Self-adaptive stability control method for grid-forming type grid-connected inverter considering strong power grid characteristics

Through improved DBN-ELM model and genetic algorithm optimization technology, the problem of grid-connected inverter instability caused by grid impedance changes is solved, and the system's adaptive stability control is realized, ensuring that the grid inverter after grid impedance changes Run stably.

CN119093463APending Publication Date: 2024-12-06NORTH CHINA ELECTRIC POWER UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310654093.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-06-05
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

After the new energy equipment is connected, the change in the grid impedance results in the lack of adaptive adjustment capabilities of the control parameters of the grid-connected inverter, which poses a risk of instability.

Method used

The improved DBN-ELM closed-loop model is used to predict the poles, and the genetic algorithm is used to optimize the number of neuron nodes in the hidden layer, and the control parameters are adjusted through an adaptive stability control method to ensure that the system remains stable after the grid impedance changes.

Benefits of technology

By adaptively adjusting the control parameters, the unstable poles are pulled back to the reference poles, so that the system can operate adaptively and stably and avoid the risk of instability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119093463A_ABST
    Figure CN119093463A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of grid-connected inverter stability control, and discloses a self-adaptive stability control method for a grid-forming type grid-connected inverter considering strong power grid characteristics, and the method comprises the steps: building a complex vector impedance circuit model of the grid-forming type grid-connected inverter; establishing a closed-loop pole mapping model, training a mapping relationship between the control parameters and the key poles by using a deep architecture, and predicting through the trained closed-loop pole mapping model to obtain the corresponding key poles; and determining a control parameter of the grid-forming grid-connected inverter when the key pole is closest to the reference pole, and adaptively adjusting the control parameter to ensure that the system tracks the reference pole when the impedance of the power grid changes so as to realize adaptive stable control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of grid-connected inverter stability control, and proposes a grid-connected inverter adaptive stability control method taking into account strong power grid characteristics. Background Art

[0002] As the proportion of new energy equipment such as wind power and photovoltaic power increases, the new power system has a trend of weakening system strength and reducing inertia. The grid impedance presents a wide range of time-varying characteristics. The control parameters of the grid-connected inverter lack the ability to adaptively adjust, and there is a risk of instability. The inventor found that it is necessary to propose an adaptive adjustment method that can adapt to changes in grid impedance. Summary of the invention

[0003] In order to solve the above problems, the present invention provides an adaptive stability control method for a grid-connected inverter taking into account the characteristics of a strong power grid. The scheme adopts an improved DBN-ELM closed-loop model to predict the poles, and uses a genetic algorithm to optimize the number of neuron nodes in each hidden layer. By adjusting multiple control parameters through an adaptive stability control method, the unstable poles can be pulled back to the reference poles after the grid impedance changes, so that the system can achieve adaptive stable operation.

[0004] The present invention provides a method for adaptive stabilization control of a grid-connected inverter taking into account the characteristics of a strong power grid, comprising the following steps:

[0005] Step S10, establishing a complex vector impedance circuit model of a grid-connected inverter;

[0006] Step S20, based on the impedance circuit model, using an improved deep belief network-extreme learning machine algorithm (DBN-ELM) model to predict the poles;

[0007] Step S30: adaptively adjusting the control parameters of the grid-connected inverter based on the pole prediction result.

[0008] Furthermore, the step S10 includes:

[0009] Step S101, establishing a complex vector equivalent output impedance of a grid-connected inverter and a Pf droop control link without considering an active power-frequency (Pf) droop control link;

[0010] Step S102, establishing a complex vector equivalent output impedance of a grid-connected inverter taking into account power droop control.

[0011] Furthermore, the step S20 includes:

[0012] Step S201, using a genetic algorithm to optimize the number of nodes in each hidden layer of the DBN-ELM model, thereby establishing an improved DBN-ELM model;

[0013] Step S202, selecting the input and output of the mapping model, and establishing a closed-loop pole mapping model for determining stability.

[0014] Furthermore, the step S30 includes:

[0015] Step S301, determining a parameter search range according to given parameters to generate a sample data set and perform offline training;

[0016] Step S302: input the measured grid impedance estimation value into the trained closed-loop pole mapping model to achieve online positioning of the control parameters.

[0017] The present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the above-mentioned method for adaptive stability control of a grid-connected inverter taking into account the characteristics of a strong power grid.

[0018] Compared with the existing technology, this technology has the following beneficial effects:

[0019] In order to solve the above problems, the present invention provides an adaptive stability control method for a grid-connected inverter taking into account the characteristics of a strong power grid. The scheme adopts an improved DBN-ELM closed-loop model to predict the poles, and uses a genetic algorithm to optimize the number of neuron nodes in each hidden layer. By adjusting multiple control parameters through an adaptive stability control method, the unstable poles can be pulled back to the reference poles after the grid impedance changes, so that the system can achieve adaptive stable operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 The general flow chart for implementing the method of the present invention is as follows;

[0022] Figure 2 The invention provides a grid-connected inverter control structure;

[0023] Figure 3 It is an equivalent block diagram of active power control closed loop of the present invention;

[0024] Figure 4The invention provides a derivation process of the equivalent virtual impedance of the control loop of the grid-connected inverter;

[0025] Figure 5 The complex vector impedance circuit model of the grid-connected inverter is constructed when the power droop control link is not considered in the present invention;

[0026] Figure 6 A complex vector impedance circuit model of a grid-connected inverter is constructed when the power droop control link is considered in the present invention;

[0027] Figure 7 It is the DBN-ELM model structure;

[0028] Figure 8 It is the flow chart of the genetic algorithm of the present invention;

[0029] Fig. 9 This is a flow chart of the control parameter search algorithm of the present invention;

[0030] Fig.10 The results comparison chart of the two control methods;

[0031] Fig.11 For L g Simulation result diagram without using adaptive stability control method when decreasing;

[0032] Figure 12(a) and 12(b) L g The control parameter change waveform diagram and the inverter output voltage d-axis component waveform diagram using the adaptive stability control method when decreasing. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] See also Figure 1 The present invention provides a method for adaptive stabilization control of a grid-connected inverter taking into account the characteristics of a strong power grid, comprising the following steps:

[0035] Step S10, establishing a complex vector impedance circuit model of a grid-connected inverter;

[0036] Step S20, predicting the poles using an improved deep belief network-extreme learning machine algorithm (DBN-ELM) based on the impedance circuit model;

[0037] Step S30: adaptively adjusting the control parameters of the grid-connected inverter based on the pole prediction result.

[0038] The step S10 of establishing a complex vector impedance circuit model of a grid-connected inverter includes:

[0039] Step S101, establishing a complex vector equivalent output impedance of a grid-connected inverter and a Pf droop control link without considering an active power-frequency (Pf) droop control link.

[0040] The grid-connected inverter with active power-frequency (Pf) droop control is taken as the research object, and the expression of its droop control regulator is shown in formula (1).

[0041]

[0042] Where: m is the droop gain; 1 / (1+s / ω f ) is a low-pass filter (LPF), which can attenuate high-frequency noise; ω f is the cutoff frequency of the low-pass filter.

[0043] When the power droop control link is not considered, the impedance matrix of the grid-connected inverter is symmetrical, and its output voltage can be expressed as

[0044]

[0045] In the formula, Z loopi is the equivalent virtual impedance of the current loop control; Z loopvc It is the sum of the equivalent virtual impedance of the voltage loop control and filter capacitors.

[0046] Therefore, the output impedance of the grid-connected inverter can be expressed as

[0047]

[0048] Where:

[0049] The expression of the equivalent virtual impedance of Pf droop control is shown in formula (4).

[0050]

[0051]

[0052] Where:

[0053] Step S102, establishing a complex vector equivalent output impedance of a grid-connected inverter taking into account power droop control.

[0054] Since the power droop control makes the output impedance of the grid-connected inverter asymmetric, the output impedance of the grid-connected inverter considering the power droop control can be expressed as

[0055]

[0056] Where: Z o+ =Z s± +Z PF+ , Z o- =Z PF- .

[0057] The step S20 uses an improved DBN-ELM model to predict the poles based on the impedance circuit model, including:

[0058] Step S201, using a genetic algorithm to optimize the number of nodes in each hidden layer of the DBN-ELM model, thereby establishing an improved DBN-ELM model.

[0059] The DBN-ELM model consists of two parts: DBN responsible for feature extraction and extreme learning machine as the regression layer. The genetic algorithm is used to optimize the number of nodes in each hidden layer of the DBN-ELM model, thereby establishing a complete DBN-ELM model and further enhancing the generalization of the model.

[0060] Step S202, selecting the input and output of the mapping model, and establishing a closed-loop pole mapping model for determining stability.

[0061] The control parameter u of the grid-connected inverter is transformed into ctrl and grid impedance L g As the input of the mapping model, the key poles of the system are the output of the mapping model, and the closed-loop pole mapping is implemented by the improved deep belief network-extreme learning machine algorithm (DBN-ELM). ctrl and grid impedance L g ) can obtain the trained improved DBN-ELM model, that is, the closed-loop pole mapping model for judging stability, which is recorded as:

[0062] y out =σ i =H(u in )(6)

[0063] In the formula, σ is the key pole in the closed-loop pole of the system. Since there are two key poles in the system and the sample data of the DBN-ELM model should be expressed as real numbers, the key pole is further divided into real and imaginary parts, so σ = [α1, f1, α2, f2] T , αi and f i They are the closed-loop poles λ i The real part of and the corresponding oscillation frequency; H(·) is the relationship function between the key poles of the system and the system parameters.

[0064] In order to obtain various combinations of input parameters with relatively low data volume to a large extent, the input u of the training set is in The expression is:

[0065]

[0066] Where, L g 、u ctrl Each parameter sample in is generated by a random number that obeys a uniform distribution. The range of random generation of parameter samples can be determined according to different types of grid-connected inverters and the actual situation of the power grid.

[0067] Based on the input u of the training set in , calculate the corresponding closed-loop poles, and the sample data set output is:

[0068]

[0069] In the formula, y out (k) represents u in (k) The real and imaginary parts of the corresponding critical poles.

[0070] The step S30 of adaptively adjusting the control parameters of the grid-connected inverter based on the pole prediction result includes:

[0071] Step S301, determining a parameter search range according to given parameters to generate a sample data set and perform offline training.

[0072] Based on the given grid impedance value and system control parameters, the search range of the control parameters is determined, that is, (L, U represent the lower limit and upper limit of each parameter), generating N control parameter combinations.

[0073] The input of the sample data set is L g and u ctrl N control parameter combinations u in The corresponding key poles calculated are used as the output of the sample data set, and the data is preprocessed. The improved DBN-ELM model is used to learn and train the sample data set to obtain a closed-loop pole mapping model.

[0074] Step S302: input the measured grid impedance estimation value into the trained closed-loop pole mapping model to achieve online positioning of the control parameters.

[0075] Detected Lg After the change, get L g The measured value is used to determine whether the real part of the corresponding key pole is greater than 0. If the real part of the key pole is greater than zero, the control parameter search method is started; otherwise, the real part keeps the control parameter unchanged.

[0076] Based on the given grid impedance value and system control parameters, the search range of the control parameters is set to generate N spl sample vectors of shape N vol =n×(N spl ) n A combination of control parameters.

[0077] The control parameters and the real-time measured grid impedance are used as the input of the improved ELM-DBN model, and the corresponding output pole α is predicted by the trained closed-loop pole mapping model. i =[α(1),…,α(N vol )], from which the control parameters corresponding to the minimum MSE can be identified.

[0078] Using a concept similar to binary division to update the upper and lower parameter boundaries of the step, the expression of the updated control parameter is shown in formula (9).

[0079]

[0080] The above process is repeated. When MSE < η (η is a predefined tolerance), it means that a set of control parameters that meet the pole tracking objective is found, and the iteration ends.

[0081] Specifically, for ease of understanding, the solution described in this embodiment is described in detail below with reference to the accompanying drawings:

[0082] like Figure 1 As shown, this embodiment provides an adaptive stability control method for a grid-connected inverter taking into account the characteristics of a strong power grid, and its main technical concept is: establishing a complex vector impedance circuit model of the grid-connected inverter and using an improved DBN-ELM model to predict the poles → adaptively adjusting the control parameters of the grid-connected inverter.

[0083] Figure 2 The control structure of the grid-connected inverter of the present invention is shown in FIG. i is the port voltage of the grid-connected inverter; v o 、i o are the output voltage and current of the grid-connected inverter; i l and i c are the currents on the filter inductor and filter capacitor of the grid-connected inverter; L f and C f are filter inductance and filter capacitance respectively; rf is the equivalent resistance of the LC filter; i l.ref is the current given signal; u g is infinite power supply voltage; is the voltage amplitude given signal; L g and r g = is the equivalent inductance and equivalent resistance of the grid. Since the resistive component in the grid impedance is more beneficial to the stability of the grid-connected system, the present invention only considers the grid inductive reactance to analyze the most unfavorable situation. The grid-connected inverter generates a phase set value (θ) through active power-frequency droop control, and the phase is the synchronization signal.

[0084] Figure 3 This is the closed-loop equivalent block diagram of the active power control of the present invention. In the figure, P* is the given value of the output power; ω* is the given angular velocity; G PF It is a droop control regulator.

[0085] Figure 4 This is the derivation process of the equivalent virtual impedance of the control loop of the grid-connected inverter. The grid-connected inverter is essentially a controlled voltage source. The grid-connected inverter adopts the control method of the dual PI regulator of the current inner loop and the voltage outer loop. The function of the current inner loop is to track the grid current command current to achieve high power factor grid connection of the inverter; the function of the voltage outer loop is to stabilize the DC side voltage. The voltage control in the grid-connected inverter provides a given current for the current control. In the figure, L f is the filter inductor of the grid-connected inverter; C f is the grid-connected inverter filter capacitor; ω r is the actual angular frequency (rad / s); G PIv is a voltage controlled PI regulator; Z PIv is the equivalent impedance of the voltage-controlled PI regulator; G PIi is the current control PI regulator transfer function; Z PIi is the equivalent impedance of the current controlled PI regulator; Z CDv is the voltage controlled negative virtual impedance; Z CDi Negative virtual impedance for current control.

[0086] Figure 5 The complex vector impedance circuit model of the grid-connected inverter is constructed when the power droop control link is not considered in the present invention. is the voltage given value; G V is the voltage control loop gain; Z loopi is the equivalent virtual impedance of the current loop control; Z loopvc It is the sum of the equivalent virtual impedance of the voltage loop control and filter capacitors.

[0087] Figure 6The impedance circuit model of the grid-connected inverter is constructed for the present invention when considering power droop control. s± is the equivalent virtual impedance of the grid-connected inverter without considering the power droop control link; Z PF± is the equivalent virtual impedance of Pf droop control.

[0088] Figure 7 This is the DBN-ELM model structure. DBN-ELM is composed of multiple layers of restricted Boltzmann machines (RBM) and one fully connected layer. In the figure, input x = {x1, x2, ..., x m-1 , x m} is an m-dimensional training sample, y = {y1, y2, ..., y n-1 ,y n} is the n-dimensional network output. The model consists of two parts: DBN responsible for feature extraction and extreme learning machine as regression layer. It combines the superior performance of DBN in feature extraction with the characteristics of fast learning speed and good generalization performance of ELM, thereby improving the prediction performance.

[0089] Figure 8 The figure is a flow chart of the genetic algorithm. First, the basic parameters of the genetic algorithm, such as the maximum genetic generation, population size, individual length, generation gap, crossover probability and mutation probability, are set to initialize the population; then, the number of nodes in each hidden layer of DBN-ELM corresponds to the individuals in the population, and real numbers are used for encoding. The selection, crossover and mutation are performed by roulette, multi-point crossover and random selection of mutation points, respectively, and the individual fitness is calculated; finally, the optimal number of hidden layer nodes is obtained when the maximum number of middle fingers is reached, and the loop terminates.

[0090] Fig. 9 The control parameter search algorithm is mainly divided into two parts: offline training and online positioning. Offline training generates a control parameter combination u based on the given grid impedance value and system control parameters. in and the corresponding key poles are used as the input and output of the sample data set, and the data is preprocessed. The improved DBN-ELM model is used to learn and train the sample data set to obtain a closed-loop pole mapping model. When applied online, first determine L g After the change, the real part of the corresponding key pole is greater than 0. If the real part of the key pole is greater than zero, the control parameter search method is started, and the control parameters and the real-time measured grid impedance are used as the input of the improved ELM-DBN model. The corresponding output poles are predicted by the trained closed-loop pole mapping model, and the control parameters corresponding to the minimum MSE are identified; otherwise, the real part keeps the control parameters unchanged.

[0091] Fig.10The results of the two control methods are compared. g When the value is reduced from 0.4pu to 0.2pu, it can be seen from the calculation results of the system poles before and after the adaptive adjustment of the control parameters is adopted that when the adaptive adjustment of the control parameters is not adopted, there is a pair of key poles with positive real parts, and the system is unstable; when the adaptive stability control method proposed in the present invention is adopted, the key poles obtained are basically consistent with the key poles of the given system, the real parts of the key poles are all negative, and the system is stable. It can be seen from the comparison of the results obtained without the adaptive adjustment of the control parameters and with the adaptive stability control method proposed in the present invention that after the adaptive adjustment method proposed in the present invention is adopted, the system can still maintain good stability after the grid impedance changes.

[0092] Fig.11 For L g As can be seen from the figure: as the grid impedance decreases, the d-axis component of the inverter output voltage oscillates at a low frequency of 11.2Hz, and the oscillation amplitude gradually increases, causing the system to become unstable. The simulation results are consistent with the closed-loop stability analysis results.

[0093] Figure 12(a) and 12(b) L g The waveform of the control parameter change and the d-axis component of the inverter output voltage using the adaptive stability control method when the impedance decreases. As can be seen from Figure 12(a), the grid impedance decreases at t=1s, and the control parameter search begins. The gradual decrease in MSE indicates that the closed-loop pole is moving towards the given closed-loop pole α*. After 0.06s, the MSE reaches the threshold. At 1.06s, the control parameter search ends and the control parameters are adjusted. As can be seen from Figure 12(b), after the grid impedance decreases, the d-axis component of the inverter output voltage tends to stabilize after a short oscillation, which is consistent with the theoretical analysis results. The simulation results show that by adjusting the searched control parameters, the unstable pole can be pulled back to the reference pole, so that the system can be restored to stability, which proves the effectiveness of the adaptive stability control method.

[0094] The embodiment of the present invention also provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the above-mentioned grid-connected inverter adaptive stability control method is completed.

[0095] The above-mentioned embodiment provides a method for adaptive stabilization control of a grid-connected inverter taking into account the strong grid characteristics, which can be realized and has broad application prospects.

[0096] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An adaptive stability control method for grid-connected inverters considering the characteristics of strong power grids. Features To ensure that the system achieves adaptive stable control when the grid impedance changes, the following steps are included: Step S10, establishing a complex vector impedance circuit model of a grid-connected inverter; Step S20, based on the impedance circuit model, using an improved deep belief network-extreme learning machine algorithm (DBN-ELM) model to predict the poles; Step S30: adaptively adjusting the control parameters of the grid-connected inverter based on the pole prediction result.

2. The method according to claim 1, It is characterized in that The step S10 comprises: Step S101, establishing a complex vector equivalent output impedance of a grid-connected inverter and a Pf droop control link without considering an active power-frequency (Pf) droop control link; Step S102, establishing a complex vector equivalent output impedance of a grid-connected inverter taking into account power droop control.

3. The method according to claim 1, It is characterized in that The step S20 includes: Step S201, using a genetic algorithm to optimize the number of nodes in each hidden layer of the DBN-ELM model, thereby establishing an improved DBN-ELM model; Step S202, selecting the input and output of the mapping model, and establishing a closed-loop pole mapping model for determining stability.

4. The method according to claim 1, It is characterized in that The step S30 includes: Step S301, determining a parameter search range according to given parameters to generate a sample data set and perform offline training; Step S302: input the measured grid impedance estimation value into the trained closed-loop pole mapping model to achieve online positioning of the control parameters.

5. A computer-readable storage medium having a program stored thereon, It is characterized in that When the program is executed by a processor, it implements an adaptive stability control method for a grid-connected inverter taking into account strong grid characteristics as described in any one of claims 1 to 4.