Complex power system power flow determination method and determination system

Abstract

The invention relates to the technical field of battery replacement, and discloses a complex power system power flow determination method and system, and the method comprises the steps: dividing a target power system into different partitions, carrying out the grid division, and generating a partition initial training set of each partition based on a grid division result; removing distributed resources which do not belong to partitions corresponding to the partition initial training set in the partition initial training set from the partition training set to obtain final partition training sets, and combining all the final partition training sets into a whole to obtain an overall training set; inputting the overall training model into a constructed machine learning model to obtain a trained machine learning model; calculating a load flow distribution result when no distributed resources exist in the target power system, and inputting the final training set of all the partitions into the trained machine learning model to determine partition load flow calculation results of all the partitions; and combining the partition load flow calculation results of all the partitions to obtain a load flow result of the target power system.

Description

Technical Field

[0001] The present invention relates to the technical field of battery replacement, and in particular to a complex power system power flow determination method and determination system. Background Art

[0002] Traditional machine learning algorithms require a large amount of data sets to achieve good performance when training models, which leads to huge demands for computing and storage resources during the training process. Especially when dealing with large-scale complex data sets, traditional machine learning training often faces the problems of slow computing speed and high resource consumption, which limits its application in practical scenarios. At the same time, the distributed computing framework has shown great potential in large-scale data processing and computing due to its advantages of horizontal scalability and parallel optimization processing. However, traditional distributed computing architectures often fail to give full play to their advantages in machine learning training and fail to adapt well to the needs of machine learning training. Therefore, it is urgent to design a new method that integrates machine learning and distributed computing to solve the defect of large demand for traditional machine learning training set data. Summary of the invention

[0003] The present invention provides a complex power system power flow determination method and determination system to solve the defect of large amount of training set data required by existing machine learning.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] In a first aspect, the present invention provides a method for determining a power flow in a complex power system, comprising:

[0006] Divide the target power system into different partitions, perform grid division on all distributed resources in each partition, and generate the initial training set of each partition based on the grid division results;

[0007] The distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set are removed from the partition training set to obtain a partition final training set, and all partition final training sets are merged into a whole to obtain a whole training set;

[0008] Inputting the overall training model into the constructed machine learning model to obtain a trained machine learning model;

[0009] Calculate the power flow distribution results when there are no distributed resources in the target power system, and input the final training sets of all the partitions into the trained machine learning model to determine the partition power flow calculation results of all the partitions;

[0010] The partition power flow calculation results of all partitions are combined to obtain the power flow results of the target power system.

[0011] Optionally, dividing the target power system into different partitions includes:

[0012] The target power system is partitioned based on the number of distributed resources in the target power system so that the number of distributed resources in each partition is equal.

[0013] Optionally, the gridding of all distributed resources in each partition includes:

[0014] Determine the output range of active power and reactive power of all distributed resources in each partition, and divide all distributed resources in each partition into grids by dividing the output range into equal intervals;

[0015] Among them, the equal interval division methods include:

[0016] The symbol k represents the serial number of the segment, k ≥ 2. When k = 2, it means that the active and reactive outputs of each distributed resource are divided into two segments, namely, the upper and lower limits of the active output and the upper and lower limits of the reactive output, recorded as and When k is other values, the active output and reactive output of different segments are as follows:

[0017]

[0018] Among them, P l and Q l They represent the active and reactive output values ​​of the distributed resource l in the current training set, and They represent the upper limit of active output, the lower limit of active output, the upper limit of reactive output and the lower limit of reactive output of distributed resources respectively, k represents the segment number, and K represents the total number of segments.

[0019] Optionally, removing the distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set from the partition training set includes:

[0020] The active output value and reactive output value of the distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set are taken according to the elimination formula, and the elimination formula satisfies the following relationship:

[0021]

[0022] Among them, P l and Q l They represent the active and reactive output values ​​of the distributed resource l in the current training set respectively, and μ(n) represents the set of distributed resources in partition n.

[0023] Optionally, the final training sets of all partitions are combined into a whole to obtain a whole training set, including:

[0024] The number of final training sets for each partition is determined by satisfying the following relationship:

[0025] S(n)=Π i∈μ(n) k 2 ;

[0026] Among them, S(n) represents the number of training sets in partition n, k 2 Indicates the number of active and reactive permutations of each distributed resource;

[0027] The number of all final training sets is determined based on the number of final training sets for each partition, and the determination method satisfies the following relationship:

[0028]

[0029] Among them, S al l represents the total number of training sets for all partitions, S(n) represents the number of training sets for partition n, and k 2 Represents the number of active and reactive combinations of each distributed resource.

[0030] Optionally, the calculation of the power flow distribution result when there are no distributed resources in the target power system includes:

[0031] All distributed resources in the target power system are removed, and the power flow distribution is calculated according to the topology, line parameters, bus parameters and load model of the target power system;

[0032] The topological structure is the connection relationship between the busbars;

[0033] The line parameters are resistance, reactance, conductance and susceptance on the line;

[0034] The bus parameters and load model include constant power generator nodes and constant power load nodes on the bus.

[0035] Optionally, the calculating power flow distribution includes:

[0036] Calculate the bus voltage of each node, the active power of each branch and the reactive power of each branch, and use the calculated reactive power as the power flow distribution.

[0037] Optionally, determine the partition power flow calculation results for all partitions, including:

[0038] Determine the bus voltage of each node and the power of each branch. The determination method satisfies the following relationship:

[0039]

[0040] in, represents the calculated bus voltage value of node i without distributed resources, represents the calculated bus voltage value of node i in the nth training set partition m, It represents the influence of the power flow calculation result of the nth training set partition m on the voltage of node i; They represent the calculated active power and reactive power of branch j without distributed resources, They represent the calculated values ​​of active power and reactive power of branch j in the nth training set partition m, respectively. They respectively represent the impact of the power flow calculation results of the nth training set partition m on the active power and reactive power of branch j.

[0041] Optionally, merging the partition power flow calculation results of all partitions to obtain the power flow result of the target power system includes:

[0042] The influence of all partition power flow calculation results is summed and added to the power flow calculation results without distributed resources. The calculation formula satisfies the following relationship:

[0043]

[0044] in, represents the voltage calculation result of node i in the nth training set, and They represent the calculation results of active power and reactive power of the nth training set branch j respectively, and M represents the number of partitions.

[0045] In a second aspect, an embodiment of the present application provides a complex power system flow determination system, including a processor and a memory;

[0046] Memory, used to store computer programs;

[0047] The processor is used to implement any method step described in the first aspect when executing a program stored in the memory.

[0048] Beneficial effects:

[0049] The complex power system flow determination method provided by the present invention reduces the complexity of the machine learning training process by dividing the power system into different partitions and generating grid-divided training sets according to the respective partitions; then directly performing machine learning training for different partitions, and then merging the training results to achieve distributed training. The use of distributed training can solve the defect of large demand for traditional machine learning training sets, greatly improve the efficiency of the machine learning training process, and is suitable for flow calculation of large-scale complex power systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 One of the flow charts of the complex power system power flow determination method according to the preferred embodiment of the present invention;

[0051] Figure 2 This is the second flow chart of the complex power system power flow determination method according to the preferred embodiment of the present invention;

[0052] Figure 3 A flowchart of a training set generation method according to a preferred embodiment of the present invention;

[0053] Figure 4 A flowchart of a method for generating a training set according to grid division according to a preferred embodiment of the present invention;

[0054] Figure 5 One of the flowcharts of the training process and testing method of the machine learning model of the preferred embodiment of the present invention;

[0055] Figure 6 This is the second flowchart of the training process and testing method of the machine learning model of the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only a 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.

[0057] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0058] See also Figure 1-6 , the embodiment of the present application provides a method for determining a complex power system flow, including:

[0059] Divide the target power system into different partitions, perform grid division on all distributed resources in each partition, and generate the initial training set of each partition based on the grid division results;

[0060] The distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set are removed from the partition training set to obtain a partition final training set, and all partition final training sets are merged into a whole to obtain a whole training set;

[0061] Inputting the overall training model into the constructed machine learning model to obtain a trained machine learning model;

[0062] Calculate the power flow distribution results when there are no distributed resources in the target power system, and input the final training sets of all the partitions into the trained machine learning model to determine the partition power flow calculation results of all the partitions;

[0063] The partition power flow calculation results of all partitions are combined to obtain the power flow results of the target power system.

[0064] In the above embodiment, the complexity of the machine learning training process is reduced by dividing the power system into different partitions and generating grid-divided training sets according to the respective partitions; then the machine learning training is directly performed for the different partitions, and the training results are merged to achieve distributed training. The use of distributed training can solve the defect of the large demand for traditional machine learning training sets, greatly improve the efficiency of the machine learning training process, and is suitable for power flow calculations in large-scale complex power systems.

[0065] like Figure 3 As shown in the figure, for the training set generation method, first, the power system is divided into different partitions, each partition contains a similar number of distributed resources; secondly, all distributed resources in each partition are divided into grids to generate their own training sets; then, the output of distributed resources that do not belong to the corresponding partition in the training set is set to 0; finally, the training sets of all partitions are merged into a whole, and the power system is trained using this training set containing the whole.

[0066] like Figure 4 As shown, the grid division method means that the active and reactive outputs of all distributed resources in each partition are divided into equal intervals according to the output range of the corresponding distributed resources.

[0067] In this embodiment, equal interval division means that the output range of active output and reactive output is divided into equal intervals. For example, when the active output range is 20-100k w , the total number of segments is 4, then the active output segment interval is 20k w; When the reactive power output range is 60-100k w The total number of segments is 4, and the segment interval of reactive power output is 10k w In the embodiment, segmentation refers to segmentation of the output range. The number of training samples can be reduced by segmentation.

[0068] The symbol k represents the serial number of the segment, k ≥ 2. When k = 2, it means that the active and reactive output of each distributed resource is divided into two segments, namely, the upper and lower limits of the reactive output, recorded as and When k is another value (for example, k=K), the reactive power output of different segments is shown in equations (1) and (2) respectively.

[0069]

[0070] Among them, P l and Q l They represent the active and reactive output values ​​of the distributed resource l in the current training set, and They represent the upper limit of active output, the lower limit of active output, the upper limit of reactive output and the lower limit of reactive output of distributed resources respectively, k represents the segment number, and K represents the total number of segments.

[0071] The output of distributed resources that do not belong to the corresponding partition n is determined according to equations (3) and (4).

[0072]

[0073] Wherein, μ(n) represents the set of distributed resources of partition n.

[0074] The number of training sets for the nth partition is shown in formula (5), and the total number of training sets for all partitions is shown in formula (6), where s represents the sequence number of the training set, s∈(1, 2, ..., S all ).

[0075] S(n)=Π i∈μ(n) k 2 (5)

[0076]

[0077] Among them, S(n) represents the number of training sets in partition n, S all Represents the total number of training sets for all partitions. k 2 Represents the number of active and reactive combinations of each distributed resource.

[0078] The training set of power flow calculation represents the combination of different outputs of distributed resources as the input of the machine learning training set, and the power flow calculation results of these different combinations are used as the output of the machine learning training set.

[0079] Machine learning methods include linear regression, polynomial regression, support vector regression, decision tree regression, random forest regression, gradient boosting regression, AdaBoost regression, K nearest neighbor regression, neural network regression, elastic network regression, ridge regression, Lass o Regression, principal component regression, Bayesian regression, deep learning regression, etc.

[0080] like Figure 5 As shown in the figure, for the training process and testing method of the machine learning model, first, the flow distribution results without distributed resources are calculated, and the grid method is used to generate respective training sets for distributed resources in different partitions; secondly, the influence of the training sets of different partitions on the flow calculation results is recorded, and then the influence of the distributed resources of all partitions on the flow calculation results is merged; finally, the distributed resources of different partitions are input into the model trained by machine learning, and the influence of the flow calculation of each partition is obtained respectively, and the results of the flow calculation of each partition are merged to obtain the final flow calculation result based on machine learning.

[0081] The calculation of power flow distribution can be directly calculated by using perfect simulation tools such as MatPower. It is necessary to build a specific model, set the line parameters, power distribution, etc. in the simulation tool, and complete the calculation after running. The calculation results will be directly output in the output column. During use, just build the model in the simulation tool.

[0082] like Figure 6 As shown in Figure 1, first, all distributed resources in the target power system are removed, and the power flow distribution is calculated based on the topology, line parameters, bus parameters and load model of the target power system, including the bus voltage V of each node. 0 , Active power P of each branch 0 and reactive power Q 0 .

[0083] The impact on the bus voltage of each node and the power of each branch is shown in equations (7 to 9).

[0084]

[0085] in, represents the calculated bus voltage value of node i without distributed resources, represents the calculated bus voltage value of node i in the nth training set partition m, It represents the influence of the power flow calculation result of the nth training set partition m on the voltage of node i; They represent the calculated active power and reactive power of branch j without distributed resources, They represent the calculated values ​​of active power and reactive power of branch j in the nth training set partition m, respectively. They respectively represent the impact of the power flow calculation results of the nth training set partition m on the active power and reactive power of branch j.

[0086] The impact of all partition power flow calculation results is summed and added to the power flow calculation results without distributed resources, as shown in equations (10 to 12).

[0087]

[0088] in, represents the voltage calculation result of node i in the nth training set, and They represent the calculation results of active power and reactive power of the nth training set branch j respectively, and M represents the number of partitions.

[0089] The embodiment of the present application also provides a complex power system flow determination system, including a processor and a memory;

[0090] Memory, used to store computer programs;

[0091] The processor is used to implement any method step in the complex power system power flow determination method when executing the program stored in the memory.

[0092] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for determining power flow in a complex power system, characterized in that: include: Divide the target power system into different partitions, perform grid division on all distributed resources in each partition, and generate the initial training set of each partition based on the grid division results; The distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set are removed from the partition training set to obtain a partition final training set, and all partition final training sets are merged into a whole to obtain a whole training set; Inputting the overall training model into the constructed machine learning model to obtain a trained machine learning model; Calculate the power flow distribution results when there are no distributed resources in the target power system, and input the final training sets of all the partitions into the trained machine learning model to determine the partition power flow calculation results of all the partitions; The partition power flow calculation results of all partitions are combined to obtain the power flow results of the target power system.

2. The complex power system power flow determination method according to claim 1, characterized in that: The target power system is divided into different partitions, including: The target power system is partitioned based on the number of distributed resources in the target power system so that the number of distributed resources in each partition is equal.

3. The complex power system power flow determination method according to claim 1, characterized in that: The grid division of all distributed resources in each partition includes: Determine the output range of active power and reactive power of all distributed resources in each partition, and divide all distributed resources in each partition into grids by dividing the output range into equal intervals; Among them, the equal interval division methods include: The symbol k represents the serial number of the segment, k ≥ 2. When k = 2, it means that the active and reactive outputs of each distributed resource are divided into two segments, namely, the upper and lower limits of the active output and the upper and lower limits of the reactive output, recorded as and When k is other values, the active output and reactive output of different segments are as follows: Among them, P l and Q l They represent the active and reactive output values ​​of the distributed resource l in the current training set, and They represent the upper limit, lower limit, upper limit and lower limit of active output of distributed resource l respectively, k represents the segment number, and K represents the total number of segments.

4. The complex power system power flow determination method according to claim 1, characterized in that: The step of removing the distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set from the partition training set includes: The active output value and reactive output value of the distributed resources in the partition initial training set that do not belong to the partition corresponding to the partition initial training set are taken according to the elimination formula, and the elimination formula satisfies the following relationship: Among them, P l and Q l They represent the active and reactive output values ​​of the distributed resource l in the current training set, μ(n) represents the set of distributed resources in partition n, and l represents the distributed resource.

5. The complex power system power flow determination method according to claim 1, characterized in that: The final training sets of all partitions are combined into a whole to obtain the overall training set, including: The number of final training sets for each partition is determined by satisfying the following relationship: S(n)=∏ i∈μ(n) k 2 ; Among them, S(n) represents the number of training sets in partition n, k 2 represents the number of active and reactive permutations of each distributed resource, μ(n) represents the set of distributed resources in partition n; The number of all final training sets is determined based on the number of final training sets for each partition, and the determination method satisfies the following relationship: Among them, S all represents the total number of training sets for all partitions, S(n) represents the number of training sets for partition n, and k 2 It represents the number of active and reactive combinations of each distributed resource, and N represents the total number of partitions.

6. The complex power system power flow determination method according to claim 1, characterized in that: The power flow distribution result when there is no distributed resource in the target power system is calculated, including: All distributed resources in the target power system are removed, and the power flow distribution is calculated according to the topology, line parameters, bus parameters and load model of the target power system; The topological structure is the connection relationship between the busbars; The line parameters are resistance, reactance, conductance and susceptance on the line; The bus parameters and load model include constant power generator nodes and constant power load nodes on the bus.

7. The complex power system power flow determination method according to claim 6, characterized in that: The calculating power flow distribution comprises: Calculate the bus voltage of each node, the active power of each branch and the reactive power of each branch, and use the calculated reactive power as the power flow distribution.

8. The complex power system power flow determination method according to claim 1, characterized in that: Determine the zone power flow calculation results for all zones, including: Determine the bus voltage of each node and the power of each branch. The determination method satisfies the following relationship: in, represents the calculated bus voltage value of node i without distributed resources, represents the calculated bus voltage value of node i in the nth training set partition m, represents the impact of the power flow calculation result of the nth training set partition m on the voltage of node i, They represent the calculated active power and reactive power of branch j without distributed resources, They represent the calculated values ​​of active power and reactive power of branch j in the nth training set partition m, respectively. They respectively represent the impact of the power flow calculation results of the nth training set partition m on the active power and reactive power of branch j.

9. The complex power system power flow determination method according to claim 1, characterized in that: The step of combining the partition power flow calculation results of all partitions to obtain the power flow result of the target power system includes: The influence of all partition power flow calculation results is summed and added to the power flow calculation results without distributed resources. The calculation formula satisfies the following relationship: in, represents the voltage calculation result of node i in the nth training set, and They represent the calculation results of active power and reactive power of the nth training set branch j, M represents the number of partitions, represents the calculated bus voltage value of node i without distributed resources, They represent the calculated active power and reactive power of branch j without distributed resources, represents the impact of the power flow calculation result of the nth training set partition m on the voltage of node i, They respectively represent the impact of the power flow calculation results of the nth training set partition m on the active power and reactive power of branch j.

10. A complex power system flow determination system, characterized in that: Including processor and memory; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 9 when executing a program stored in a memory.

Images