MMC capacitor voltage dynamic balance control method based on K-means algorithm

Abstract

The invention relates to an MMC capacitor voltage dynamic balance control method based on a K-means algorithm, and the method comprises the steps: S1, determining the number of submodules needed to be input in a current control period through a modulation strategy at the beginning of each control period; s2, the capacitor voltage of each sub-module on the bridge arm at the current moment is collected; s3, using a K-means algorithm to cluster the capacitor voltage of the sub-module; s4, respectively executing corresponding operations according to the current directions of the bridge arms; s5, sorting the clusters according to the size of the mass center of each cluster; s6, according to the current direction of the bridge arm and the sorting result of the clusters, the corresponding sub-modules are input in sequence; compared with a traditional direct voltage balancing method based on capacitor voltage sorting, the method has the advantages that the switching frequency is reduced, and the capacitor voltage volatility is remarkably reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of power electronics, and in particular relates to a MMC capacitor voltage dynamic balancing control method based on a K-means algorithm. Background Art

[0002] In recent years, modular multilevel converters (MMC) have been widely used in high-voltage direct current transmission, high-power motor drive and other fields due to their advantages such as high efficiency, low harmonics, and flexible scalability. As China has increased its support for green energy transformation and smart grid construction in its planning, the application of MMC in my country has developed rapidly, especially in large-scale offshore wind power access, cross-regional grid interconnection, and ultra-high voltage direct current (UHVDC) transmission projects. MMC has shown important technical advantages. However, in practical applications, the problem of dynamic balance of sub-module capacitor voltage still exists. Unbalanced capacitor voltage can lead to unstable operation of the MMC system, damage to switching devices, deterioration of power quality, module overload failure, and even serious equipment damage and system paralysis.

[0003] There are three main types of MMC sub-module capacitor voltage balancing control methods: the first type is to achieve sub-module capacitor voltage balancing by adding an external voltage balancing circuit, which is simple but increases the hardware cost and complexity of the system; the second type is a direct voltage balancing method based on capacitor voltage sorting. Although the principle is simple and easy to implement, it will cause large capacitor voltage fluctuations when the calculation amount is large, the switching frequency is high, and the number of sub-modules is large; the third type is to add two additional voltage closed-loop controllers, namely independent voltage balancing control and voltage stabilization control, to the MMC sub-module to avoid switching losses, but it is more difficult to adjust the controller parameters, especially when the number of sub-modules is large. Summary of the invention

[0004] The present invention proposes a MMC capacitor voltage dynamic balancing control method based on the K-means algorithm, which aims to solve the problem of large capacitor voltage fluctuation and switching loss caused by the second type of direct voltage balancing method based on MMC submodule capacitor voltage sorting in the background technology when the number of submodules is large. This method can not only reduce the number of switching times of submodules during MMC operation to a certain extent and reduce the overall switching frequency, but also reduce capacitor voltage fluctuation when the number of submodules is large.

[0005] A method for dynamic balancing control of MMC capacitor voltage based on K-means algorithm comprises the following steps:

[0006] S1: At the beginning of each control cycle, the number of submodules required for the current control cycle is determined through the modulation strategy;

[0007] S2: Collect the capacitor voltages U of each sub-module on the arm at the current moment ci ;

[0008] S3: Use the K-means algorithm to cluster the sub-module capacitor voltages

[0009] S4: Perform corresponding operations respectively according to the arm current direction

[0010] S5: Sort the clusters according to the magnitudes of the centroids of each cluster

[0011] S6: Input the corresponding sub-modules in sequence according to the sorting result of the clusters according to the arm current direction

[0012] Further, S1 includes the following steps: Calculate the number N of MMC sub-modules to be input on each arm at a certain moment according to the modulation strategy selected by the control system on , each arm contains N sub-modules. If N on = N, then all MMC sub-modules on this arm are put into use; if N on = 0, then all MMC sub-modules on this arm are cut off; if 0 < N on < N, then enter step S2

[0013] Further, S3 includes the following steps: Perform K-means algorithm clustering on the dataset of sub-module capacitor voltages U ci to separate several clusters

[0014] Further, the K-means algorithm includes: First, initialize: Select k initial clustering centroids, evenly distributed around the rated voltage of the sub-module capacitor. For the sub-module capacitor voltage U ci , calculate its distance from each cluster centroid and assign it to the cluster with the closest distance, recalculate each cluster centroid, and repeat the assignment and update steps until the specified number of iterations d is reached

[0015] Further, use the Euclidean distance as the distance metric for the distance, and the calculation formula is as follows:

[0016]

[0017] where U ci is the sub-module capacitor voltage, and C i is the centroid of the i-th cluster

[0018] Further, for recalculating each cluster centroid, the new cluster centroid is the average value of all capacitor voltages within the cluster, and the calculation formula is as follows:

[0019]

[0020] Where S i is the number of capacitor voltage data points in the i-th cluster.

[0021] Further, S4 includes: setting the bridge arm current i arm Direction and Figure 1 The same as in the figure, the bridge arm current i arm The positive direction is from the high pressure end to the low pressure end. arm When greater than 0, if the number of data points S in the cluster S with the smallest centroid a Much larger than the number of submodules N required on , then execute S3 again for cluster S; otherwise, execute S5; when i arm When it is less than 0, if the number of data points in the cluster B with the largest centroid is B a Much larger than the number of submodules N required on , then execute S3 again for cluster B; otherwise, execute S5.

[0022] Further, S6 includes: setting the bridge arm current i arm Direction and Figure 1 Same as in, when i arm When it is greater than 0, N clusters are selected in sequence starting from the cluster with the smaller centroid. on submodules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors far from the centroid are selected from the current cluster. arm When it is less than 0, select N clusters starting from the cluster with the larger centroid. on Sub-modules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors that are far from the centroid are preferentially selected from the current cluster for extraction.

[0023] The invention is characterized in that it makes full use of the advantages of K-means algorithm clustering and clustering, clusters the submodule capacitor voltage, and the voltage values ​​in each cluster are concentrated and do not need to be sorted. It only needs to compare the size of the cluster centroid and then sort the clusters, which solves the shortcomings of the traditional sorting, such as accurate comparison and multiple actions, and can reduce the number of submodule switching, reduce the switching frequency and reduce the loss; by adjusting the parameters k and d in the algorithm, and the number of submodules N calculated according to the modulation strategy on , achieving dynamic adjustment; in addition, the present invention also significantly reduces the capacitor voltage fluctuation rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the prior art and the drawings required for use in the embodiments. The following drawings are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A model diagram of a modular multi-level converter system of the present invention;

[0026] Figure 2 It is a flow chart of the capacitor voltage dynamic balancing control method of the present invention;

[0027] Figure 3 This is a capacitor voltage waveform diagram after the upper bridge arm submodule of phase A of the modular multilevel converter of the present invention is stabilized;

[0028] Figure 4 This is a trigger pulse diagram of the IGBT of the upper bridge arm submodule of phase A of the modular multilevel converter of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all other embodiments obtained by ordinary technicians in this field without creative work based on the embodiments of the present invention are within the scope of protection of the present invention.

[0030] The embodiment of the present application provides a method for dynamic balanced control of MMC capacitor voltage based on K-means algorithm, which is applicable to balanced control of capacitor voltage of MMC submodules. The method can reduce the number of switching times of submodules during MMC operation to a certain extent, reduce the overall switching frequency, and effectively reduce capacitor voltage fluctuation when there are a large number of submodules. At the same time, the method also supports dynamic adjustment algorithm to achieve more efficient balanced control.

[0031] As a typical embodiment of the present invention, a 49-level modular multi-level converter system is taken as an example for description.

[0032] The present invention uses Matlab / Simulink simulation platform to simulate the modular multilevel converter, and the system uses the nearest level approximation modulation. In the simulation model, each bridge arm of the modular multilevel converter contains 48 submodules, no redundant submodules are configured, and the rated voltage of the DC side is U dc =48kV, the rated voltage of each submodule capacitor is U cn=1kV, capacitance C0 = 670μF, the initial value of the capacitance voltage of each submodule of the bridge arm is distributed within ±30%U cn , the rated capacitance is distributed within ±30% C0, the bridge arm inductance is L0 = 10mH, the sampling frequency is 10kHz, and the control period is 100μs. The specific implementation is as follows:

[0033] According to S1 in the manual, at the beginning of each control cycle, the number of submodules N required for the current control cycle is determined by the nearest level approximation. on ;

[0034] According to S2, the capacitor voltage U of each submodule on the bridge arm at the current moment is collected ci ;

[0035] According to S3, the submodule capacitor voltage U ci This data set is clustered by K-means algorithm to divide into several clusters. K-means algorithm includes the following steps: First, initialization: select 3 initial cluster centroids, which are 950, 1000, and 1050 respectively. The value of k must be greater than or equal to 1. The larger the k value, the smaller the capacitance fluctuation rate, but the number of switching times will increase accordingly. The selection of k value usually needs to be reasonably set according to the number of system submodules; then for the submodule capacitance voltage U ci , calculate its distance to each cluster centroid and assign it to the cluster with the closest distance, recalculate each cluster centroid, and repeat the assignment and update steps until the specified number of iterations d = 3 is reached.

[0036] Specifically, the Euclidean distance is used as the distance metric, and the calculation formula is as follows:

[0037]

[0038] Among them U ci is the submodule capacitor voltage, C i is the centroid of the ith cluster.

[0039] Specifically, each cluster centroid is recalculated, and the new cluster centroid is the mean value of all capacitor voltages in the cluster, and the calculation formula is as follows:

[0040]

[0041] Where S i is the number of capacitor voltage data points in the i-th cluster.

[0042] According to S4: let the bridge arm current i arm Direction and Figure 1 The same as in the figure, the bridge arm current i arm The positive direction is from the high pressure end to the low pressure end. armWhen greater than 0, if the number of data points S in the cluster S with the smallest centroid a Much larger than the number of submodules N required on , then execute S3 again for cluster S, and the submodule capacitor voltage U in cluster S ci This data set is clustered using the K-means algorithm to divide it into several clusters; otherwise, S5 is executed to sort the clusters according to the size of the centroid of each cluster; when i arm When it is less than 0, if the number of data points in the cluster B with the largest centroid is B a Much larger than the number of submodules N required on , then execute S3 again for cluster B; otherwise, execute S5.

[0043] Specifically, it can be selected as 5 times or 10 times according to the number of submodules. In the example, due to the small number of modules, 5 times can be selected, that is, S a >5N on Can be considered as much greater than.

[0044] According to S6: let the bridge arm current i arm Direction and Figure 1 Same as in, when i arm When it is greater than 0, select N clusters starting from the cluster with the smaller centroid. on Sub-modules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors far from the centroid are selected from the current cluster to charge these sub-module capacitors. arm When it is less than 0, select N clusters starting from the cluster with the larger centroid. on Sub-modules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors farther from the centroid are preferentially selected from the current cluster to discharge the capacitors of these sub-modules.

[0045] Depend on Figure 3 It can be seen that under the implementation of the above capacitor voltage control method, the capacitor voltage control effect is good, and the maximum fluctuation rate ε = 1.4% can meet the demand. Figure 4 It can be seen that under this capacitor voltage control method, the trigger pulses of the switching device are relatively sparse, and the average switching frequency of a single IGBT is 1.2kHz, which reduces the switching frequency.

[0046] Therefore, the MMC capacitor voltage dynamic balancing control method based on the K-means algorithm proposed in the present invention can reduce the switching frequency, effectively reduce the capacitor voltage fluctuation rate, and improve the capacitor voltage balancing effect compared with the traditional direct voltage balancing method based on capacitor voltage sorting, which further verifies the accuracy of the proposed method.

Claims

1. A method for dynamic balancing control of MMC capacitor voltage based on K-means algorithm, characterized in that: include: S1: At the beginning of each control cycle, the number of submodules required for the current control cycle is determined through the modulation strategy; S2: Collect the capacitor voltage U of each submodule on the bridge arm at the current moment ci ; S3: Use K-means algorithm to cluster the collected sub-module capacitor voltage; S4: Execute corresponding operations according to the direction of the bridge arm current; S5: Sort the clusters according to the size of the centroid of each cluster; S6: According to the direction of the bridge arm current and the sorting result of the clusters, the corresponding sub-modules are put into operation in sequence.

2. The MMC capacitor voltage dynamic balancing control method based on the K-means algorithm according to claim 1, characterized in that: The K-means algorithm in step S3 includes: first initialization: selecting k initial cluster centroids, evenly distributed around the rated voltage of the submodule capacitor, for the submodule capacitor voltage U ci , calculate its distance to each cluster centroid and assign it to the cluster with the closest distance, recalculate each cluster centroid, and repeat the assignment and update steps until the specified number of iterations d is reached.

3. The MMC capacitor voltage dynamic balancing control method based on the K-means algorithm according to claim 2, characterized in that: The distance is calculated using the Euclidean distance as the distance metric, and the calculation formula is as follows: Among them U ci is the submodule capacitor voltage, C i is the centroid of the ith cluster; For each cluster centroid, the new cluster centroid is the mean of all capacitor voltages in the cluster, and the calculation formula is as follows: Where S i is the number of capacitor voltage data points in the i-th cluster.

4. The MMC capacitor voltage dynamic balancing control method based on the K-means algorithm according to claim 1, characterized in that: In step S4, the bridge arm current i arm The positive direction is from the high pressure end to the low pressure end. arm When greater than 0, if the number of data points S in the cluster S with the smallest centroid a Much larger than the number of submodules N required on , then execute S3 again for cluster S; Otherwise, execute S5; when i arm When it is less than 0, if the number of data points in the cluster B with the largest centroid is B a Much larger than the number of submodules N required on , then execute S3 again for cluster B; otherwise, execute S5.

5. The MMC capacitor voltage dynamic balancing control method based on K-means algorithm according to claim 1, characterized in that: In step S6, the bridge arm current i arm The positive direction is from the high pressure end to the low pressure end. arm When it is greater than 0, select N clusters starting from the cluster with the smaller centroid. on submodules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors far from the centroid are selected from the current cluster. arm When it is less than 0, select N clusters starting from the cluster with the larger centroid. on Sub-modules are put into use and selected cluster by cluster. If the remaining number of available modules is not enough to meet the needs of the next cluster, capacitors that are far from the centroid are preferentially selected from the current cluster for extraction.

Images