A control method for intelligent single-unit converter

Abstract

The present invention discloses a control method for an intelligent single-cell converter, including: constructing a network line operation model, designing an intelligent online identification scheme for power grid parameters; introducing edge computing technology, constructing a local computing optimization scheme; researching the power grid support technology of distributed grid-connected converters, and forming an auxiliary algorithm for realizing the intelligence of single-cell converters. The present invention belongs to the field of new energy grid-connected technology, and specifically provides a control method for an intelligent single-cell converter, which is used to solve the problems of low intelligence level and heavy computational burden of single-cell converters, as well as the difficulty of identifying power grid impedance parameters, introduces modern artificial intelligence technology, improves the intelligence level of single-cell converters, develops the auxiliary functions of single-cell converters for intelligent monitoring of power grid status and effective identification of power grid impedance parameters, and eliminates the influence of power grid parameter changes on the control system.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy grid-connected technology, and specifically refers to a control method for an intelligent single-unit converter. Background Art

[0002] As the installed capacity of renewable energy continues to increase, its coupling effect on the power grid is becoming increasingly significant, bringing new challenges to the safe and stable operation of the power grid. Renewable energy has significant volatility and intermittency. The large-scale access of power electronic equipment at the source-load end has significantly reduced the system inertia. The traditional power grid has a low degree of acceptance and poor operational reliability. The grid-connected converter is the interface equipment for the access of new energy. Energy conversion and frequency-voltage regulation are all achieved through the control of the grid-connected converter. Therefore, the cluster of grid-connected converter equipment and its intelligent control are key technologies to overcome the above-mentioned difficulties, achieve a high proportion of friendly access to new energy, and provide reliable support for the power grid. The development of intelligent, highly reliable, and collaboratively optimized grid-connected converter equipment has become a consensus in academia and industry, and it is urgent to tackle the problem.

[0003] With the rapid development of power grids and information technology, converter equipment and its intelligent control play an increasingly important role, and converter equipment and its intelligent control have penetrated into every link of power generation, transmission, distribution and use. The future smart grid with high renewable energy penetration integrates wind energy, solar energy, energy storage, and flexible loads. It is a complex dynamic system across multiple time scales that integrates chemical, mechanical and electromagnetic processes. In the above context, there is an urgent need to develop more active control functions for converter equipment and provide intelligent support and flexible auxiliary services according to the operating status of the power grid. However, the existing solutions have the following fundamental defects.

[0004] (1) Low level of intelligence and heavy computing burden for single machines. Currently, there is no mature solution for intelligent functions such as intelligent sensing, status detection and monitoring, data link sharing and intercommunication in single-machine grid-connected converter equipment, and the potential of intelligent monitoring and scheduling of single converters has not been fully utilized.

[0005] (2) Failure to fully utilize the coordinated output of “source-load”. Renewable energy can only support the power grid by “reducing production and conservative output”, which fails to fully utilize the installed capacity of renewable energy and sacrifices the system conversion efficiency to a certain extent; it fails to fully utilize the flexible loads driven by the inverter (such as motor drives, etc.) and mobile energy storage devices (such as electric vehicles, etc.) to participate in the active regulation of source-load power balance. Summary of the invention

[0006] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a control method for an intelligent single-cell converter, which is used to solve the problems of low intelligence and heavy computational burden of single-cell converters. The parameters of the power grid with high new energy penetration rate change frequently, and the monitoring, protection and control schemes based on the impedance parameters of the power grid are difficult to apply. In the future power grid with high new energy penetration rate, each distributed power source is connected to the power grid through a power electronic converter, and in order to adapt to the intermittent production capacity characteristics of new energy, the source-end converter often needs to flexibly switch the working mode and actively disconnect / connect to the grid, which further makes the impedance parameters of the power grid difficult to identify. The existing single-cell converter can only perform a preset single function, which is difficult to meet the above-mentioned operating requirements. Therefore, introducing modern artificial intelligence technology, improving the intelligence level of single-cell converters, and developing auxiliary functions of single-cell converters for intelligent monitoring of the power grid status and effective identification of the power grid impedance parameters are effective ways to eliminate the impact of power grid parameter changes on the control system and ensure efficient and stable operation of the power grid.

[0007] For large-scale new energy grid-connected converter cluster systems, we focus on achieving highly intelligent individual machines and conduct in-depth research in order to realize technical solutions for smart, efficient and reliable grid-connected consumption of new energy.

[0008] The technical solution is as follows:

[0009] The present invention provides a control method for an intelligent single-unit converter, comprising the following steps:

[0010] Step S1, constructing a network line operation model and designing an intelligent online identification scheme for power grid parameters:

[0011] Firstly, the grid parameter impedance and moment of inertia parameters of the power electronic converter cluster are collected and analyzed, the influence of the distribution and operation mode of the "source-load" power electronic interface converter on the grid state parameters is explored, and the parameter change rules are summarized to form a grid state parameter relationship comparison data set.

[0012] Then, a model of the power grid line and network state parameters with high renewable energy penetration is constructed based on the data set. Modern artificial intelligence technologies such as deep learning and reinforcement learning are introduced to build a digital twin system of the smart grid to realize online identification of the power grid state and parameters. The recursive least squares method is used to find the best function match for the data by minimizing the sum of squares of the errors. The least squares method can be used to easily obtain unknown data and minimize the sum of squares of the errors between the obtained data and the actual data. Online observation saves all observations and uses the classical least squares method to solve them, which consumes a lot of computing resources and memory, so the recursive least squares method is needed to update the parameters of the linear model online.

[0013] Finally, the actual and identified parameters are compared to further revise the intelligent identification algorithm.

[0014] Step S2: Introduce edge computing technology and construct a local computing optimization solution:

[0015] First, according to the online identification algorithm obtained in step S1, the local parameter operation and the global information identification operation set are stripped off, and the local data and the global identification software interface are standardized according to the operation set.

[0016] Then, edge computing technology is introduced to use the local inverter controller to perform operations on local data sets, and global information synchronization is performed according to the standard unified interface to reduce the computing burden of the server.

[0017] Finally, the influence of time delay on the intelligent state monitoring and parameter identification algorithm is explored, and the online parameter identification algorithm is further modified.

[0018] Step S3, research on distributed grid-connected converter power grid support technology:

[0019] In practical applications, in order to adapt to the irregular and intermittent production characteristics of new energy, the disconnection / grid-connection status and production mode of multiple distributed power sources change over time. The project intends to analyze and study the operating status characteristics of the converter, summarize the converter status under multiple working conditions, and build a converter grid-connected status and production mode identification scheme.

[0020] Secondly, the distributed power grid support technology (voltage / frequency) of converter clusters is studied, and according to the converter status under multiple working conditions, a support control scheme that can adapt to seamless connection of multiple states is studied.

[0021] Finally, combined with step S1 and step S2, an auxiliary algorithm for realizing the intelligence of the single-cell converter is formed.

[0022] As a further elaboration of the scheme, the control scheme in step S3 is an improved droop control method based on virtual impedance optimization, and the control scheme includes two components: virtual impedance adaptive adjustment and feeder impedance identification.

[0023] The virtual impedance adaptive adjustment specifically includes the following operations:

[0024] First, the local measurement unit collects the data of PCC voltage and current. The impedance identification module receives the data at this point. The recursive least squares method is used to calculate and output the impedance identification result.

[0025] Secondly, Z L The optimal virtual impedance coordination value is calculated by the central controller. The central controller of each distributed power source improves the droop control by introducing virtual impedance. The traditional droop control is used to ensure active power sharing.

[0026] In this step, in order to correct the imbalance of the impedance between DG1 (distributed generation) and DG2, the adaptive virtual impedance controller is used according to ZL Real-time adjustment of virtual impedance Z v , so that the equivalent line impedance Z of each DG unit (distributed generation unit) ei equal.

[0027] Furthermore, the core part of the proposed reactive power sharing strategy, the feeder impedance identification scheme, is responsible for the identification of feeder impedance, which specifically includes the following operations:

[0028] In order to reduce the workload and calculation amount, it has two working modes: recognition mode and standby mode.

[0029] Only when a pulse trigger signal is received will it enter the identification mode. Otherwise, it will always run in standby mode, waiting for a trigger signal with a higher control level.

[0030] In practical applications, when the measured output reactive power deviation between inverters exceeds the upper limit, the identification mode can be triggered. In the identification mode, the recursive least squares algorithm is used to estimate the Z of each DG unit. L The voltage and current measurement unit first obtains the inverter output voltage V o The value of m groups P CC The voltage value V pcc and current value I pcc .

[0031] The relationship when the inverter is running stably is shown in equation (1):

[0032]

[0033] In formula (1), V pcc Indicates P CC Voltage vector; I pcc Represents P CC Current vector; V o Represents the inverter output voltage vector; Z L is the feeder impedance Z L =R L +jX L .

[0034] Formula (1) can be simplified to a matrix form Y=AX.

[0035]

[0036] In order to describe the deviation between the identified value and the actual impedance value of the feeder, the cost function is defined as formula (3), as follows:

[0037]

[0038] In order to minimize the value of the cost function, the partial derivative of J must be calculated. Let The optimal feeder impedance identification value X is obtained as shown in the following formula (4).

[0039]

[0040] The traditional least squares method can only realize offline recognition because it needs batch sampling calculation. In order to realize online recognition and prevent the main line saturation of the recognition result, a recursive least squares algorithm with forgetting factor is proposed, as shown in the following formula (5).

[0041]

[0042]

[0043] In the formula, X k is the feeder impedance identification result at the kth moment. k+1 Y is the feeder impedance identification result at time (k+1). k+1 and a k+1 They are the new voltage measurement data and current measurement data from time k to time (k+1). k+1 is the correction term. λ is the forgetting factor. The higher its value, the smaller the impact on the final identification result and the smaller the impact of the prior voltage and current measurement data. This patent takes λ as 0.9. By using multiple iterations of formula (5), the online identification of feeder impedance can be realized. Without injecting specific subharmonics, the non-intrusive recursive least squares algorithm with forgetting factor is used to realize feeder impedance identification, which ensures the stability of feeder impedance and provides high-quality power supply to the load.

[0044] In order to achieve power sharing, virtual impedance is introduced. The idea of ​​reconstructing line impedance is to introduce a virtual voltage drop through current feedforward. After the introduction of virtual impedance, the simplified electrical block diagram of the distributed generator set can be equivalent to Figure 3 . Virtual impedance Z vi With feeder impedance Z Li The sum is the final equivalent impedance Z ei .

[0045] Traditional droop control adjusts the output voltage amplitude and frequency of each inverter according to the active and reactive power output of each DG. The droop control equation can be expressed as shown in the following equation (6):

[0046] f=f n +n(P n -P)#(6)

[0047] U=U 0 -nQ

[0048] In formula (6), m is the active frequency droop characteristic coefficient, n is the reactive voltage droop characteristic coefficient, and f n is the rated frequency of the power grid, Pn is the output active power of the micro power supply at rated frequency, U 0 is the voltage amplitude when the reactive power output by the micro power supply is 0, and P and Q are the measured values ​​of the active and reactive power output by the inverter power supply, respectively.

[0049] Adaptive virtual impedance is added to the traditional droop control strategy to compensate for the voltage drop caused by line differences and eliminate the impact of impedance differences on reactive power distribution. Directly adding virtual impedance values ​​requires known line impedance parameters. To avoid using line impedance values, the virtual impedance is adjusted by load power. The voltage drop generated on the virtual impedance compensates for the voltage drop caused by DG line impedance differences, achieving reasonable power distribution. When the load power factor is different, the required virtual impedance value is different, so the virtual impedance value should be adaptively adjusted as the load power factor changes.

[0050] The DG adjusts the virtual impedance value through the integral controller based on the difference between the received reactive power output reference value and the actual output reactive power. The virtual impedance is superimposed on the voltage reference value obtained by the droop control. The integrator can automatically adjust the virtual impedance according to the reactive power, without being affected by the inconsistency of the DG line impedance, to adapt to the change of the operating state and realize the reasonable distribution of reactive power. When the virtual impedance is adjusted to the point where the actual output reactive power of the DG is equal to the reactive power reference value, the virtual impedance value will no longer change. The virtual impedance value will only change when the load power factor changes.

[0051] In order to achieve power sharing, it is necessary to ensure that the introduced virtual impedance makes the final equivalent impedance of each DG consistent. e1 =Z e2 =Z e . Z vi The selection of is crucial to system performance. When the enable end of the virtual impedance controller receives a trigger signal, the local controller of the DG unit receives the virtual impedance information. In order to avoid unnecessary voltage loss, we take the average value of all feeder impedances as the equivalent impedance reference value, as shown in equation (7), and then use equation (8) to determine the optimal virtual impedance.

[0052]

[0053] Z vi =Z e -Z Li =(R e -R Li )+j(X e -X Li )#(8)

[0054] In the above formula, Ze is the globally consistent equivalent impedance after compensation, and Zvi is the optimal value of the virtual impedance.

[0055] The beneficial effects achieved by the present invention using the above structure are as follows:

[0056] 1. Introducing modern artificial intelligence and edge computing technologies into the control and design of single-cell converters, breaking the traditional monitoring and control architecture, significantly improving the intelligence level of single-cell converters, and providing new ideas for the single-cell design of grid-connected converters to meet future needs.

[0057] 2. At the same time, the influence of impedance difference on reactive power distribution is eliminated, and the virtual impedance is adaptively adjusted as the load power factor changes.

[0058] 3. Single-cell converters are the basic components of future power grids with high penetration rates of renewable energy. By introducing modern artificial intelligence algorithms and developing source-end converter grid support technologies, the underlying grid-connected equipment can realize intelligent status collection, monitoring and voltage / frequency support functions for the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A research roadmap for intelligent converters that provides a control method for an intelligent single converter for this solution;

[0060] Figure 2 Schematic diagram of improved droop control based on virtual impedance optimization in this scheme;

[0061] Figure 3 Virtual impedance control strategy for AC microgrid. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0063] Example 1

[0064] Single-cell converters are the basic components of future power grids with high renewable energy penetration. By introducing modern artificial intelligence algorithms and developing source-end converter grid support technologies, the underlying grid-connected equipment can realize intelligent status collection, monitoring and voltage / frequency support functions for the power grid.

[0065] Please refer to Figure 1-3 As shown, this embodiment provides a control method for an intelligent single-unit converter, and the specific method is as follows:

[0066] First, the grid parameter impedance and moment of inertia parameters of the power electronic converter cluster are collected and analyzed, the influence of the distribution and operation mode of the "source-load" power electronic interface converter on the grid state parameters is explored, and the parameter change law is summarized to form a grid state parameter relationship comparison data set. According to the data set, a high-energy penetration grid line and network state parameter model is constructed, and the recursive least squares method is introduced to find the best function match for the data by minimizing the sum of squares of the error. The least squares method can be used to easily obtain unknown data and minimize the sum of squares of the errors between the obtained data and the actual data. The online identification of the grid state and parameters is realized. By comparing the actual and identified parameters, the intelligent online identification algorithm is further corrected.

[0067] According to the online identification algorithm obtained above, the local parameter operation and global information identification operation set are stripped, and the local data and global identification software interface are standardized according to the operation set. The edge computing technology is introduced, and the local converter controller is used to perform operations on the local data set, and the global information is synchronized according to the standard unified interface to reduce the computing burden of the server. The impact of delay on the intelligent state monitoring and parameter identification algorithm is explored, and the online parameter identification algorithm is further modified.

[0068] The control structure of the method is as follows Figure 2 As shown. Virtual impedance adaptive adjustment and feeder impedance identification are two components of this control technology. First, the local measurement unit collects data on PCC voltage and current. The impedance identification module receives data at this point. The recursive least squares method is used to calculate and output the impedance identification result. Secondly, ZL is sent to the central controller to calculate the optimal virtual impedance coordination value. Each DG central controller improves the droop control by introducing virtual impedance composition. Traditional droop control is used to ensure active power sharing. In order to correct the imbalance of DG1 and DG2 line impedance, the adaptive virtual impedance controller adjusts the virtual impedance Z in real time according to ZL. v , so that the equivalent line impedance Z of each DG unit ei equal.

[0069] Figure 3 is the virtual impedance control strategy of AC microgrid. After the introduction of virtual impedance, the simplified electrical block diagram of distributed generator set can be equivalent to Figure 3 .

[0070] Feeder impedance identification, the core part of the proposed reactive power sharing strategy, is responsible for the identification of feeder impedance. In order to reduce the workload and calculation, it has two working modes: identification mode and standby mode. It will enter the identification mode only when a pulse trigger signal is received. Otherwise, it will always run in standby mode, waiting for a trigger signal with a higher control level. In practical applications, the identification mode can be triggered when the measured output reactive power deviation between inverters exceeds the upper limit. In the identification mode, the Z of each DG unit is estimated using the recursive least squares algorithm. L The voltage and current measurement unit first obtains the inverter output voltage V o The value of the PCC voltage V pcc and current value I pcc When the inverter is running stably, there is a relationship as shown in equation (1).

[0071] The relationship when the inverter is running stably is shown in equation (1):

[0072]

[0073] In formula (1), V pcc Indicates P CC Voltage vector; I pcc Represents P CC Current vector; V o Represents the inverter output voltage vector; Z L is the feeder impedance Z L =R L +jX L .

[0074] Formula (1) can be simplified to a matrix form Y=AX.

[0075]

[0076] In order to describe the deviation between the identified value and the actual impedance value of the feeder, the cost function is defined as formula (3), as follows:

[0077]

[0078] In order to minimize the value of the cost function, the partial derivative of J must be calculated. Let The optimal feeder impedance identification value X is obtained as shown in the following formula (4).

[0079]

[0080] The traditional least squares method can only realize offline recognition because it needs batch sampling calculation. In order to realize online recognition and prevent the main line saturation of the recognition result, a recursive least squares algorithm with forgetting factor is proposed, as shown in the following formula (5).

[0081]

[0082]

[0083] In the formula, X k is the feeder impedance identification result at the kth moment. k+1 Y is the feeder impedance identification result at time (k+1). k+1 and a k+1 They are the new voltage measurement data and current measurement data from time k to time (k+1). k+1 is the correction term. λ is the forgetting factor. The higher its value, the smaller the impact on the final identification result and the smaller the impact of the prior voltage and current measurement data. This patent takes λ as 0.9. By using multiple iterations of formula (5), the online identification of feeder impedance can be realized. Without injecting specific subharmonics, the non-intrusive recursive least squares algorithm with forgetting factor is used to realize feeder impedance identification, which ensures the stability of feeder impedance and provides high-quality power supply to the load.

[0084] In order to achieve power sharing, virtual impedance is introduced. The idea of ​​reconstructing line impedance is to introduce a virtual voltage drop through current feedforward. After the introduction of virtual impedance, the simplified electrical block diagram of the distributed generator set can be equivalent to Figure 3 . Virtual impedance Z vi With feeder impedance Z Li The sum is the final equivalent impedance Z ei .

[0085] Traditional droop control adjusts the output voltage amplitude and frequency of each inverter according to the active and reactive power output of each DG. The droop control equation can be expressed as shown in the following equation (6):

[0086] f=f n +m(P n -P)#(6)

[0087] U=U 0 -nQ

[0088] In formula (6), m is the active frequency droop characteristic coefficient, n is the reactive voltage droop characteristic coefficient, and f n is the rated frequency of the power grid, P n is the output active power of the micro power supply at rated frequency, U 0 is the voltage amplitude when the reactive power output by the micro power supply is 0, and P and Q are the measured values ​​of the active and reactive power output by the inverter power supply, respectively.

[0089] Adaptive virtual impedance is added to the traditional droop control strategy to compensate for the voltage drop caused by line differences and eliminate the impact of impedance differences on reactive power distribution. Directly adding virtual impedance values ​​requires known line impedance parameters. To avoid using line impedance values, the virtual impedance is adjusted by load power. The voltage drop generated on the virtual impedance compensates for the voltage drop caused by DG line impedance differences, achieving reasonable power distribution. When the load power factor is different, the required virtual impedance value is different, so the virtual impedance value should be adaptively adjusted as the load power factor changes.

[0090] The DG adjusts the virtual impedance value through the integral controller based on the difference between the received reactive power output reference value and the actual output reactive power. The virtual impedance is superimposed on the voltage reference value obtained by the droop control. The integrator can automatically adjust the virtual impedance according to the reactive power, without being affected by the inconsistency of the DG line impedance, to adapt to the change of the operating state and realize the reasonable distribution of reactive power. When the virtual impedance is adjusted to the point where the actual output reactive power of the DG is equal to the reactive power reference value, the virtual impedance value will no longer change. The virtual impedance value will only change when the load power factor changes.

[0091] In order to achieve power sharing, it is necessary to ensure that the introduced virtual impedance makes the final equivalent impedance of each DG consistent. e1 =Z e2 =Z e . Z vi The selection of is crucial to system performance. When the enable end of the virtual impedance controller receives a trigger signal, the local controller of the DG unit receives the virtual impedance information. In order to avoid unnecessary voltage loss, we take the average value of all feeder impedances as the equivalent impedance reference value, as shown in equation (7), and then use equation (8) to determine the optimal virtual impedance.

[0092]

[0093] Z vi =Z e -Z Li =(R e -R Li )+j(X e -X Li )#(8)

[0094] In the above formula, Z e is the globally consistent equivalent impedance after compensation, Z vi is the optimal value of virtual impedance.

[0095] It should be noted that although the embodiments of the present invention have been shown and described, for ordinary technicians in this field, without departing from the purpose of the invention, structural methods and embodiments similar to the technical solution are designed without creativity and should all fall within the scope of protection of the present invention.

Claims

1. A control method for an intelligent single-unit converter, characterized in that: The following steps are involved: Step S1, constructing a network line operation model and designing an intelligent online identification scheme for power grid parameters; Step S2, introducing edge computing technology and constructing a local computing optimization solution; Step S3, analyzing the grid support technology of the distributed grid-connected converter, and analyzing the support control scheme that can adapt to the seamless connection of multiple states according to the converter states under multiple working conditions, and combining with step S1 and step S2 to form an auxiliary algorithm for realizing the intelligence of the single converter; The specific scheme of step S1 is as follows: S1.1, summarize the parameter variation rules to form a grid state parameter relationship comparison data set; S1.2, construct a high-renewable energy penetration power grid line and network state parameter model based on the data set, introduce modern artificial intelligence technologies such as deep learning and reinforcement learning, build a digital twin system for smart grids, realize online identification of power grid states and parameters, and use the least squares method to update the parameters of the linear model online; S1.3, compare the actual and identified parameters and further modify the intelligent identification algorithm; The specific scheme of step S2 is as follows: S2.1, according to the online identification algorithm obtained in step S1, stripping off the local parameter operation and the global information identification operation set, and standardizing the local data and the global identification software interface according to the operation set; S2.2, introduce edge computing technology, use local converter controllers to perform operations on local data sets, and synchronize global information according to standard unified interfaces, reduce the computing burden of servers, and further revise intelligent identification algorithms; The control scheme in step S3 is an improved droop control method based on virtual impedance optimization, and the control scheme includes two components: virtual impedance adaptive adjustment and feeder impedance identification; Virtual impedance adaptive adjustment: used to correct the imbalance of feeder impedance between distributed power sources DG1 and DG2. The adaptive virtual impedance controller adjusts the feeder impedance Z according to the feeder impedance Z. L Real-time adjustment of virtual impedance Z v , so that the equivalent impedance of each DG unit is Z ei equal; Feeder impedance identification: When the measured output reactive power deviation between inverters exceeds the upper limit, the feeder impedance Z of each DG unit is estimated using the recursive least squares algorithm. L ; Without injecting specific subharmonics, a non-intrusive recursive least squares algorithm with a forgetting factor is used to identify the feeder impedance, ensuring the stability of the feeder impedance and providing high-quality power to the load. To achieve power sharing, virtual impedance is introduced. The idea of ​​reconstructing line impedance is to introduce a virtual voltage drop through current feedforward. After the virtual impedance is introduced, the virtual impedance Z vi With feeder impedance Z Li The sum is the final equivalent impedance Z ei ; Adaptive virtual impedance is added to the droop control strategy to compensate for the voltage drop caused by line differences and eliminate the impact of impedance differences on reactive power distribution. Directly adding virtual impedance values ​​requires known line impedance parameters. To avoid using line impedance values, the virtual impedance is adjusted by load power. The voltage drop generated on the virtual impedance compensates for the voltage drop caused by DG line impedance differences. When the load power factor is different, the required virtual impedance value is different, so the virtual impedance value is adaptively adjusted as the load power factor changes. DG adjusts the virtual impedance value through the integral controller based on the difference between the received reactive power output reference value and the actual output reactive power. The virtual impedance is superimposed on the voltage reference value obtained by the droop control. The virtual impedance is automatically adjusted according to the reactive power through the integrator and is not affected by the inconsistent impedance of the DG line. When the virtual impedance is adjusted to the point where the actual output reactive power of DG is equal to the reactive power reference value, the virtual impedance value no longer changes. When the load power factor changes, the virtual impedance value changes accordingly.

2. The control method of an intelligent single-unit converter according to claim 1, characterized in that: In said S1.1, the impedance and moment of inertia parameters of the power electronic converter cluster grid parameters are collected and analyzed, and the influence of the distribution and operation mode of the "source-load" power electronic interface converter on the grid state parameters is analyzed; The recursive least squares method in S1.2 is used to find the best function matching of the data by minimizing the sum of squares of errors; the unknown data is obtained by using the least squares method, and the sum of squares of errors between the obtained data and the actual data is made to be a predetermined value; Online observation saves all observations and uses the classical least squares method to solve; The specific scheme of step S3 is as follows: S3.1, the off-grid / on-grid status and power generation mode of multiple distributed power sources change over time, analyze the operating status characteristics of the converter, summarize the converter status under multiple working conditions, and build a converter grid-connected status and power generation mode identification scheme; S3.2, analyze the converter cluster distributed power grid support technology; S3.3, combined with step S1 and step S2, forms an auxiliary algorithm for realizing the intelligence of the single-cell converter.

3. The control method of an intelligent single-unit converter according to claim 2, characterized in that: The virtual impedance adaptive adjustment specifically includes the following operations: First, the local measurement unit collects the PCC voltage and current data, and the impedance identification module receives the data and uses the recursive least squares method to calculate and output the feeder impedance identification result; Secondly, the feed line impedance Z L It is sent to the central controller to calculate the optimal virtual impedance value. The central controller of each distributed power source introduces virtual impedance to form an improved droop control, and adopts the improved droop control to ensure active power sharing.

4. The control method of an intelligent single-unit converter according to claim 3 is characterized in that: The feeder impedance identification scheme, as the core part of the reactive power sharing strategy, is responsible for the identification of feeder impedance and includes two working modes: identification mode and standby mode. The scheme will enter the identification mode only when a pulse trigger signal is received; otherwise, it will always operate in standby mode, waiting for a trigger signal with a higher control level.

5. The control method of an intelligent single-unit converter according to claim 4, characterized in that: The voltage and current measurement unit first obtains the inverter output voltage V o The value of m groups P CC The voltage value V pcc and current value I pcc , after entering the recognition mode, the working principle is as follows: The relationship when the inverter is running stably is shown in equation (1): In formula (1), V pcc Indicates P CC Voltage vector; I pcc Represents P CC Current vector; V o Represents the inverter output voltage vector; Z L The feeder impedance Z L = R L + jX L ; Formula (1) is simplified to the matrix form Y = AX; In order to describe the deviation between the identified value and the actual impedance value of the feeder, the cost function is defined as formula (3) as follows: In order to minimize the value of the cost function, the partial derivative of J must be calculated, and ∂J / ∂X = 0 is set to obtain the optimal feeder impedance identification value X, as shown in the following formula (4): In order to realize online recognition and prevent the main line saturation of the recognition result, a recursive least squares algorithm with forgetting factor is proposed, as shown in the following formula (5): In formula (5), X k is the feeder impedance identification result at the kth moment; X k+1 is the feeder impedance identification result at time (k+1), y k+1 and a k+1 are the new voltage measurement data and current measurement data from time k to time (k+1); k+1 is the correction term; λ is the forgetting factor. The higher its value, the smaller the impact on the final recognition result and the smaller the impact of the prior voltage and current measurement data. λ is taken as 0.9; By using formula (5) for multiple iterations, the online identification of feeder impedance can be realized. Droop control is to adjust the output voltage amplitude and frequency of each inverter according to the active and reactive power output of each DG. The droop control equation is expressed as shown in the following formula (6): In formula (6), t is the active frequency droop characteristic coefficient, n is the reactive voltage droop characteristic coefficient, and f n is the rated frequency of the power grid, P n is the output active power of the micro power source at the rated frequency, U0 is the voltage amplitude when the output reactive power of the micro power source is 0, and P and Q are the measured values ​​of the output active and reactive power of the inverter power source respectively; To achieve power sharing, it is necessary to ensure that the introduced virtual impedance makes the final equivalent impedance of each DG consistent: that is, Z e1 =Z e2 =Z e ; When the enable end of the virtual impedance controller receives the trigger signal, the local controller of the DG unit receives the virtual impedance information; the average value of all feeder impedances is used as the equivalent impedance reference value, as shown in formula (7), and then the optimal virtual impedance is determined using formula (8): In the above formulas (7) and (8), Z e is the globally consistent equivalent impedance after compensation, Z vi is the optimal value of virtual impedance.

Images