Secondary voltage-frequency controller parameter setting method based on system identification

Abstract

The invention provides a secondary voltage-frequency controller parameter setting method based on system identification, and relates to the technical field of micro-grid control, and the method comprises the steps: carrying out the sampling of data used for system identification according to a preset sampling rule, and obtaining an original data sequence; according to the original data sequence, determining an identification model structure and order adaptive to the black box characteristics of the micro-grid, and outputting a model framework; based on the model framework, performing iterative solution by using an LM algorithm, and outputting an initial identification model for describing the micro-grid system; the initial identification model is verified, and a final identification model meeting the fitting degree requirement is obtained; and setting the parameters of the secondary voltage-frequency controller based on the final identification model. According to the scheme, the control of secondary voltage and frequency regulation can be realized without completely depending on the internal mechanism information of the micro-grid.

Description

Technical Field

[0001] This invention relates to the field of microgrid control technology, and in particular to a method for tuning the parameters of a secondary voltage-frequency controller based on system identification. Background Technology

[0002] With rapid socio-economic development and a rapidly improving standard of living, human demand for energy is increasing daily. However, due to the rapid depletion of fossil fuels and severe damage to the ecological environment, traditional energy sources are increasingly unable to meet the needs of human production and daily life. Faced with a severe energy crisis, there is an urgent need to vigorously develop new energy sources such as wind, solar, and tidal energy. Among these, new energy power generation technologies provide an important way to solve energy problems. Currently, wind power, photovoltaic power generation, and other new energy power generation industries have achieved industrial operation in major countries around the world. As an important application of distributed generation technology, microgrid technology has solved the access problem for various distributed power sources and has developed rapidly in recent years. Compared with traditional large power grids, microgrids are small power grids composed of multiple power sources, loads, and corresponding control and protection devices. They can be connected to the large power grid or operated in isolation, and can flexibly switch between these two operating modes. This can effectively improve the power system's ability to accommodate distributed generation systems and the utilization efficiency of distributed energy, making it an important component of the future smart grid. Undoubtedly, current microgrid technology cannot perfectly solve all the problems brought about by distributed generation, but the proposal of microgrids still points to a path for the large-scale and efficient utilization of clean energy and has broad research prospects.

[0003] In peer-to-peer microgrids built using virtual synchronous generators (VSGs) or droop-controlled inverters, this system has become a common application due to its ease of expansion and high reliability. However, both droop control and virtual synchronous generator control are voltage-frequency differential control methods, meaning the voltage and frequency within the system change with load variations. Therefore, in peer-to-peer microgrid systems, it is necessary to employ secondary voltage and frequency regulation control methods to maintain the system's voltage and frequency at normal levels, thereby improving the power supply quality and reliability of the microgrid system.

[0004] Current methods for secondary voltage and frequency regulation control in microgrids primarily rely on inverter mechanism models to control and regulate the microgrid. However, the topology of microgrids is becoming increasingly complex, making it difficult to accurately obtain detailed internal information, or the obtained information is often significantly simplified. These characteristics mean that traditional controller parameter tuning methods used in motor and inverter control are increasingly unsuitable for the parameter design of secondary frequency and voltage regulation controllers in microgrid systems. Summary of the Invention

[0005] In view of this, and to address the above shortcomings, it is necessary to propose a secondary voltage-frequency controller parameter tuning method based on system identification, so as to achieve secondary voltage and frequency regulation control without completely relying on the internal mechanism information of the microgrid.

[0006] In a first aspect, the present invention provides a method for tuning the parameters of a secondary voltage-frequency controller based on system identification, comprising:

[0007] The data used for system identification is sampled according to the preset sampling rules to obtain the original data sequence;

[0008] Based on the original data sequence, determine the identification model structure and order adapted to the black box characteristics of the microgrid, and output the model framework;

[0009] Based on the aforementioned model framework, the LM algorithm is used for iterative solution to output an initial identification model for describing the microgrid system.

[0010] The initial identification model is validated to obtain the final identification model that meets the fitting requirements;

[0011] Based on the final identification model, the parameters of the secondary voltage-frequency controller are tuned.

[0012] Preferably, in the step of sampling the data used for system identification according to a preset sampling rule, the sampling frequency satisfies ;in, Used to characterize the sampling frequency, Used to characterize signal frequency.

[0013] Preferably, after obtaining the original data sequence, the process further includes:

[0014] For each type of original data sequence, a steady-state data segment is selected to calculate the mean, and the difference between the original data sequence and the mean is calculated to obtain the corrected data after baseline offset removal;

[0015] The corrected data is then subjected to noise filtering to obtain a data sequence for model construction.

[0016] Preferably, in the step of determining the identification model structure and order adapted to the black-box characteristics of the microgrid based on the original data sequence, the model structure is determined to be an OE model.

[0017] Preferably, determining the order of the identification model adapted to the black-box characteristics of the microgrid based on the original data sequence includes:

[0018] Iterative testing was performed on combinations of denominator and numerator orders;

[0019] Determine the degree of fit between the step response of each model order corresponding to each combination and the measured step response of the microgrid;

[0020] From various combinations of model orders that meet the degree of fit, the combination of model orders with the lowest order is selected as the identification model order that adapts to the black-box characteristics of microgrids.

[0021] Preferably, the iterative solution using the LM algorithm includes:

[0022] Based on the structure and order of the identification model, determine the transfer function and the parameter vector to be identified;

[0023] Set the initial values ​​for the recognition model;

[0024] Calculate the Jacobian matrix of the frequency domain response relative to the variable, and construct the incremental normal equation of the variable based on the prediction error between the model output and the measured output at the k-th iteration;

[0025] Solve the incremental normal equation and calculate the prediction error between the model output and the measurement output at the (k+1)th iteration;

[0026] Based on the prediction error at the (k+1)th iteration, perform iterative updates and convergence judgment, and output the initial identification model after convergence.

[0027] Preferably, the verification of the initial identification model includes:

[0028] The initial identification model and the microgrid experimental device are provided with the same step input, and their respective step response outputs are output.

[0029] Calculate the best fit based on the respective step response outputs;

[0030] Determine whether the optimal fit meets the preset requirements;

[0031] If the optimal fit meets the preset requirements, the current initial identification model is deemed to have passed the verification; if the optimal fit does not meet the preset requirements, the order of the identification model is re-determined and identification is performed again until the optimal fit requirements are met.

[0032] Preferably, the step of calculating the best fit based on the respective step response output includes:

[0033] The best fit is calculated according to the following formula:

[0034] In the formula, Used to characterize the best fit. Used to characterize the step response output by the initial identification model. Used to characterize the measured output of the microgrid experimental setup Used to characterize the average value of the measured output.

[0035] In a second aspect, the present invention provides a computing device including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it performs any of the methods described in the first aspect.

[0036] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform any of the methods described in the first aspect.

[0037] As can be seen from the above technical solution, the secondary voltage-frequency controller parameter tuning method based on system identification provided in this embodiment of the invention first samples the data used for system identification according to a preset sampling rule to obtain the original data sequence. Then, based on the original data sequence, the identification model structure and order adapted to the microgrid core phase characteristics are determined, and the model framework is output. Further, based on the model framework, the LM algorithm is used for iterative solution to output an initial identification model for describing the microgrid system. Then, the initial identification model is verified to obtain a final identification model that meets the fitting requirements. Finally, based on the final identification model, the secondary voltage-frequency controller parameters can be tuned. Therefore, this solution adopts an identification modeling method, which can perform identification modeling by measuring input and output port data, no longer relying entirely on the internal mechanism information of the microgrid, avoiding the complex process brought about by traditional mechanism modeling, and thus realizing the tuning of the secondary voltage-frequency controller parameters. This ensures that this solution is applicable to the parameter design of secondary frequency and voltage regulation controllers in microgrid systems. Attached Figure Description

[0038] Figure 1 This is a flowchart of a method for tuning secondary voltage-frequency controller parameters based on system identification, provided in an embodiment of the present invention.

[0039] Figure 2 This is a topology diagram of a microgrid experimental platform.

[0040] Figure 3 This is an experimental waveform for optimizing a microgrid frequency control system under a specific condition.

[0041] Figure 4 Optimize experimental waveforms for a microgrid voltage control system under another condition. Detailed Implementation

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] In peer-to-peer microgrids built using virtual synchronous generators (VSGs) or droop-controlled inverters, this system has become a common application due to its ease of expansion and high reliability. However, both droop control and virtual synchronous generator control are voltage-frequency differential control, meaning the voltage and frequency within the system change with load variations. Therefore, in peer-to-peer microgrid systems, it is necessary to employ secondary voltage and frequency regulation control methods to maintain the system's voltage and frequency at normal levels, thereby improving the power supply quality and reliability of the microgrid system. Current secondary voltage and frequency regulation control methods in microgrids are mainly based on inverter mechanism models to control and regulate the microgrid. However, the topology of microgrids is becoming increasingly complex, making it difficult to accurately obtain detailed internal information, or the information obtained is often significantly simplified. These characteristics determine that traditional controller parameter tuning methods used in motor and inverter control are increasingly difficult to apply to the parameter design of secondary frequency and voltage regulation controllers in microgrid systems. Therefore, researching more flexible and effective secondary voltage and frequency regulation control methods that do not entirely rely on the internal mechanisms of microgrids is of great significance. This will solve the parameter tuning problem of secondary voltage and frequency regulation controllers encountered in the large-scale application of microgrid systems, and achieve adaptive adjustment of the secondary voltage and frequency regulation controller parameters according to changes in the microgrid system topology and operating state. Specifically, such as... Figure 1 As shown, the present invention provides a method for tuning secondary voltage-frequency controller parameters based on system identification, which may include the following steps:

[0044] Step 101: Sample the data used for system identification according to the preset sampling rules to obtain the original data sequence;

[0045] Step 102: Determine the identification model structure and order adapted to the black box characteristics of the microgrid based on the original data sequence, and output the model framework;

[0046] Step 103: Based on the aforementioned model framework, use the LM algorithm for iterative solution to output an initial identification model for describing the microgrid system;

[0047] Step 104: Verify the initial identification model to obtain the final identification model that meets the fitting requirements;

[0048] Step 105: Based on the final identification model, tune the parameters of the secondary voltage-frequency controller.

[0049] In this embodiment, the data used for system identification is first sampled according to a preset sampling rule to obtain the original data sequence. Then, based on the original data sequence, the identification model structure and order adapted to the microgrid's core phase characteristics are determined, and the model framework is output. Further, based on the model framework, the LM algorithm is used for iterative solving to output an initial identification model for describing the microgrid system. The initial identification model is then verified to obtain a final identification model that meets the fitting requirements. Finally, based on the final identification model, the parameters of the secondary voltage-frequency controller can be tuned. Therefore, this scheme adopts an identification modeling method that can perform identification modeling by measuring input and output port data, no longer relying entirely on the internal mechanism information of the microgrid, avoiding the complex process of traditional mechanism modeling, and thus enabling the tuning of the secondary voltage-frequency controller parameters. This ensures that this scheme is applicable to the parameter design of secondary frequency and voltage regulation controllers in microgrid systems.

[0050] The following is a detailed explanation of each step.

[0051] Step 101: Sample the data used for system identification according to the preset sampling rules to obtain the original data sequence.

[0052] In this step, the sampled data is the most original and crucial information used for system identification; therefore, the quality of the sampled data directly affects the identification result. For data acquisition during a step response process, the system needs to be in a steady state. The choice of sampling frequency directly affects the identification accuracy; the selection should follow the sampling theorem, i.e., the sampling frequency satisfies... ;in, Used to characterize the sampling frequency, It is used to characterize the signal frequency. However, an excessively high sampling frequency can lead to excessive data analysis and computation. Therefore, a typical sampling frequency can be chosen to be 10 times the signal frequency.

[0053] In addition, this plan should also build an experimental platform. The operating mode of the experimental platform is set as follows: disconnect the smart circuit breaker at the point of common coupling (PCC) to disconnect the microgrid from the main grid and enter islanded mode; send the initial active power reference value (e.g., 20kW) and reactive power reference value (e.g., 10kVar) to the energy storage converter through the microgrid central controller MGCC, maintain the steady-state operation of the system for 30 minutes, and ensure the stable operation of the equipment.

[0054] Specifically, when collecting data, two types of step experiments can be considered to simulate system power disturbance scenarios, namely: (1) Active power reference step experiment: keeping the reactive power reference value unchanged, the MGCC steps the active power reference value from 20kW to 50kW in the 5th second, and the collection time is from the 3rd second to the 13th second to cover the steady state segment of 2 seconds before the step and the dynamic segment of 8 seconds after the step; (2) Reactive power reference step experiment: keeping the active power reference value unchanged, the MGCC steps the reactive power reference value from 10kVar to 40kVar in the 5th second, and the collection time is the same as above. Specifically, the input data can be the active power reference value and reactive power reference value issued by the MGCC, and the output data can be the PCC point frequency, PCC point voltage amplitude, active power and reactive power at the energy storage converter port. The sampling frequency is set at 10 times the highest frequency of the signal. For example, if the signal frequency is 50Hz, the sampling frequency is set to 500Hz to avoid data aliasing. In this way, the original data sequence can be obtained by sampling, such as the frequency data sequence [f1, f2, ..., f 5000 Voltage amplitude sequence [U1, U2, ..., U 5000 ].

[0055] Furthermore, after obtaining the original data sequence, preprocessing is considered to eliminate baseline offset and noise interference. This can be achieved as follows:

[0056] For each type of original data sequence, a steady-state data segment is selected to calculate the mean, and the difference between the original data sequence and the mean is calculated to obtain the corrected data after baseline offset removal;

[0057] The corrected data is then subjected to noise filtering to obtain a data sequence for model construction.

[0058] In this embodiment, for each type of data collected, the mean value of the steady-state segment before the step jump is calculated, and the mean value is subtracted from the original data to obtain the corrected data. Furthermore, a first-order low-pass filter is used to filter out high-frequency measurement noise, and a five-point moving average method is used to filter out switching ripple, ultimately obtaining the preprocessed data sequence.

[0059] Step 102: Determine the identification model structure and order adapted to the black box characteristics of the microgrid based on the original data sequence, and output the model framework.

[0060] When identifying models, they can be mainly divided into two categories: equation error models and output error models. Equation error models are mainly represented by exogenous input autoregressive (ARX) models and exogenous input autoregressive moving average (ARMAX) models, while output error models are represented by OE (output error, OE) models and Box Jenkins (BJ) models. The former is mainly suitable for situations where the type of noise is known, while the latter is mainly suitable for situations where noise mainly exists in the output data. Since the system transfer function and noise transfer function in ARX and ARMAX models are not independent of each other, but share a portion of the same transfer function A(s), using these models in this paper would make the identification process very complicated. Considering that the switching noise and low-frequency harmonics of the microgrid black box model will not have a significant impact on model identification, in order to balance the accuracy and simplicity of the model, this paper chooses the OE model for parameter identification, where the noise transfer (H(s)=1).

[0061] After determining the model structure, it is necessary to select an appropriate model order. An iterative testing method can be used to test the fit between the step response of different model orders and the measured step response of the microgrid, and select the model that meets the required fit. Furthermore, considering the importance of computational complexity and speed in model testing, it is recommended to choose a lower model order as the identification result, provided the fit condition is met.

[0062] For example, based on the original data sequence obtained in step 101, iterative tests are performed on combinations with denominator orders of 2-4 and numerator orders of 1-3, and the goodness of fit is used for evaluation. Then, the combination with the lowest order that meets the goodness of fit requirement is selected from the combinatorial options. The goodness of fit can be expressed by the formula... Calculated.

[0063] Step 103: Based on the model framework, use the LM algorithm for iterative solution to output an initial identification model for describing the microgrid system.

[0064] In this step, the core of model parameter identification lies in selecting a suitable algorithm to identify the transfer function. There are many algorithms for transfer function identification, such as the step response method and the impulse response method, which use response curves to identify the transfer function, and iterative algorithms like the least squares method to obtain the identification parameters. The former, also known as the graphical method, requires high accuracy of the response curve, and the selection of key points such as the maximum slope on the curve is greatly influenced by subjective factors. However, when identifying the transfer function in this paper, there is an important prerequisite: understanding the time-domain form of the transfer function, which involves performing an inverse Laplace transform on the transfer function and indirectly using the identification algorithm to obtain the parameters in the time-domain form. But in most cases, only the order information of the numerator and denominator of the transfer function can be obtained, not its specific form in the time domain. Therefore, we consider analyzing a transfer function identification method based on the least squares method, which has greater flexibility and applicability.

[0065] The least squares method, also known as the least squares method, is based on the principle of minimizing the sum of squared errors. This involves sampling the input and output data of the system to be identified. The input data is applied to the system identification function, and then the system's measured output data is subtracted from the obtained output. The sum of the squares of these differences yields the system identification function with the minimum error.

[0066] The least squares method can be represented by the cost function (COF) shown below:

[0067] Where u(k) is the model input, y(k) is the model output, G(s) is the transfer function for identification, and N is the total number of samples. The criterion for minimizing the cost function is the least squares criterion, and the method for obtaining parameter estimates using this criterion is the least squares algorithm. This algorithm is not limited to identifying linear systems, but is also applicable to the field of nonlinear least squares, including iterative least squares algorithms represented by the Gauss-Newton algorithm and the Levenberg-Marquardt algorithm.

[0068] Furthermore, the Gauss-Newton (GS) method, as a nonlinear least squares optimization algorithm, is developed based on Newton's method. It effectively solves the problem of the excessive computational complexity of Newton's method. The basic idea of ​​the Gauss-Newton method is to first select initial values, and then use Taylor series expansion to approximate the nonlinear regression model. Through multiple iterations, the identification results are continuously corrected until the identification error meets the requirements or the number of iterations reaches the set upper limit, at which point the iteration stops and the identification results are output. The specific steps can be implemented as follows:

[0069] Assume the system to be identified consists of continuous nonlinear functions, and n sets of observation data have been collected from this system. , ,..., ,satisfy In this formula, x is the input data for recognition, and y is the output data for recognition. It is a vector consisting of m parameters to be identified, and n≥m.

[0070] In order to find identification The optimal solution requires summing the squared errors, i.e. ,in Represented as residuals: .

[0071] Further find E with respect to The partial derivative of , and set it equal to 0, then we have:

[0072] In nonlinear systems, the right-hand side partial derivatives of the above equations are functions of the variables and parameters, and there is no closed-form solution. Therefore, an initial value needs to be provided, and an approximation is made using an iterative method:

[0073] In this equation, Indicates the k-th iteration The estimated value. The iterative formula is expanded using a Taylor series, linearized, and the first two terms of the expansion are retained:

[0074] in, At this point, the residual can be expressed as:

[0075] Substituting the residual expression into E with respect to After simplification, the partial derivative equations of can be obtained as follows:

[0076] To facilitate computer iteration, it is rewritten in matrix form:

[0077]

[0078] Therefore, the increment of the GS method in each iteration is:

[0079] in, Representative system function The Jacobian matrix.

[0080] When using the GS method for iteration, it is necessary to assume the matrix. It is full rank. However, in actual iterative solutions, it is often impossible to guarantee that the matrix is ​​full rank. All matrices are full rank. Once the matrix becomes a singular matrix, the iterative algorithm cannot continue. Furthermore, the GS method has high requirements for the choice of initial values; improper initial value selection can also lead to non-convergence. To address the potential issues of non-invertible Jacobian matrices and non-convergence in the GS method, the Levenberg-Marquardt algorithm (LM method) is considered to effectively solve these problems. The LM method features fast convergence, no impact of initial value selection on the number of iterations, and significantly higher accuracy than similar iterative algorithms, making it an improved GS method.

[0081] Compared to the GS method, the LM method adds a damping term to the recursive formula:

[0082] in Represents the identity matrix. This is the damping coefficient. Initial value. Choose as When the matrix When it is a non-invertible matrix, That is, each time The value is magnified 10 times, and then recalculated. The invertibility of the matrix is ​​checked until it becomes invertible before proceeding with further iterations. Considering the numerous advantages of the LM method compared to Newton's method and the GS method, this scheme chooses the LM method to identify the transfer function.

[0083] It is important to note that both the LM and GS methods involve solving for the Jacobian matrix in their recursive formulas, which requires differentiating the functions in the identification system. Since the transfer function, as a special function, does not have a corresponding derivative form, it needs to be transformed into a solution using the amplitude-frequency characteristics of the transfer function. An example analysis is given using a transfer function with a third-order denominator and a second-order numerator, as shown in the following equation. , , , , These are the parameters to be identified. Substituting into this equation, we can obtain the amplitude-frequency expression for the step response corresponding to this equation:

[0084] Furthermore, when using the LM algorithm for iterative solution, it can be implemented in the following way:

[0085] Based on the structure and order of the identification model, determine the transfer function and the parameter vector to be identified;

[0086] Set the initial values ​​for the recognition model;

[0087] Calculate the Jacobian matrix of the frequency domain response relative to the variable, and construct the incremental normal equation of the variable based on the prediction error between the model output and the measured output at the k-th iteration;

[0088] Solve the incremental normal equation and calculate the prediction error between the model output and the measurement output at the (k+1)th iteration;

[0089] Based on the prediction error at the (k+1)th iteration, perform iterative updates and convergence judgment, and output the initial identification model after convergence.

[0090] In this embodiment, the LM iteration method is considered to be used to determine the estimated values ​​of each parameter in the following equation:

[0091] For example, the iteration steps can be as follows:

[0092] Step 1: Given initial parameters, magnification factor and The termination control error constant is selected. The initial parameter estimates;

[0093] Step 2: Calculation Relative to variables Jacobian matrix and calculate To build about Incremental normal equation .here, f is the prediction error between the model output and the measurement output at the k-th iteration. s ≥f c ;

[0094] Step 3: Solve the incremental normal equation to obtain And calculate the prediction error between the model output and the measured output at the (k+1)th iteration. ;

[0095] Step 4: If Then let ;if Stop the iteration and output the result. Otherwise, let... Return to step 2 to solve; if ,make Resolve the incremental normal equation to obtain Then return to step 4.

[0096] Step 104: Verify the initial identification model to obtain the final identification model that meets the fitting requirements.

[0097] In this step, to determine whether the initial identification model can accurately reproduce the dynamic response of the microgrid, we consider validating the initial identification model. Specifically, this can be achieved in the following way:

[0098] The initial identification model and the microgrid experimental device are provided with the same step input, and their respective step response outputs are output.

[0099] Calculate the best fit based on the respective step response outputs;

[0100] Determine whether the optimal fit meets the preset requirements;

[0101] If the optimal fit meets the preset requirements, the current initial identification model is deemed to have passed the verification; if the optimal fit does not meet the preset requirements, the order of the identification model is re-determined and identification is performed again until the optimal fit requirements are met.

[0102] The optimal fit can be calculated using the following formula:

[0103] In the formula, Used to characterize the best fit. Used to characterize the step response output by the initial identification model. Used to characterize the measured output of the microgrid experimental setup Used to characterize the average value of the measured output.

[0104] Step 105: Based on the final identification model, tune the parameters of the secondary voltage-frequency controller.

[0105] In this step, based on the final identification model obtained, the parameters of the secondary voltage-frequency controller can be tuned by collecting data from the input and output ports of the microgrid system.

[0106] The feasibility of this scheme and the accuracy of its model parameter identification will be further verified through experiments.

[0107] 1. Microgrid Experimental Platform

[0108] Considering such Figure 2A series of model parameter identification experiments were conducted on the microgrid experimental platform shown. This platform includes an Energy Management System (EMS), a Microgrid Central Controller (MGCC), two photovoltaic inverters, three energy storage converters, and active and reactive loads. All power nodes in the system are connected via smart gateway circuit breakers. To achieve communication between the microgrid central controller and the underlying devices, the system is built on the EtherCAT bus and uses a conversion protocol to achieve communication between the Energy Management System and the smart gateway circuit breakers. The main equipment in the system is described below:

[0109] (1) The inverter control board in the system is based on TI's TMS320F28335 chip, which has floating-point operation capability and a maximum frequency of 150MHz. The inverter adopts a three-phase half-bridge structure. The DC side is connected to the DC power supply through a DC filter, and the bridge arm output side is connected to the AC grid through an LC filter, isolation transformer, fuse, etc. Its main functions include analog signal sampling and processing, protection function, control function, and communication function.

[0110] (2) The MGCC in the system is built using Beckhoff's CX5130 master station, which is located between the bottom inverter and the upper control system and plays a key role. Its main functions include coordinating and controlling the microgrid, fault detection and handling, and providing a high-speed communication network.

[0111] (3) EMS is a higher-level control system of microgrid, capable of centralized energy distribution within the system, and also has monitoring and control functions. The main functions of EMS include providing various prediction algorithms, real-time optimization schemes and power dispatch plans for each unit.

[0112] During the experiment, the microgrid operated in islanded mode, and the microgrid central controller (MGCC) sent power reference commands to the energy storage converter via the communication system. Voltage and frequency signals from the point of common coupling (PCC) were read from the smart circuit breaker for system identification. Based on the transfer function to be identified in the model, active power reference step experiments and reactive power reference step experiments were conducted respectively.

[0113] 2. Active power step test

[0114] Under isolated microgrid load operation, the Microgrid Central Controller (MGCC) was used to issue active power reference values ​​to the energy storage converters in the microgrid system. During the experiment, the frequency, voltage amplitude, and power signals at the energy storage converter ports and the point of common coupling (PCC) were collected through the smart circuit breaker at the energy storage converter and PCC. Simultaneously, the issued reference power values ​​were recorded.

[0115] The same sampling frequency should be used when measuring and recording these data. The experimental waveforms show that at 5 seconds, when a 30kW load is connected to the microgrid, the microgrid frequency decreases. At 9.1 seconds, the microgrid MGCC adjusts the reference active power of the two parallel inverters from 0kW to 50kW, and the voltage at the point of common coupling (PCC) rises from 49.95Hz to 50.12Hz. During the experiment, the steady-state voltage amplitude of the system remains almost constant, indicating a low degree of coupling in the system.

[0116] 3. Identification of the frequency power transfer function

[0117] In the active power reference change experiment, the frequency waveform at the point of common coupling (PCC) shows slight fluctuations due to frequency sampling accuracy issues, but these do not affect the overall observation. First, the input power and output frequency data are processed to eliminate frequency fluctuations caused by the previous load; only the frequency response before and after the power reference value change is selected for identification.

[0118] Transfer function obtained through identification algorithm The response curve of the identification function was compared with the measured curve. The comparison results showed that the fitting degree was as high as 95.57%, indicating that the model response obtained by identification fits the sampled waveform in the experiment very well.

[0119]

[0120] 4. Identification of amplitude power transfer function

[0121] The voltage amplitude waveform at the PCC point acquired by the PCC intelligent circuit breaker contains a large amount of ripple. To improve the fitting accuracy of parameter identification, the voltage amplitude needs to be filtered. This scheme uses a smoothing function to smooth the waveform, and the active power reference value P0ST is also filtered using the same filtering parameters for model identification, thereby obtaining the transfer function between the active power reference and the voltage amplitude at the PCC point, as shown in the following equation:

[0122] Experimental results show that the voltage amplitude response waveform at the point of common coupling (PCC) during a 9.1-second step change in the power reference value indicates a relatively small coupling relationship between the active frequency and reactive voltage in the system. Therefore, when the active reference value changes, the voltage amplitude change at the PCC is almost negligible. This also results in the voltage amplitude sampling data, after bias removal, consisting almost entirely of small high-frequency ripple, leading to a low fit of only 20% for the identified model. However, considering the relatively small coupling of the model itself, this... It can be ignored in the model.

[0123] 5. Reactive power reference step test

[0124] Under the load operation conditions of an isolated microgrid, the active power reference value will be... The reactive power reference step signal was maintained at 0, and the microgrid central controller (MGCC) was used to send a reactive power reference step signal to the energy storage converter in the microgrid system. During the experiment, the voltage, frequency, voltage amplitude, and power signals at the energy storage inverter ports and the point of common coupling (PCC) were sampled and recorded using the same method, while the transmitted reference power value was also recorded. In the experiment, the frequency of the 30kW active load microgrid decreased at 2.8 seconds, and the voltage amplitude fluctuated at the moment of switching. At 5.8 seconds, the MGCC modified the reference reactive power of the two parallel-feed inverters from 0kW to 50kW, and the voltage amplitude at the PCC increased from 311V to 340V. Due to the slight coupling between active and reactive power in the system, the system frequency dropped slightly when the reactive power reference command was changed.

[0125] a) Identification of amplitude reactive power transfer function

[0126] For voltage amplitude data, bias removal processing is required, and then the 4-8 second interval is selected as the identification data. The identified transfer function is shown in the following equation, which is a second-order model. The comparison between the response waveform of the identified model and the sampled waveform shows that the identified model response fits the sampled waveform well, with a fit degree of 94.04%.

[0127]

[0128] b) Frequency-based reactive power transfer function

[0129] When the reactive power reference value Q0 issued by the microgrid central controller (MGCC) changes abruptly, the frequency change of the microgrid system is only -0.01Hz, indicating a very small coupling between reactive power and frequency. Through system identification, the transfer function from the reactive power reference value to the frequency can be obtained, as shown in the following equation. The comparison between the reactive power-frequency model response waveform and the sampled waveform shows that the model fit is 80.84%.

[0130]

[0131] c) Experiment on system secondary frequency and voltage recovery control

[0132] Experiments were conducted to compare the parameters designed using the traditional mechanistic analysis-based parameter tuning method with those designed using the method proposed in this scheme, in order to verify the effectiveness and speed of the proposed method.

[0133] Parametric controllers designed based on traditional mechanistic analysis may lead to frequency instability because the model does not consider the influence of micro-power sources and loads participating in non-microgrid systems. In the experiment, multiple inverters connected in parallel were first started and run under no-load for a period of time before an active load was applied after 3.5 seconds. After the load was applied, the system frequency dropped. At 5.4 seconds, secondary frequency regulation was initiated, but due to unreasonable secondary frequency regulation parameter design, the system oscillated. At 16 seconds, the parameters of the secondary frequency regulation controller were modified to those described in this paper, and the system stabilized. At 24 seconds, the system load was removed. After the load was removed, the system frequency quickly rose and returned to the set value. The experimental process is as follows: Figure 3 As shown.

[0134] Furthermore, the parameters designed based on traditional mechanistic analysis were compared with the parameters based on the identification model proposed in this paper. Data collected during the experiment are as follows: Figure 4 To verify the effectiveness of the secondary voltage controller parameters designed in this paper, an experimental analysis of the controller's regulation performance was conducted using a reactive load. Due to limitations of laboratory equipment, only a capacitive load was applied during the experiment. During the experiment, two parallel energy storage inverters were first started, initially under no-load conditions. After approximately 5 seconds, a capacitive reactive load was connected. It can be seen that after loading, the inverter's output reactive power is negative, and the system voltage rises. After approximately 8 seconds, the secondary voltage controller was started, but due to unreasonable controller parameter design, the system voltage amplitude became unstable. After 16 seconds, the controller parameters were changed to those designed in this application, and the system voltage gradually stabilized. After approximately 22 seconds, the reactive load was disconnected, and the system voltage stabilized after a brief oscillation. The experiment demonstrates that the secondary voltage controller parameters designed using the proposed method have good stability and can solve the difficulties in designing secondary voltage controllers for complex microgrid systems.

[0135] In summary, this scheme takes an islanded microgrid system as the research object, proposing a system-level identification modeling method for microgrids and a secondary voltage and frequency regulation strategy based on the identification model. First, the identification modeling method treats the microgrid system as a dual-input, dual-output black-box system, allowing for identification and modeling by measuring data from the input and output ports, avoiding the complex process of traditional mechanism modeling. Second, the black-box model obtained from the identification is used to optimize the design of the secondary voltage and frequency controller of the microgrid system. This can provide a solution and methodological reference for the parameter tuning problem of the secondary voltage and frequency regulation controller encountered when commercial power sources are widely deployed in microgrid systems. The offline identification method is used to design the parameters of the microgrid secondary voltage and frequency controller. Furthermore, the optimal identification algorithm was explored to obtain the model function with the highest fitting degree, and the evaluation index of the identified model was quantified to establish the model's fitting degree function. Then, using the operating data sampled from the input and output ports of the microgrid, the transfer function in the identified model was identified by combining the optimization algorithm. The identified model was embedded into the open-loop transfer function of the system to optimize the design of the microgrid's secondary voltage and frequency controller. This solved the problems of difficult mechanism modeling and low parameter tuning accuracy of the secondary controller in the widespread application of commercial micro-power sources in microgrid systems. By constructing a low-order black box model driven by data, the precise decoupling design and adaptive adjustment of the voltage-frequency controller were realized, effectively improving the voltage and frequency stability of the microgrid and providing an efficient and reliable parameter tuning method for the dynamic performance optimization of microgrids.

[0136] The present invention also provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it runs the method as described in the above embodiments.

[0137] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method as described in any of the above embodiments.

[0138] The device and method embodiments of the present invention are based on the same inventive concept. For detailed description, please refer to the method embodiments, which will not be repeated here.

[0139] The modules or units in the device of this invention can be merged, divided, and deleted according to actual needs. The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the processes of the above embodiments and making equivalent changes according to the claims of this invention still fall within the scope of the invention.

Claims

1. A method for parameter tuning of a secondary voltage-frequency controller based on system identification, characterized in that, include: The data used for system identification is sampled according to the preset sampling rules to obtain the original data sequence; Based on the original data sequence, determine the identification model structure and order adapted to the black box characteristics of the microgrid, and output the model framework; Based on the aforementioned model framework, the LM algorithm is used for iterative solution to output an initial identification model for describing the microgrid system. The initial identification model is validated to obtain the final identification model that meets the fitting requirements; Based on the final identification model, the parameters of the secondary voltage-frequency controller are tuned.

2. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 1, characterized in that, The sampling frequency satisfies ; wherein, is used to represent the sampling frequency, is used to represent the signal frequency.

3. The system identification based secondary voltage-frequency controller parameter tuning method of claim 2, wherein, After obtaining the original data sequence, the following further steps are included: For each type of original data sequence, a steady-state data segment is selected to calculate the mean, and the difference between the original data sequence and the mean is calculated to obtain the corrected data after baseline offset removal; The corrected data is then subjected to noise filtering to obtain a data sequence for model construction.

4. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 1, characterized in that, In the process of determining the identification model structure and order adapted to the black box characteristics of the microgrid based on the original data sequence, the model structure is determined to be an OE model.

5. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 4, characterized in that, The step of determining the order of the identification model adapted to the black-box characteristics of the microgrid based on the original data sequence includes: Iterative testing was performed on combinations of denominator and numerator orders; Determine the degree of fit between the step response of the model order corresponding to each combination and the measured step response of the microgrid; From various combinations of model orders that meet the degree of fit, the combination of model orders with the lowest order is selected as the identification model order that adapts to the black-box characteristics of microgrids.

6. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 1, characterized in that, The iterative solution using the LM algorithm includes: Based on the structure and order of the identification model, determine the transfer function and the parameter vector to be identified; Set the initial values ​​for the recognition model; Calculate the Jacobian matrix of the frequency domain response relative to the variable, and construct the incremental normal equation of the variable based on the prediction error between the model output and the measured output at the k-th iteration; Solve the incremental normal equation and calculate the prediction error between the model output and the measurement output at the (k+1)th iteration; Based on the prediction error at the (k+1)th iteration, perform iterative updates and convergence judgment, and output the initial identification model after convergence.

7. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 6, characterized in that, The verification of the initial identification model includes: The initial identification model and the microgrid experimental device are provided with the same step input, and their respective step response outputs are output. Calculate the best fit based on the respective step response outputs; Determine whether the optimal fit meets the preset requirements; If the optimal fit meets the preset requirements, the current initial identification model is deemed to have passed the verification; if the optimal fit does not meet the preset requirements, the order of the identification model is re-determined and identification is performed again until the optimal fit requirements are met.

8. The method for tuning secondary voltage-frequency controller parameters based on system identification according to claim 7, characterized in that, The calculation of the best fit based on the respective step response outputs includes: The best fit is calculated according to the following formula: ； In the formula, Used to characterize the best fit. Used to characterize the step response output by the initial identification model. Used to characterize the measured output of the microgrid experimental setup Used to characterize the average value of the measured output.

9. A computing device, comprising a memory and a processor, wherein executable code is stored in the memory, and when the processor executes the executable code, it performs the method as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method as described in any one of claims 1-8.

Images