Multi-circuit DC limit power transmission capability evaluation method and system based on extreme gradient lifting, and computer readable storage medium
By building a time domain simulation model and using particle swarm algorithm and extreme gradient enhancement XGboost algorithm, the calculation complexity and adaptability problems of multi-back DC extreme transmission capacity evaluation are solved, and the rapid and accurate evaluation effect is achieved, suitable for new energy asynchronous Internet scenario.
Patent Information
- Application Number
- CN202411121009.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-15
AI Technical Summary
The prior art has complex calculations, long time-consuming and difficult to adapt to generator output fluctuations and load changes when evaluating multi-return DC power transmission capacity, especially in the asynchronous interconnection scenarios of new energy.
Using the multi-return DC extreme transmission capacity evaluation method based on extreme gradient enhancement, the sample set is generated and the extreme transmission capacity evaluation model is constructed by constructing a time domain simulation model, particle swarm algorithm and XGboost algorithm, and the extreme gradient enhancement XGboost algorithm is used for learning, forming Multiple return DC ultimate transmission capacity evaluation model.
It improves the flexibility and speed of evaluation, can quickly obtain results when the generator output or load changes, meets the accuracy of actual grid operation planning, and is suitable for diverse scenarios.
Smart Images

Figure CN119029853A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power systems, and in particular to a method, system and computer-readable storage medium for evaluating the maximum transmission capacity of multiple direct current circuits based on extreme gradient boosting. Background Art
[0002] In the application of clean energy such as photovoltaics and hydropower, they are faced with the problem that they are located in high altitude areas, far away from the main load centers, and difficult to transmit and consume. Improving the ability of the power grid to consume clean energy such as photovoltaics and hydropower is the key to achieving reasonable resource allocation and coordinated operation between regional power grids and ensuring the overall safety and stability of the power grid. In the context of cross-regional and long-distance power transmission, the transmission of new energy via multiple DC lines has greater economic advantages than transmission via multiple AC lines, and currently shows broad application prospects in actual power grids.
[0003] At present, most multi-circuit DC transmission projects use hybrid DC transmission formed by mixing conventional DC and flexible DC. Conventional DC converters consume a lot of reactive power during stable operation. If the current I flowing from the AC system to the commutation bus is increased, the active power P transmitted by DC and the reactive power Q consumed will also increase. On the one hand, excessive reactive power consumption causes the voltage U of the commutation bus to drop and cannot be maintained within the normal voltage range. On the other hand, when the increase in I is not enough to offset the voltage drop caused by the increase in reactive power consumption, the DC transmission power will show a downward trend, so the DC converter station has a maximum power operation point. Although conventional DC converters are generally equipped with reactive power compensation devices on the AC side, their limit transmission capacity is still restricted by static voltage stability and maximum power curves. Flexible DC converter stations can achieve active and reactive power decoupling control, but their transmission capacity is still restricted by the capacity of the converter station. If the limit transmission capacity of multi-circuit DC cannot be accurately and effectively evaluated, it will cause difficulties in dispatching the transmission and consumption of new energy power generation, and the static safety of the power grid cannot be guaranteed. Therefore, it is very important to objectively evaluate the limit transmission capacity of multi-circuit DC. Existing studies mainly use model-driven methods such as the optimal power flow method and the continuous power flow method to calculate and evaluate the maximum transmission capacity of multiple DC lines. These methods often have the disadvantages of complex calculations and long time consumption. In addition, when applied to actual power grids, the model dimension is high and cannot adapt to diverse scenarios such as generator output fluctuations and load level changes. In addition, some studies currently use data-driven methods supported by machine learning to evaluate the maximum transmission capacity of regional power grids interconnected by AC. Studies have shown that these methods can achieve relatively reliable evaluation of the maximum transmission capacity and have good application potential in actual power grids. However, for the case where the sending-end power grid containing large-scale new energy is asynchronously interconnected with the receiving-end power grid through multiple conventional / flexible DC lines, how to use data-driven methods to efficiently and accurately evaluate the maximum transmission capacity of multiple DC lines is still a technical gap. Summary of the invention
[0004] In view of this, the present invention provides a method, system and computer-readable storage medium for evaluating the maximum transmission capacity of multiple DC circuits based on extreme gradient boosting, so as to at least solve the problems existing in the existing technology for evaluating the maximum transmission capacity of multiple DC circuits, namely, that it is time-consuming, only applicable to specific scenarios and difficult to be applied to actual power grids.
[0005] In order to achieve the above object, the present invention adopts the following technical solution:
[0006] A method for evaluating the maximum transmission capacity of multi-circuit direct current based on extreme gradient improvement comprises the following steps:
[0007] A multi-circuit DC limit transmission capacity evaluation model is constructed, and the multi-circuit DC limit transmission capacity evaluation model is used to evaluate the multi-circuit DC limit transmission capacity; wherein, the specific contents of constructing the multi-circuit DC limit transmission capacity evaluation model include:
[0008] A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity;
[0009] Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints;
[0010] The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ;
[0011] Determine the hyperparameters of the extreme gradient boosting XGboost algorithm;
[0012] The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
[0013] Preferably, the specific contents of batch adjustment of regional generator processing and load level parameters include:
[0014] According to the principle that the power factor remains unchanged before and after the change, the output and load levels of all generators in the area are multiplied by a random correction coefficient between 0.95 and 1.05; the principles for adjusting the output and load levels of generators are:
[0015]
[0016] Where s is the correction factor, 0.95≤s≤1.05, φ u is the power factor angle, P Gu0 and Q Gu0 are respectively the active power and reactive power of the generator output before adjustment in the uth region, P Gu and Q Gu are respectively the active power and reactive power of the generator output after adjustment in the uth region; P Lu and Q Lu are the active consumption and reactive consumption of the load in the u-th area respectively, u=1,2…a, a is the number of areas included in the actual power grid.
[0017] Optimally, the power flow constraints of pure AC nodes are:
[0018]
[0019] The power flow constraints of the AC nodes connected to the LCC converter station are:
[0020]
[0021] The power flow constraints of the AC node connected to the MMC converter station are:
[0022]
[0023] Among them, P i and Q i are the active power and reactive power of AC node i respectively; φ k is the power factor angle of the kth LCC converter station; U dk and I dk are the DC voltage and current of the kth LCC converter station or MCC converter station respectively; θ k is the power factor angle of the kth MMC converter station; δ k Y is the fundamental voltage phase output by the kth MMC converter station; k is the equivalent admittance of the kth MMC converter station; M k is the modulation ratio of the kth MMC converter station; n is the number of pure AC nodes in the power grid; m is the number of LCC converter stations; N is the number of MMC converter stations;
[0024] Generator output constraints:
[0025]
[0026] Converter station capacity constraints:
[0027]
[0028] Converter transformer ratio constraints:
[0029] K vmin ≤K v ≤K vmax (7)
[0030] Node voltage constraints:
[0031] U imin ≤U i ≤U imax (8).
[0032] Among them, P Gi is the active power output of the generator; Q Gi is the reactive power of the generator output; K v is the transformation ratio of the converter transformer; the subscripts min and max represent the maximum and minimum values of the corresponding parameters respectively.
[0033] Preferably, the optimal multi-circuit DC limit transmission capacity of each simulation file under the static safety and stability constraints is calculated according to the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p The specific contents include:
[0034] The active power transmitted by multiple DC circuits is used to reflect the maximum transmission capacity of multiple DC circuits, which are represented by x1, x2, …, x m , where m is the number of DC transmission times;
[0035] The fitness function f(x) of the particle swarm algorithm is expressed as: f(x) = -(x1+x2+...+x m );
[0036] After iterative optimization using the particle swarm algorithm, the minimum value of f(x) under the static safety and stability constraints is calculated and recorded as y;
[0037] The active output of the generator P after adjustment is G1 ,P G2 ,…,P Ga , Active power consumption of load P L1 ,P L2 ,…,P La , Generator reactive output Q G1 ,Q G2 ,…,Q Ga , reactive power consumption of load QL1 ,Q L2 ,…,Q La Combined to form a sample set S p :
[0038] S p ={[P G1 ,Q G1 ,P L1 ,Q L1 ...,P Ga ,Q Ga ,P La ,Q La ,y] (p)},p=1,2,...,q
[0039] Here, q is the number of samples.
[0040] Preferably, the loss function used by the XGboost algorithm to construct the t-th decision tree model is the mean square error loss function, and the goal of training each decision tree model is to minimize the loss function:
[0041]
[0042] Among them, y p is the true value of the pth sample, is the model prediction value obtained by inputting the pth sample into the tth decision tree, q is the number of samples, γ is the regular term coefficient in the loss function, and the number of regression tree leaf nodes T greater than the preset threshold is penalized;
[0043] The converged XGboost model is tuned by the grid search method: after the search range of each hyperparameter is given, the grid search traverses the given hyperparameter value combination and performs a ten-fold cross validation: all the data are randomly divided into ten non-overlapping subsets, nine of which are selected as training sets in each iteration, and multiple regression prediction models are constructed by multiple sets of hyperparameter combinations. The remaining subset is used as a test set to evaluate the accuracy of these models, thereby selecting the optimal hyperparameters and obtaining the final multi-circuit DC limit transmission capacity assessment model.
[0044] Preferably, the accuracy of the multi-circuit DC limit transmission capacity assessment model is evaluated by the determination coefficient and the root mean square error:
[0045] The coefficient of determination is:
[0046]
[0047] The root mean square error is:
[0048]
[0049] in, is the predicted value of the p-th sample by the XGboost-based limit transmission capacity regression prediction model; y p is the true value of the sample's ultimate transmission capacity; where p = 1, 2, 3, ..., q, q is the number of samples; R 2 ∈[0,1], the closer it is to 1, the better the prediction effect of the multi-circuit DC limit transmission capacity assessment model; the smaller the RMSE, the better the prediction effect.
[0050] A multi-circuit DC limit transmission capacity assessment system based on extreme gradient boosting, comprising: a model building module;
[0051] The model building module is used to build a multi-circuit DC limit transmission capacity evaluation model, and evaluate the multi-circuit DC limit transmission capacity through the multi-circuit DC limit transmission capacity evaluation model; wherein, the specific contents of building the multi-circuit DC limit transmission capacity evaluation model include:
[0052] A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity;
[0053] Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints;
[0054] The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ;
[0055] Determine the hyperparameters of the extreme gradient boosting XGboost algorithm;
[0056] The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
[0057] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.
[0058] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method, system and computer-readable storage medium for evaluating the maximum transmission capacity of multiple DC circuits based on extreme gradient improvement, which have the following beneficial effects:
[0059] Compared with existing methods such as the continuous power flow method and the optimal power flow method, the present invention constructs a limit transmission capacity regression prediction model based on the XGboost algorithm through a large number of samples. When the generator output or load level changes, there is no need to solve the limit transmission capacity through a time-consuming optimization algorithm. Instead, the generator output and load level related data are directly input into the regression prediction model to obtain the result, which improves the flexibility and rapidity of multi-circuit DC limit transmission capacity evaluation. At the same time, the present invention adopts ten-fold cross validation as a parameter tuning method to improve the generalization performance of the model. In addition, the simulation file generation method of the present invention is based on actual engineering applications, and its accuracy also meets the needs of actual power grid operation planning, and has practical value. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 creative work.
[0061] Figure 1 A flowchart of constructing a multi-circuit DC limit transmission capacity assessment model provided by the present invention;
[0062] Figure 2 A structural diagram of a regression prediction model based on XGboost provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0064] The present invention provides a method for evaluating the maximum transmission capacity of multiple direct current circuits based on extreme gradient improvement, comprising the following steps:
[0065] A multi-circuit DC limit transmission capacity evaluation model is constructed to evaluate the multi-circuit DC limit transmission capacity; among them, Figure 1As shown in the figure, the specific contents of constructing a multi-circuit DC limit transmission capacity assessment model include:
[0066] A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity;
[0067] Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints;
[0068] The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ;
[0069] Determine the hyperparameters of the extreme gradient boosting XGboost algorithm;
[0070] The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
[0071] It should be noted that:
[0072] In this embodiment, PSD-BPA is used to perform time domain simulation, and the obtained simulation result file contains the power system flow calculation results: the voltage amplitude and phase of each node, and the active and reactive power transmitted on each line.
[0073] In order to further implement the above technical solution, the specific contents of batch adjustment of regional generator processing and load level parameters include:
[0074] In order to simulate the changes in generator output and load level fluctuations and make the constructed model applicable to various scenarios, all generator outputs and load levels in the area are multiplied by a random correction coefficient between 0.95 and 1.05 according to the principle that the power factor remains unchanged before and after the change; the principles for adjusting the generator output and load level are:
[0075]
[0076] Where s is the correction factor, 0.95≤s≤1.05, φ uis the power factor angle, P Gu0 and Q Gu0 are respectively the active power and reactive power of the generator output before adjustment in the uth region, P Gu and Q Gu are respectively the active power and reactive power of the generator output after adjustment in the uth region; P Lu and Q Lu are the active consumption and reactive consumption of the load in the u-th area respectively, u=1,2…a, a is the number of areas included in the actual power grid.
[0077] In order to further implement the above technical solution, the power flow constraints of pure AC nodes are:
[0078]
[0079] The power flow constraints of the AC nodes connected to the LCC converter station are:
[0080]
[0081] The power flow constraints of the AC node connected to the MMC converter station are:
[0082]
[0083] Among them, P i and Q i are the active power and reactive power of AC node i respectively; φ k is the power factor angle of the kth LCC converter station; U dk and I dk are the DC voltage and current of the kth LCC converter station or MCC converter station respectively; θ k is the power factor angle of the kth MMC converter station; δ k Y is the fundamental voltage phase output by the kth MMC converter station; k is the equivalent admittance of the kth MMC converter station; M k is the modulation ratio of the kth MMC converter station; n is the number of pure AC nodes in the power grid; m is the number of LCC converter stations; N is the number of MMC converter stations;
[0084] Generator output constraints:
[0085]
[0086] Converter station capacity constraints:
[0087]
[0088] Converter transformer ratio constraints:
[0089] K vmin≤K v ≤K vmax (7)
[0090] Node voltage constraints:
[0091] U imin ≤U i ≤U imax (8).
[0092] Among them, P Gi is the active power output of the generator; Q Gi is the reactive power of the generator output; K v is the transformation ratio of the converter transformer; the subscripts min and max represent the maximum and minimum values of the corresponding parameters respectively.
[0093] In order to further implement the above technical solution, the optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated according to the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p The specific contents include:
[0094] The active power of multiple DC transmission is used to reflect the maximum transmission capacity of multiple DC transmission. That is, the variable to be optimized in the particle swarm algorithm is the active power of multiple DC transmission, which is expressed as x1, x2, …, x m , where m is the number of DC transmission times;
[0095] The fitness function f(x) of the particle swarm algorithm is expressed as: f(x) = -(x1+x2+...+x m );
[0096] By continuously calling the power flow calculation program and checking whether the constraints in the results are met, the particle swarm algorithm iteratively searches for the best solution within the search range defined by the constraints, and finally calculates the minimum value of f(x) under the static safety and stability constraints, which is recorded as y.
[0097] The active output of the generator P after adjustment is G1 ,P G2 ,…,P Ga , Active power consumption of load P L1 ,P L2 ,…,P La , Generator reactive output Q G1 ,Q G2 ,…,Q Ga , reactive power consumption of load Q L1 ,Q L2 ,…,Q La Combined to form a sample set S p :
[0098] Sp ={[P G1 ,Q G1 ,P L1 ,Q L1 ...,P Ga ,Q Ga ,P La ,Q La ,y] (p)},p=1,2,...,q
[0099] Here, q is the number of samples.
[0100] It should be noted that:
[0101] Since the generator output and load level are fixed in each simulation file, the particle swarm algorithm calculates the multi-circuit DC limit transmission capacity under specific scenarios.
[0102] In order to further implement the above technical solution, the loss function used by the XGboost algorithm in S5 to construct the t-th decision tree model is the mean square error loss function. The goal of training each decision tree model is to minimize the loss function:
[0103]
[0104] Among them, y p is the true value of the pth sample, is the model prediction value obtained by inputting the pth sample into the tth decision tree, q is the number of samples, γ is the regular term coefficient in the loss function, and the number of regression tree leaf nodes T greater than the preset threshold is penalized;
[0105] The converged XGboost model is tuned by the grid search method: after the search range of each hyperparameter is given, the grid search traverses the given hyperparameter value combination and performs a ten-fold cross validation: all the data are randomly divided into ten non-overlapping subsets, nine of which are selected as training sets in each iteration, and multiple regression prediction models are constructed by multiple sets of hyperparameter combinations. The remaining subset is used as a test set to evaluate the accuracy of these models, thereby selecting the optimal hyperparameters and obtaining the final multi-circuit DC limit transmission capacity assessment model.
[0106] It should be noted that:
[0107] XGboost is an ensemble learning algorithm based on decision trees. The core idea of the XGboost algorithm is to first train a decision tree model, then add more decision trees to fit the residuals of the first tree, and finally combine the results of each decision tree into a k (x) and weight coefficient α kAfter multiplication, sum is taken to obtain the final regression prediction result. The structure diagram of the regression prediction model based on XGboost constructed in this embodiment is as follows: Figure 2 shown.
[0108] In this embodiment, the meaning, search range and search results of each hyperparameter of the XGboost algorithm are as follows:
[0109]
[0110] In this embodiment, the XGboost algorithm integrates 50 decision trees, and the learning rate is 0.4 when the loss function gradient decreases; 60% of the sample size is used to build a decision tree each time, and sampling with replacement is used to ensure the differences between each decision tree; each tree is split into 6 layers at most; when the loss function value of the node decreases by less than 0.1, the decision tree stops splitting. The final output prediction result is the weighted sum of the prediction results of all trees.
[0111] In order to further implement the above technical solution, the accuracy of the multi-circuit DC limit transmission capacity assessment model is evaluated by the determination coefficient and root mean square error:
[0112] The coefficient of determination is:
[0113]
[0114] The root mean square error is:
[0115]
[0116] in, is the predicted value of the p-th sample by the XGboost-based limit transmission capacity regression prediction model; y p is the true value of the sample's ultimate transmission capacity; where p = 1, 2, 3, ..., q, q is the number of samples; R 2 ∈[0,1], the closer it is to 1, the better the prediction effect of the multi-circuit DC limit transmission capacity assessment model; the smaller the RMSE, the better the prediction effect.
[0117] A multi-circuit DC limit transmission capacity assessment system based on extreme gradient boosting, comprising: a model building module;
[0118] The model building module is used to build a multi-circuit DC limit transmission capacity evaluation model, and evaluate the multi-circuit DC limit transmission capacity through the multi-circuit DC limit transmission capacity evaluation model; wherein, the specific contents of building the multi-circuit DC limit transmission capacity evaluation model include:
[0119] A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity;
[0120] Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints;
[0121] The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ;
[0122] Determine the hyperparameters of the extreme gradient boosting XGboost algorithm;
[0123] The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
[0124] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.
[0125] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A method for evaluating the maximum transmission capacity of multi-circuit DC based on extreme gradient boosting, characterized in that: The following steps are involved: A multi-circuit DC limit transmission capacity evaluation model is constructed, and the multi-circuit DC limit transmission capacity evaluation model is used to evaluate the multi-circuit DC limit transmission capacity; wherein, the specific contents of constructing the multi-circuit DC limit transmission capacity evaluation model include: A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity; Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints; The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ; Determine the hyperparameters of the extreme gradient boosting XGboost algorithm; The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
2. According to claim 1, a method for evaluating the maximum transmission capacity of multiple DC circuits based on extreme gradient boosting is characterized in that: The specific contents of batch adjustment of regional generator processing and load level parameters include: According to the principle that the power factor remains unchanged before and after the change, the output and load levels of all generators in the area are multiplied by a random correction coefficient between 0.95 and 1.05; the principles for adjusting the output and load levels of generators are: Where s is the correction factor, 0.95≤s≤1.05, φ u is the power factor angle, P Gu0 and Q Gu0 are respectively the active power and reactive power of the generator output before adjustment in the uth region, P Gu and Q Gu are respectively the active power and reactive power of the generator output after adjustment in the uth region; P Lu and Q Lu are the active consumption and reactive consumption of the load in the u-th area respectively, u=1,2…a, a is the number of areas included in the actual power grid.
3. The method for evaluating the maximum transmission capacity of multiple DC circuits based on extreme gradient boosting according to claim 1 is characterized in that: Power flow constraints for pure AC nodes: Among them, U i and U j are the node voltages of node i and node j respectively; G ij is the mutual conductance between node i and node j; B ij is the mutual susceptance between node i and node j; The power flow constraints of the AC nodes connected to the LCC converter station are: The power flow constraints of the AC node connected to the MMC converter station are: Among them, P i and Q i are the active power and reactive power of AC node i respectively; φ k is the power factor angle of the kth LCC converter station; U dk and I dk are the DC voltage and current of the kth LCC converter station or MCC converter station respectively; θ k is the power factor angle of the kth MMC converter station; δ k Y is the fundamental voltage phase output by the kth MMC converter station; k is the equivalent admittance of the kth MMC converter station; M k is the modulation ratio of the kth MMC converter station; n is the number of pure AC nodes in the power grid; m is the number of LCC converter stations; N is the number of MMC converter stations; Generator output constraints: Converter station capacity constraints: Converter transformer ratio constraints: K vmin ≤K v ≤K vmax (7) Node voltage constraints: IN imin ≤U i ≤U imax (8). Among them, P Gi is the active power output of the generator; Q Gi is the reactive power of the generator output; K v is the transformation ratio of the converter transformer; the subscripts min and max represent the maximum and minimum values of the corresponding parameters respectively.
4. The method for evaluating the multi-circuit DC power transmission capacity based on extreme gradient boosting according to claim 1 is characterized in that: The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p The specific contents include: The active power transmitted by multiple DC circuits is used to reflect the maximum transmission capacity of multiple DC circuits, which are represented by x1, x2, …, x m , where m is the number of DC transmission times; The fitness function f(x) of the particle swarm algorithm is expressed as: f(x) = -(x1+x2+...+x m ); After iterative optimization using the particle swarm algorithm, the minimum value of f(x) under the static safety and stability constraints is calculated and recorded as y; The active output of the generator P after adjustment is G1 ,P G2 ,…,P Ga , Active power consumption of load P L1 ,P L2 ,…,P La , Generator reactive output Q G1 ,Q G2 ,…,Q Ga , reactive power consumption of load Q L1 ,Q L2 ,…,Q La Combined to form a sample set S p : S p ={[P G1 ,Q G1 ,P L1 ,Q L1 ...,P Ga ,Q Ga ,P La ,Q La ,y] (p) },p=1,2,...,q Here, q is the number of samples.
5. The method for evaluating the maximum transmission capacity of multiple DC circuits based on extreme gradient boosting according to claim 1 is characterized in that: The loss function used by the XGboost algorithm to construct the t-th decision tree model is the mean square error loss function. The goal of training each decision tree model is to minimize the loss function: Among them, y p is the true value of the pth sample, is the model prediction value obtained by inputting the pth sample into the tth decision tree, q is the number of samples, γ is the regular term coefficient in the loss function, and the number of regression tree leaf nodes T greater than the preset threshold is penalized; The converged XGboost model is tuned by the grid search method: after the search range of each hyperparameter is given, the grid search traverses the given hyperparameter value combination and performs a ten-fold cross validation: all the data are randomly divided into ten non-overlapping subsets, nine of which are selected as training sets in each iteration, and multiple regression prediction models are constructed by multiple sets of hyperparameter combinations. The remaining subset is used as a test set to evaluate the accuracy of these models, thereby selecting the optimal hyperparameters and obtaining the final multi-circuit DC limit transmission capacity assessment model.
6. The method for evaluating the multi-circuit DC power transmission capacity based on extreme gradient boosting according to claim 1 is characterized in that: The accuracy of the multi-circuit DC limit transmission capacity assessment model is evaluated by the determination coefficient and root mean square error: The coefficient of determination is: The root mean square error is: in, is the predicted value of the p-th sample by the XGboost-based limit transmission capacity regression prediction model; y p is the true value of the sample's ultimate transmission capacity; where p = 1, 2, 3, ..., q, q is the number of samples; R 2 ∈[0,1], the closer it is to 1, the better the prediction effect of the multi-circuit DC limit transmission capacity assessment model; the smaller the RMSE, the better the prediction effect.
7. A multi-circuit DC limit transmission capacity assessment system based on extreme gradient boosting, based on a multi-circuit DC limit transmission capacity assessment method based on extreme gradient boosting as claimed in any one of claims 1 to 6, characterized in that: include: Model building module; The model building module is used to build a multi-circuit DC limit transmission capacity evaluation model, and evaluate the multi-circuit DC limit transmission capacity through the multi-circuit DC limit transmission capacity evaluation model; wherein, the specific contents of building the multi-circuit DC limit transmission capacity evaluation model include: A time-domain simulation model is built based on the multi-circuit DC grid structure of the target power grid, and the regional generator output and load level parameters are adjusted in batches to generate a simulation file for evaluating the multi-circuit DC limit transmission capacity; Determine the static safety and stability constraints that affect the multi-circuit DC transmission capacity, including equality constraints and inequality constraints. The equality constraints include: power flow constraints of pure AC nodes, power flow constraints of AC nodes connected to LCC converter stations, and power flow constraints of AC nodes connected to MMC converter stations; the inequality constraints include: generator output constraints, converter station capacity constraints, converter transformer ratio constraints, and node voltage constraints; The optimal multi-circuit DC limit transmission capacity of each simulation file under static safety and stability constraints is calculated based on the particle swarm algorithm, and the sample set S is formed by combining the adjusted generator output and load level parameters. p ; Determine the hyperparameters of the extreme gradient boosting XGboost algorithm; The sample set S p The data are input into the XGboost algorithm in batches for learning until the XGboost model obtained through learning converges, and the converged XGboost model is used as a multi-circuit DC limit transmission capacity evaluation model.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.
Citation Information
Patent Citations
Hybrid supervised learning-based power transmission section power limit rapid evaluation method
CN114936527A
Risk behavior information prediction method and system based on gradient boosting decision tree
CN116502742A
Feature selection method for power transmission line parameter prediction and related device
CN117633496A
