, calculating the difference between the transient voltage stability evaluation index of the power grid bus of the two selected buses, and determining whether to accept the new bus; When the iteration ends, the grid bus transient voltage stability evaluation index of the current bus is output as the transient voltage safety boundary threshold.

4. The method for dynamic reactive power partitioning based on transient voltage constraints according to claim 1, characterized in that: The determination of the fault severity of the power grid based on the transient voltage safety boundary threshold and the screening of the primary critical fault set include: According to the pre-configured power system fault number sequence, the fault simulation is performed on each bus in the current bus set to obtain the transient voltage response curve of each bus and calculate the transient voltage stability index of each bus; Determine whether the transient voltage stability index of each bus corresponding to any power system fault is greater than the transient voltage safety boundary threshold. If it is greater, the corresponding power system fault is added to the primary critical fault set.

5. The method for dynamic reactive power partitioning based on transient voltage constraints according to claim 4, characterized in that: The clustering analysis of the primary key fault set using a clustering algorithm and outputting the clustering results of the primary key fault set through iterative optimization includes: Obtain the fault feature group of each sample in the primary key fault set, and establish a fault feature quantity matrix based on the index value of each feature quantity in the fault feature group. The fault feature group includes the transient voltage stability index of the regional power grid, the fault type, the fault location, the load reactive power characteristic index, and the reactive power fluctuation of the new energy source; Initialize the network structure and training parameters of the self-organizing feature map neural network based on the primary key faults and fault feature groups; Training a self-organizing feature map neural network based on the feature vectors of primary critical faults and updating the training parameters; After iterative training of the self-organizing feature map neural network, the clustering results of the primary key fault set are output.

6. The method for dynamic reactive power partitioning based on transient voltage constraints according to claim 1, characterized in that: Determining the key bus corresponding to the primary key fault based on the clustering results of the primary key fault set and the regional power grid transient voltage stability evaluation index includes: Calculate the average eigenvector of the clustering results of each primary critical fault set and the Euclidean distance between each fault sample in the clustering results and the average eigenvector, and take the fault sample with the smallest Euclidean distance as the core critical fault; According to the core critical faults in the clustering results of each primary critical fault set and the regional power grid transient voltage stability evaluation index under the corresponding core critical faults, the critical bus of the corresponding clustering result is determined.

7. The method for dynamic reactive power partitioning based on transient voltage constraints according to claim 1, characterized in that: The distance measurement algorithm is used to cyclically calculate the voltage electrical distance between the partition center and the bus to be observed, and the electrical distance matrix obtained includes: The transient voltage electrical distance between each partition center and the bus to be observed is calculated based on the Euclidean distance algorithm, and the cyclic calculation of the transient voltage electrical distance is implemented to obtain the electrical distance matrix.

8. The method for dynamic reactive power partitioning based on transient voltage constraints according to claim 1, characterized in that: The initial partitioning of the power grid based on the electrical distance matrix includes: Based on the transient voltage electrical distance, the columns in the electrical distance matrix are arranged in descending order to generate a sorting matrix; The partition center and the busbar to be observed corresponding to each row in the sorting matrix are merged to obtain several initial partitions.

9. A method for dynamic reactive power partitioning based on transient voltage constraints according to claim 8, characterized in that: The merging and adjusting of the initial partitions based on the pre-built modularity function and the reactive reserve coefficient of the initial partition to determine the optimal reactive partition includes: Calculate the modularity function value of the initial partition based on the modularity function; obtain the reactive power reserve coefficient of the initial partition; Determine whether the reactive power reserve coefficient of any initial partition meets the constraint conditions. If not, merge the initial partition with the adjacent initial partition and determine whether the merged initial partition meets the constraint conditions. Repeat the determination until the constraint conditions are met. The modularity function values ​​of all initial partitions that meet the constraints are calculated, and the number of partitions and the partition adjustment results corresponding to the maximum modularity function value are taken as the optimal reactive partition.

10. A method for dynamic reactive power partitioning based on transient voltage constraints according to claim 9, characterized in that: The obtaining of the reactive power reserve coefficient of the initial partition includes: Determine the effective dynamic reactive power reserve value in the initial partition and obtain the total reactive power demand value of all loads in the initial partition; The reactive power reserve coefficient of the initial partition is obtained by dividing the effective dynamic reactive power reserve value by the total reactive power demand of all loads.

Images