Adaptive control methods, devices and equipment for grid-type main photovoltaic virtual synchronous machines
By constructing a grid-connected system model and a frequency shaping domain, and coordinating the inertia-damping-filtering links, the problem of virtual inertia parameter selection relying on experience in existing technologies is solved, and more flexible frequency adjustment and inertia support capabilities are achieved.
Patent Information
- Application Number
- CN202611172006.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-04
- Publication Date
- 2026-11-10
AI Technical Summary
In existing network-type virtual synchronous machine control technology, the selection of virtual inertia parameters and virtual damping parameters relies on experience and does not fully consider the impact of active power imbalance, resulting in insufficient flexibility in parameter adjustment and difficulty in maintaining effective frequency response and inertia support capability at multiple operating points.
A target grid-connected system model is constructed, candidate control parameters are determined, inertia support parameters are identified through the equivalent inertia transfer function, a frequency shaping domain is constructed by combining small-signal stability constraints and frequency response characteristics at multiple operating points, and an adjustment amount is generated and corrected using a data-driven parameter adaptive adjustment strategy to coordinate the dynamic response of the inertia-damping-filtering link.
It improves the inertia support capability of the main photovoltaic virtual synchronous machine in the grid-type system, ensures that the parameter combination meets the constraints of small signal stability and frequency response characteristics at multiple operating points, reduces parameter overshoot, and gives full play to the frequency regulation capability.
Smart Images

Figure CN123073707A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of virtual synchronous machine control technology, and in particular to an adaptive control method, device and equipment for a grid-type main photovoltaic virtual synchronous machine. Background Technology
[0002] With the continued advancement of the "dual carbon" goals, a new power system dominated by new energy sources is being rapidly constructed. The penetration rate of new energy sources such as photovoltaics and wind power in the power grid is constantly increasing, while the proportion of traditional synchronous generators in the system is relatively decreasing, consequently weakening the grid's voltage support, frequency support, and inertia support capabilities. Grid-based control technology, because it enables new energy converters to possess voltage source characteristics and actively participate in grid dynamic regulation, is better suited to the operational needs of new energy sources transitioning from auxiliary to primary power sources, and is considered an important technical path for adapting to a new energy-dominated power system. Among these, grid-based virtual synchronous machine control, by simulating the swing equations of a synchronous generator to configure virtual inertia and virtual damping parameters for the converter, allows the converter to exhibit dynamic response characteristics similar to a synchronous generator when power disturbances occur in the grid, and is one of the mainstream grid-based control schemes.
[0003] In related technologies, to address the insufficient adaptability of fixed virtual synchronous machine parameter control under different operating conditions, various adaptive control schemes have been proposed. Regarding parameter selection, virtual inertia and virtual damping parameters are typically chosen empirically as control variables, and these are adjusted online based on the system state. For parameter constraints, independent upper and lower limits are usually used to constrain the control parameters, or a two-dimensional feasible region is derived based on a simplified model and a single operating point as the parameter adjustment range. Meanwhile, reinforcement learning has also been introduced into the field of virtual synchronous machine adaptive control in recent years, typically by designing reward functions with multiple indices to coordinate various performance objectives and output continuous parameter adjustments.
[0004] However, the above technologies still have the following shortcomings: First, since the rapid frequency support capability of the grid-type virtual synchronous machine is not solely determined by the virtual inertia and virtual damping parameters, the measurement method of active power imbalance and the corresponding filtering circuit also affect the dynamic response speed and support effect of the converter. Simply selecting and adjusting the virtual inertia and virtual damping parameters based on experience, without analyzing the impact of other control parameters on the inertia support capability, makes it difficult to fully utilize the flexibility and inertia support capability of the grid-type main photovoltaic virtual synchronous machine. Second, using independent upper and lower limits to constrain the adjustment parameters, or deriving a two-dimensional feasible region based on a simplified model and a single operating point as the parameter adjustment range, not only makes it difficult to characterize the coupling boundaries of multiple control parameters at multiple operating points, but also makes it difficult to ensure that the adjusted parameter combination always remains within the feasible range that satisfies the target frequency response characteristics. When the system operating point changes significantly, the feasible region of parameters designed based on a single operating point may fail, leading to parameter out-of-bounds errors and a decrease in control performance. Third, when designing a reward function that includes multiple indicators, there are often mutual constraints between different indicators. Such methods rely on empirical weight adjustments and it is difficult to guarantee that the adjusted parameter combination always meets the frequency response requirements and frequency tuning resource constraints of the target. This can easily lead to the agent over-optimizing on some indicators while exceeding the acceptable range on others. Summary of the Invention
[0005] This application provides an adaptive control method, apparatus, and device for a grid-type main photovoltaic virtual synchronous machine, which can fully utilize the flexibility of the grid-type main photovoltaic virtual synchronous machine, improve inertia support capability, and fully utilize the frequency regulation capability of the grid-type main photovoltaic virtual synchronous machine while reducing parameter out-of-bounds errors. The technical solution includes at least the following: On the one hand, an adaptive control method for a grid-connected main photovoltaic virtual synchronous machine is provided, comprising: constructing a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points, and determining candidate control parameters to be analyzed; determining an equivalent inertia transfer function based on the grid-connected main photovoltaic virtual synchronous machine model, and identifying inertia support parameters from the candidate control parameters according to the equivalent inertia transfer function, wherein the inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth; constructing a frequency shaping domain in the parameter space composed of the inertia support parameters based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points; acquiring the dynamic state information of the target grid-connected system, generating the adjustment amount of the inertia support parameters based on a data-driven parameter adaptive adjustment strategy and obtaining the adjusted inertia support parameter combination, and using the frequency shaping domain to correct the adjusted inertia support parameter combination to update the control parameters of the grid-connected main photovoltaic virtual synchronous machine.
[0006] Optionally, the grid-type main photovoltaic virtual synchronous machine model is a full-order average value model that includes active-frequency loop, reactive-voltage loop, virtual impedance, voltage inner loop, current inner loop, LC filter, transmission line and grid equivalent links.
[0007] Optionally, in the active power-frequency loop, the active power imbalance is first filtered, and then the filtered active power imbalance is input into the swing equation; the active power imbalance before filtering is expressed as:
[0008] in, This represents the active power imbalance before filtering. This is an active power command. The actual active power injected into the grid by the main grid-connected photovoltaic virtual synchronous machine at the point of common coupling; the dynamic relationship of the filtered active power imbalance is expressed as:
[0009] in, This represents the filtered active power imbalance. Indicates time, The active power imbalance filter bandwidth is denoted by , and the cutoff angular frequency of the low-pass filter for the active power imbalance channel is denoted by . The expression for the oscillation equation of the grid-type main photovoltaic virtual synchronous machine model is as follows:
[0010] in, For virtual inertia parameters, For virtual damping parameters, This refers to the internal angular frequency of the converter. This is the rated angular frequency.
[0011] Optionally, the expression for the equivalent inertia transfer function is:
[0012] in, The equivalent inertia transfer function, For complex frequency variables, This is the rated power.
[0013] Optionally, the candidate control parameters include virtual inertia parameters, virtual damping parameters, active power imbalance filter bandwidth, voltage inner loop proportional gain, voltage inner loop integral gain, current inner loop proportional gain, current inner loop integral gain, virtual resistance, and virtual reactance; the step of identifying inertia support parameters from the candidate control parameters based on the equivalent inertia transfer function includes: performing sensitivity analysis and closed-loop stability boundary analysis on the candidate control parameters based on the equivalent inertia transfer function, so as to use the virtual inertia parameters, virtual damping parameters, and active power imbalance filter bandwidth as the inertia support parameters.
[0014] Optionally, the step of constructing a frequency shaping domain within the parameter space constituted by the inertia support parameters, based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points, includes: establishing a linearized model for each operating point and each combination of inertia support parameters within the parameter space constituted by the inertia support parameters.
[0015]
[0016] in, For the combination of inertia support parameters, , For virtual inertia parameters, For virtual damping parameters, The active power imbalance filter bandwidth. For the running point, For state increment, This represents the increment of external disturbances. State increment The first derivative with respect to time, This is the output increment that includes frequency deviation and active power increment. , , , These represent the combinations of inertia support parameters. At the running point The full-order closed-loop state matrix, disturbance input matrix, output matrix, and direct transfer matrix are calculated. Based on the linearized model, the small-signal stability index, frequency response characteristic index, and frequency modulation resource index of each inertia support parameter at each operating point are calculated. The small-signal stability index includes the abscissa of the closed-loop state matrix spectrum. The frequency response characteristic index includes the maximum absolute frequency deviation, maximum absolute frequency change rate, steady-state frequency deviation, and settling time. The frequency modulation resource index includes peak support power and support energy. Simultaneously, small-signal stability constraints, maximum absolute frequency deviation constraints, maximum absolute frequency change rate constraints, steady-state frequency deviation constraints, settling time constraints, peak support power constraints, and support energy constraints are set. In the parameter space, combinations of inertial support parameters that satisfy the small-signal stability constraint, the maximum absolute frequency deviation constraint, the maximum absolute frequency change rate constraint, the steady-state frequency deviation constraint, the settling time constraint, the peak support power constraint, and the support energy constraint at all representative operating points are selected to form an initial discrete frequency shaping domain. The initial discrete frequency shaping domain is subjected to a two-stage discrete scan to obtain a final discrete frequency shaping domain. The final discrete frequency shaping domain is conservatively continuous using an orthogonal element reconstruction method based on the full vertex consistency criterion to obtain a conservative continuous frequency shaping domain. The frequency shaping domain is constructed based on the final discrete frequency shaping domain and the conservative continuous frequency shaping domain.
[0017] Optionally, the dynamic state information of the target grid-connected system is represented as follows:
[0018] in, This refers to the dynamic status information of the target grid-connected system. for Frequency deviation at time, for Rate of change of frequency at time, This indicates the control time step; the adjustment amount of the inertia support parameter is expressed as:
[0019] in, This refers to the adjustment amount of the inertia support parameter. for The adjustment amount of the virtual inertia parameter at any given time. for The adjustment amount of the virtual damping parameter at any given time. for The adjustment amount of the active power imbalance filter bandwidth at all times, and the adjusted inertia support parameter combination are as follows: The step of correcting the adjusted inertia support parameter combination using the frequency shaping domain includes: performing motion correction mapping using the frequency shaping domain.
[0020] in, The corrected inertia support parameter combination, This refers to the frequency shaping domain. Indicates motion correction mapping; when Located in the frequency shaping domain hour, ;when Located in the frequency shaping domain When outside, Set as frequency shaping domain Internal and The nearest combination of inertia support parameters.
[0021] Optionally, the data-driven parameter adaptive adjustment strategy employs a dual-delay deep deterministic strategy gradient algorithm.
[0022] On the other hand, an adaptive control device for a grid-connected main photovoltaic virtual synchronous machine is provided, comprising: a model building module for building a model of the grid-connected main photovoltaic virtual synchronous machine of the target grid-connected system at multiple operating points and determining candidate control parameters to be analyzed; a parameter identification module for determining an equivalent inertia transfer function based on the grid-connected main photovoltaic virtual synchronous machine model, and identifying inertia support parameters from the candidate control parameters according to the equivalent inertia transfer function, wherein the inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth; a frequency shaping domain construction module for constructing a frequency shaping domain within the parameter space constituted by the inertia support parameters, based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points; and a state acquisition and correction module for acquiring the dynamic state information of the target grid-connected system, generating the adjustment amount of the inertia support parameters based on a data-driven parameter adaptive adjustment strategy and obtaining the adjusted inertia support parameter combination, and using the frequency shaping domain to correct the adjusted inertia support parameter combination to update the control parameters of the grid-connected main photovoltaic virtual synchronous machine.
[0023] In another aspect, a computer device is provided, the computer device comprising: a memory and a processor, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement the aforementioned adaptive control method for a grid-connected main photovoltaic virtual synchronous machine.
[0024] The beneficial effects of the technical solution provided in this application include at least the following: In this embodiment, by constructing a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points and determining the candidate control parameters to be analyzed, the equivalent inertia transfer function is determined based on the grid-connected main photovoltaic virtual synchronous machine model. Based on the equivalent inertia transfer function, virtual inertia parameters, virtual damping parameters, and active power imbalance filter bandwidth are identified from the candidate control parameters as inertia support parameters. This avoids relying solely on experience to select virtual inertia parameters and virtual damping parameters as adaptive control variables. By incorporating the active power imbalance filter bandwidth into the coordinated adjustment, the low-pass filter is dynamically coordinated with the inertia-damping adjustment, fully leveraging the flexibility of the grid-connected main photovoltaic virtual synchronous machine and improving its inertia support capability. Then, within the parameter space composed of inertia support parameters, a frequency shaping domain is constructed based on small-signal stability constraints, frequency response characteristics, and frequency regulation resource constraints at multiple operating points. This domain can accurately characterize the coupling boundaries between virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth at multiple operating points. Next, the dynamic state information of the target grid-connected system is obtained. Based on a data-driven parameter adaptive adjustment strategy, the adjustment amount of the inertia support parameters is generated, and the adjusted inertia support parameter combination is obtained. Then, the frequency shaping domain is used to correct the adjusted inertia support parameter combination. Without designing complex reward functions or empirical weight adjustments, it can effectively ensure that the corrected inertia support parameter combination simultaneously satisfies small-signal stability constraints, target frequency response characteristics, and frequency regulation resource constraints at multiple operating points. Thus, it can fully leverage the frequency regulation capability of the grid-connected main photovoltaic virtual synchronous machine while reducing parameter out-of-bounds errors. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is a flowchart of an adaptive control method for a grid-type main photovoltaic virtual synchronous machine provided in an embodiment of this application.
[0027] Figure 2 This is a structural block diagram of a grid-type main photovoltaic virtual synchronous machine adaptive control device provided in an embodiment of this application.
[0028] Figure 3 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation
[0029] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The terms “connected” or “linked” and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.
[0030] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0031] Figure 1 This is a flowchart of an adaptive control method for a grid-type main photovoltaic virtual synchronous machine provided in an embodiment of this application. Figure 1 As shown, the method includes: Step S1: Construct a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points, and determine the candidate control parameters to be analyzed.
[0032] Step S2: Determine the equivalent inertia transfer function based on the grid-type main photovoltaic virtual synchronous machine model, and identify the inertia support parameters from the candidate control parameters according to the equivalent inertia transfer function.
[0033] The inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth.
[0034] Step S3: Within the parameter space composed of inertia support parameters, construct a frequency shaping domain based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points.
[0035] Step S4: Obtain the dynamic status information of the target grid-connected system, generate the adjustment amount of the inertia support parameters based on the data-driven parameter adaptive adjustment strategy, and obtain the adjusted inertia support parameter combination. Use the frequency shaping domain to correct the adjusted inertia support parameter combination in order to update the control parameters of the grid-connected main photovoltaic virtual synchronous machine.
[0036] In this embodiment, by constructing a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points and determining the candidate control parameters to be analyzed, the equivalent inertia transfer function is determined based on the grid-connected main photovoltaic virtual synchronous machine model. Based on the equivalent inertia transfer function, virtual inertia parameters, virtual damping parameters, and active power imbalance filter bandwidth are identified from the candidate control parameters as inertia support parameters. This avoids relying solely on experience to select virtual inertia parameters and virtual damping parameters as adaptive control variables. By incorporating the active power imbalance filter bandwidth into the coordinated adjustment, the low-pass filter is dynamically coordinated with the inertia-damping adjustment, fully leveraging the flexibility of the grid-connected main photovoltaic virtual synchronous machine and improving its inertia support capability. Then, within the parameter space composed of inertia support parameters, a frequency shaping domain is constructed based on small-signal stability constraints, frequency response characteristics, and frequency regulation resource constraints at multiple operating points. This domain can accurately characterize the coupling boundaries between virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth at multiple operating points. Next, the dynamic state information of the target grid-connected system is obtained. Based on a data-driven parameter adaptive adjustment strategy, the adjustment amount of the inertia support parameters is generated, and the adjusted inertia support parameter combination is obtained. Then, the frequency shaping domain is used to correct the adjusted inertia support parameter combination. Without designing complex reward functions or empirical weight adjustments, it can effectively ensure that the corrected inertia support parameter combination simultaneously satisfies small-signal stability constraints, target frequency response characteristics, and frequency regulation resource constraints at multiple operating points. Thus, it can fully leverage the frequency regulation capability of the grid-connected main photovoltaic virtual synchronous machine while reducing parameter out-of-bounds errors.
[0037] For example, in step S1, the grid-type main photovoltaic virtual synchronous machine model is a full-order average value model that includes an active power-frequency loop, a reactive power-voltage loop, virtual impedance, an inner voltage loop, an inner current loop, an LC filter, transmission lines, and equivalent grid components. In the grid-type main photovoltaic virtual synchronous machine model, the DC side can use an equivalent constant voltage source, and external disturbances can be represented by changes in active power command or load.
[0038] For example, in the active power-frequency loop, the active power imbalance is first filtered, and then the filtered active power imbalance is input into the swing equation.
[0039] The active power imbalance before filtering is expressed as:
[0040] in, This represents the active power imbalance before filtering. This is an active power command. The actual active power injected into the grid by the main photovoltaic virtual synchronous machine in the grid-connected system at the common coupling point.
[0041] The dynamic relationship of the filtered active power imbalance is expressed as follows:
[0042] in, This represents the filtered active power imbalance. Indicates time, The active power imbalance filter bandwidth represents the cutoff angular frequency of the low-pass filter for the active power imbalance channel. By setting up the filtering stage, high-frequency power disturbances can be suppressed, and active power command step fluctuations can be mitigated.
[0043] For example, the expression for the swing equation of the grid-type main photovoltaic virtual synchronous machine model is as follows:
[0044] in, For virtual inertia parameters, For virtual damping parameters, This refers to the internal angular frequency of the converter. This is the rated angular frequency.
[0045] In this embodiment, the candidate control parameters include virtual inertia parameters, virtual damping parameters, active power imbalance filter bandwidth, voltage inner loop proportional gain, voltage inner loop integral gain, current inner loop proportional gain, current inner loop integral gain, virtual resistance, and virtual reactance. Specifically, the candidate control parameters can be expressed as:
[0046] in, For the set of candidate control parameters, and These are the voltage inner loop proportional gain and voltage inner loop integral gain, respectively. and These are the proportional gain and integral gain of the current inner loop, respectively. and These are virtual resistance and virtual reactance, respectively.
[0047] For example, in step S2, to analyze the key control parameters affecting the inertia support characteristics, under small power angle and inductive grid conditions, the increase in output active power at the common coupling point can be expressed as:
[0048] in, The output active power increment at the common coupling point. For complex frequency variables, For synchronization power factor, This refers to the change in angular frequency inside the converter. The disturbance is the grid angular frequency. By combining the active power imbalance filtering circuit and the oscillation equation, the transmission relationship from the grid angular frequency disturbance to the output active power and the internal angular frequency of the converter can be determined.
[0049] For example, the expression for the equivalent inertia transfer function is:
[0050] in, The equivalent inertia transfer function, For complex frequency variables, The rated power is given. According to the equivalent inertia transfer function, the inertia support characteristics exhibited by the main grid-type photovoltaic virtual synchronous generator are jointly influenced by the virtual inertia parameter, virtual damping parameter, and active power imbalance filter bandwidth. Specifically, the virtual inertia parameter mainly affects the response strength of the converter's power increment to the rate of frequency change; the virtual damping parameter mainly affects the power regulation related to frequency deviation; the active power imbalance filter bandwidth determines the effective bandwidth of the power imbalance response. A lower filter bandwidth can suppress high-frequency disturbances but reduces the fast support capability, while a higher filter bandwidth can improve the response speed but increases the system's sensitivity to high-frequency dynamics and noise. Therefore, there is a coupling relationship between the virtual inertia parameter, virtual damping parameter, and active power imbalance filter bandwidth.
[0051] For example, in step S2, the inertia support parameters are identified from the candidate control parameters according to the equivalent inertia transfer function, including: performing sensitivity analysis and closed-loop stability boundary analysis on the candidate control parameters according to the equivalent inertia transfer function, so as to use the virtual inertia parameter, virtual damping parameter and active power imbalance filtering bandwidth as the inertia support parameters.
[0052] Specifically, for the candidate control parameter set Any candidate control parameter in It can calculate the equivalent inertia transfer function with respect to candidate control parameters within the target matching frequency band. Normalized sensitivity:
[0053] in, The equivalent inertia transfer function with respect to candidate control parameters Normalized sensitivity, The imaginary unit, For the angular frequency to be analyzed, This is the index of the candidate control parameter. Indicates candidate control parameters Parameter values after relative perturbation , The disturbance coefficient is... and These represent the equivalent inertia transfer functions for the candidate control parameters when they are taken at values before and after the disturbance, respectively. Normalization can reduce the impact of differences in parameter dimensions and nominal values on the comparison results.
[0054] In the target matching frequency band Within the band, sensitivity analysis can be performed using the in-band average sensitivity index and the in-band worst-case sensitivity index, respectively:
[0055]
[0056] in, The in-band average sensitivity index is used to characterize the overall impact of candidate control parameters within the target matching frequency band. This is the worst-case sensitivity index within the frequency band, used to characterize the maximum impact of a parameter within that frequency band. and These are the lower and upper angular frequencies of the target matching frequency band, respectively.
[0057] To avoid selecting inertia support parameters solely based on the amplitude-frequency characteristics of the target channel, we can examine low-frequency oscillation mode damping and the full-order closed-loop stability boundary. For the first... Conjugate oscillation modes = +j Its damping ratio can be expressed as:
[0058] in, For the first Damping ratio of each conjugate oscillation mode, and The first The real and imaginary parts of each conjugate oscillation mode are determined. The conjugate oscillation mode with the largest real part within the target frequency band is selected as the dominant oscillation mode, and the candidate control parameters are determined accordingly. Dominant oscillation mode damping ratio sensitivity:
[0059] in, For candidate control parameters Dominant oscillation mode damping ratio sensitivity and Candidate control parameters The damping ratio of the dominant oscillation mode before and after the disturbance. Simultaneously, closed-loop stability boundary analysis can be performed using the abscissa of the closed-loop state matrix spectrum and its offset index.
[0060]
[0061] in, The x-coordinate of the closed-loop state matrix spectrum is represented by... Represents the full-order closed-loop state matrix. The closed-loop state matrix The 1 eigenvalue, The offset index is the horizontal axis of the closed-loop state matrix spectrum. and They represent candidate control parameters. The horizontal axis of the closed-loop state matrix spectrum before and after the disturbance.
[0062] For example, in step S2, the aforementioned indices for each candidate control parameter can be calculated and compared. The parameter that ranks highly among multiple indices and whose value is significantly higher than other candidate control parameters is identified as the inertia support parameter. This allows for the identification of virtual inertia parameters. Virtual damping parameters And active power imbalance filter bandwidth These are the key parameters affecting the characteristics of inertia support, namely, inertia support parameters.
[0063] For example, in step S3, within the parameter space composed of inertia support parameters, a frequency shaping domain is constructed based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points, including the following steps: Step S31: Within the parameter space defined by the inertia support parameters, establish a linearized model for each operating point and each combination of inertia support parameters:
[0064]
[0065] in, For the combination of inertia support parameters, , For virtual inertia parameters, For virtual damping parameters, The active power imbalance filter bandwidth. For the running point, For state increment, This represents the increment of external disturbances. State increment The first derivative with respect to time, This is the output increment that includes frequency deviation and active power increment. , , , These represent the combinations of inertia support parameters. At the running point The full-order closed-loop state matrix, disturbance input matrix, output matrix, and direct transfer matrix are defined. Changes in the operating point and parameter combinations will alter the state-space matrix of the linearized model, thereby affecting the transient response characteristics of the grid-type main photovoltaic virtual synchronous machine.
[0066] Step S32: Calculate the small-signal stability index, frequency response characteristic index, and frequency modulation resource index of each inertia support parameter at each operating point based on the linearized model. The small-signal stability index includes the abscissa of the closed-loop state matrix spectrum. The frequency response characteristic index includes the maximum absolute frequency deviation, the maximum absolute frequency change rate, the steady-state frequency deviation, and the settling time. The frequency modulation resource index includes the peak support power and the support energy. At the same time, set the small-signal stability constraint, the maximum absolute frequency deviation constraint, the maximum absolute frequency change rate constraint, the steady-state frequency deviation constraint, the settling time constraint, the peak support power constraint, and the support energy constraint.
[0067] For example, the small-signal stability constraint can be expressed as:
[0068] in, Combination of inertia support parameters At the running point The horizontal axis of the full-order closed-loop state matrix spectrum.
[0069] For example, it can be set for Moment of inertia support parameter combination At the running point The frequency deviation below, For the analysis of time windows, where, To analyze the final moment of the time window, the maximum absolute frequency deviation, the maximum absolute rate of change of frequency, and the steady-state frequency deviation can be expressed as follows:
[0070]
[0071]
[0072] in, Combination of inertia support parameters At the running point The maximum absolute frequency deviation below, Combination of inertia support parameters At the running point The maximum absolute rate of change of frequency, To analyze the combination of inertia support parameters at the end of the time window At the running point The frequency deviation below, which represents the combination of inertia support parameters. At the running point The steady-state frequency deviation.
[0073] Using the steady-state frequency value as the recovery reference frequency, the frequency recovery judgment error band and settling time are defined as follows:
[0074]
[0075] in, For frequency recovery determination error band, This is the proportional coefficient for the frequency recovery determination error band. To adjust the time.
[0076] For example, it can be set For the active power increment of the main grid-type photovoltaic virtual synchronous generator, the peak supported power and supported energy can be expressed as follows:
[0077]
[0078] in, For peak support power, To provide energy.
[0079] Based on the above small-signal stability index, frequency response characteristic index, and frequency modulation resource index, small-signal stability constraints, maximum absolute frequency deviation constraints, maximum absolute frequency change rate constraints, steady-state frequency deviation constraints, settling time constraints, peak support power constraints, and support energy constraints can be set.
[0080] Step S33: Select inertial support parameter combinations that satisfy small-signal stability constraints, maximum absolute frequency deviation constraints, maximum absolute frequency change rate constraints, steady-state frequency deviation constraints, settling time constraints, peak support power constraints, and support energy constraints at all representative operating points in the parameter space to form the initial discrete frequency shaping domain.
[0081] For example, the initial discrete frequency shaping domain can be represented as:
[0082] in, This is the initial discrete frequency shaping domain. The parameter space to be scanned. This represents a set of representative operating points. The set of representative operating points can be set based on factors such as the pre-fault active power level of the main grid-connected photovoltaic system, the grid short-circuit ratio, the grid reactance-resistance ratio, power command changes, and load changes. The set constraint thresholds can be determined based on the capacity of the target grid-connected system, the availability of frequency regulation resources, grid connection standards, and operational requirements. The parameter scan range, scan step size, frequency response analysis window, and parameter adaptive adjustment period can all be set according to the actual application scenario.
[0083] Step S34: Perform a two-stage discrete scan on the initial discrete frequency shaping domain to obtain the final discrete frequency shaping domain.
[0084] For example, when using a two-stage discrete scan, in the first stage, in the maximum possible parameter space By setting wide discrete candidate parameters within each section, a coarse scan grid is obtained. For any candidate parameter combination, the steady-state operating point is solved at each operating point, and a full-order linearized model is established. The spectral abscissa, time-domain response, and supporting resource indices are calculated to obtain the first-stage discrete frequency shaping domain. , .
[0085] In the second stage, based on the value range of each parameter in the discrete frequency shaping domain of the first stage, adjacent coarse grid layers are expanded on both sides to obtain a fine scan space. Set a fine scan grid within the fine scan space. The final discrete frequency shaping domain is obtained. , .
[0086] Step S35: The final discrete frequency shaping domain is conservatively continuous by using the orthogonal element reconstruction method based on the full vertex consistency criterion to obtain the conservative continuous frequency shaping domain.
[0087] Since the discrete frequency shaping domain consists of a finite number of discrete parameter samples, a conservative continuous transformation of the discrete frequency shaping domain can be achieved by using an orthogonal element reconstruction method based on the full vertex consistency criterion to accommodate continuous parameter adjustment processes. A three-dimensional local orthogonal element can be represented as:
[0088] in, For adjacent grid indices in the direction of virtual inertia parameters, and This represents the neighboring parameter values of the virtual inertia parameter in the scanned mesh. For adjacent grid indices in the direction of virtual damping parameters, and This represents the neighboring parameter values of the virtual damping parameter in the scanned grid. This represents the adjacent grid index along the active power imbalance filter bandwidth direction. and This represents the nearest parameter value to the filtering bandwidth of the active power imbalance in the scanned grid.
[0089] For any locally orthogonal element Let the set of all its vertices be denoted as . .when Belongs to the final discrete frequency shaping domain When this local orthogonal element is incorporated into the conservative continuous frequency shaping domain, , .
[0090] Step S36: Construct a frequency shaping domain based on the final discrete frequency shaping domain and the conservative continuous frequency shaping domain.
[0091] Some discrete parameter combinations may not be included in the conservative continuous frequency shaping domain; these parameter combinations can be denoted as... , and the conservative continuous frequency shaping domain By merging, a frequency shaping domain for motion correction is obtained. , .
[0092] It should be noted that the construction of the frequency shaping domain is not limited to two-stage grid scanning. Adaptive sampling, neural network classifiers, or other parameter domain analysis methods can also be used, but the parameter combination must satisfy the target frequency response characteristics and frequency modulation resource constraints.
[0093] For example, in step S4, the data-driven adaptive parameter adjustment strategy employs the dual-delay deep deterministic policy gradient algorithm (TD3 algorithm). In other embodiments, the data-driven adaptive parameter adjustment strategy may also employ other intelligent optimization algorithms capable of outputting continuous parameter adjustment amounts, such as the deep deterministic policy gradient algorithm (DDPG algorithm), the SAC (Soft Actor-Critic) algorithm, or the proximal policy optimization algorithm (PPO algorithm), and this application does not impose any limitations on this.
[0094] In this embodiment, the dynamic state information of the target grid-connected system is represented as follows:
[0095] in, For the dynamic status information of the target grid-connected system, for Frequency deviation at time, for Rate of change of frequency at time, This indicates the control time step.
[0096] The adjustment amount for the inertia support parameter is expressed as follows:
[0097] in, This is the adjustment amount for the inertia support parameters. for The adjustment amount of the virtual inertia parameter at any given time. for The adjustment amount of the virtual damping parameter at any given time. for The adjustment amount of the active power imbalance filter bandwidth at all times, and the adjusted inertia support parameter combination are as follows: .
[0098] During the training phase, the data-driven adaptive parameter adjustment strategy can generate adjustments to virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth based on the acquired frequency, rate of change of frequency, and time information. This leads to the formation of an adjusted combination of inertia support parameters based on the initial parameter settings. .
[0099] For example, in step S4, the adjusted inertia support parameter combination is corrected using the frequency shaping domain, including: performing motion correction mapping using the frequency shaping domain.
[0100] in, The corrected inertia support parameter combination, For frequency shaping domain, This represents the motion correction mapping.
[0101] when Located in the frequency shaping domain hour, ;when Located in the frequency shaping domain When outside, Set as frequency shaping domain Internal and The nearest combination of inertia support parameters.
[0102] The corrected inertia support parameters are combined and used to update the parameters of the grid-type main photovoltaic virtual synchronous machine controller to obtain the corresponding frequency dynamic response. The response is then compared with the benchmark response of the fixed parameter controller under the same disturbance scenario. Rewards are constructed based on the degree of improvement of indicators such as frequency deviation, frequency change rate, and settling time to update the data-driven strategy.
[0103] After training is completed, the online control phase can generate adjustment amounts for the inertia support parameters based on real-time status information. The adjusted inertia support parameter combination is then corrected in the frequency shaping domain to update the control parameters of the main photovoltaic virtual synchronous machine in the grid.
[0104] For example, for local orthogonal elements in a continuous conservative frequency shaping domain, the Euclidean projection points of the uncorrected parameter combinations to each local orthogonal element can be calculated first, and the closest projection point can be selected. For the remaining discrete parameter combinations, the Euclidean distance between the uncorrected parameter combinations and each remaining discrete parameter combination can be calculated, and the closest discrete parameter combination can be selected. Finally, the candidate correction point closest to the uncorrected parameter combination is selected from the above candidate correction points. It should be noted that the action correction distance is not limited to Euclidean distance; weighted Euclidean distance or a hierarchical correction strategy that meets engineering priorities can also be used according to actual needs. This application does not impose any restrictions on this.
[0105] In some embodiments, the action correction mapping can serve as a constraint layer between the data-driven parameter adaptive adjustment strategy and the grid-type main photovoltaic virtual synchronous machine controller. This constraint layer allows for the filtering and correction of output parameter combinations without altering the basic structure of the upper-layer data-driven parameter adaptive adjustment strategy.
[0106] The following are device embodiments of this application. For details not described in detail in the device embodiments, please refer to the above method embodiments.
[0107] Figure 2 This is a structural block diagram of an adaptive control device for a grid-type main photovoltaic virtual synchronous machine provided in an embodiment of this application. Figure 2 As shown, the adaptive control device 100 of the main photovoltaic virtual synchronous machine in the grid configuration includes a model construction module 101, a parameter identification module 102, a frequency shaping domain construction module 103, and a state acquisition and correction module 104.
[0108] Among them, the model building module 101 is used to build a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points and determine the candidate control parameters to be analyzed.
[0109] The parameter identification module 102 is used to determine the equivalent inertia transfer function based on the grid-type main photovoltaic virtual synchronous machine model, and to identify the inertia support parameters from the candidate control parameters according to the equivalent inertia transfer function. The inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth.
[0110] The frequency shaping domain construction module 103 is used to construct a frequency shaping domain within a parameter space composed of inertia support parameters, based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points.
[0111] The status acquisition and correction module 104 is used to acquire the dynamic status information of the target grid-connected system, generate the adjustment amount of the inertia support parameters based on the data-driven parameter adaptive adjustment strategy, and obtain the adjusted inertia support parameter combination. The adjusted inertia support parameter combination is corrected using the frequency shaping domain to update the control parameters of the grid-connected main photovoltaic virtual synchronous machine.
[0112] It should be noted that the adaptive control device for the grid-type main photovoltaic virtual synchronous machine provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the adaptive control device for the grid-type main photovoltaic virtual synchronous machine provided in the above embodiments and the adaptive control method embodiment for the grid-type main photovoltaic virtual synchronous machine belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.
[0113] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods are possible. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.
[0114] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a terminal device (which may be a personal computer, mobile phone, or communication device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0115] Figure 3 This is a schematic diagram of the structure of the computer device provided in an embodiment of this application. For example... Figure 3 As shown, the computer device 200 includes a processor 201 and a memory 202.
[0116] Processor 201 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 201 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 201 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 201 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the screen. In some embodiments, processor 201 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.
[0117] The memory 202 may include one or more computer-readable storage media, which may be non-transitory. The memory 202 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 202 is used to store at least one instruction, which is executed by the processor 201 to implement the adaptive control method for a grid-connected main photovoltaic virtual synchronous machine provided in the embodiments of this application.
[0118] Those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the computer device 200, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.
[0119] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of a computer device, enables the computer device to execute the adaptive control method of the grid-type main photovoltaic virtual synchronous machine provided in this application.
[0120] This application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the adaptive control method for the grid-type main photovoltaic virtual synchronous machine provided in this application.
[0121] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. An adaptive control method for a grid-connected main photovoltaic virtual synchronous machine, characterized in that, include: Construct a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points, and determine the candidate control parameters to be analyzed; The equivalent inertia transfer function is determined based on the grid-type main photovoltaic virtual synchronous machine model, and the inertia support parameters are identified from the candidate control parameters according to the equivalent inertia transfer function. The inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth. Within the parameter space comprised of the inertia support parameters, a frequency shaping domain is constructed based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points. The dynamic state information of the target grid-connected system is obtained, and the adjustment amount of the inertia support parameter is generated based on the data-driven parameter adaptive adjustment strategy to obtain the adjusted inertia support parameter combination. The adjusted inertia support parameter combination is corrected using the frequency shaping domain to update the control parameters of the grid-connected main photovoltaic virtual synchronous machine.
2. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 1, characterized in that, The grid-type main photovoltaic virtual synchronous machine model is a full-order average value model that includes active-frequency loop, reactive-voltage loop, virtual impedance, voltage inner loop, current inner loop, LC filter, transmission line and grid equivalent links.
3. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 2, characterized in that, In the active-frequency loop, the active power imbalance is first filtered, and then the filtered active power imbalance is input into the swing equation. The active power imbalance before filtering is expressed as: in, This represents the active power imbalance before filtering. Active power command, The actual active power injected into the grid by the main grid-type photovoltaic virtual synchronous machine at the common coupling point; The dynamic relationship of the filtered active power imbalance is expressed as follows: in, This represents the filtered active power imbalance. Indicates time, The active power imbalance filter bandwidth is represented by the cutoff angular frequency of the active power imbalance channel low-pass filter. The expression for the swing equation of the grid-type main photovoltaic virtual synchronous machine model is as follows: in, For virtual inertia parameters, For virtual damping parameters, This refers to the internal angular frequency of the converter. This is the rated angular frequency.
4. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 3, characterized in that, The expression for the equivalent inertia transfer function is: in, The equivalent inertia transfer function, For complex frequency variables, This is the rated power.
5. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 4, characterized in that, The candidate control parameters include virtual inertia parameters, virtual damping parameters, active power imbalance filter bandwidth, voltage inner loop proportional gain, voltage inner loop integral gain, current inner loop proportional gain, current inner loop integral gain, virtual resistance, and virtual reactance. The step of identifying inertia support parameters from the candidate control parameters based on the equivalent inertia transfer function includes: performing sensitivity analysis and closed-loop stability boundary analysis on the candidate control parameters based on the equivalent inertia transfer function, so as to use the virtual inertia parameter, virtual damping parameter, and active power imbalance filter bandwidth as the inertia support parameters.
6. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 1, characterized in that, The construction of a frequency shaping domain within the parameter space defined by the inertia support parameters, based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points, includes: Within the parameter space defined by the inertia support parameters, a linearized model is established for each operating point and each combination of inertia support parameters: in, For the combination of inertia support parameters, , For virtual inertia parameters, For virtual damping parameters, The active power imbalance filter bandwidth. For the running point, For state increment, This represents the increment of external disturbances. State increment The first derivative with respect to time, This is the output increment that includes frequency deviation and active power increment. , , , These represent the combinations of inertia support parameters. At the running point The full-order closed-loop state matrix, disturbance input matrix, output matrix, and direct transfer matrix are given below. Based on the linearized model, the small-signal stability index, frequency response characteristic index, and frequency modulation resource index of each inertia support parameter at each operating point are calculated. The small-signal stability index includes the abscissa of the closed-loop state matrix spectrum. The frequency response characteristic index includes the maximum absolute frequency deviation, the maximum absolute frequency change rate, the steady-state frequency deviation, and the settling time. The frequency modulation resource index includes the peak support power and the support energy. At the same time, small-signal stability constraints, maximum absolute frequency deviation constraints, maximum absolute frequency change rate constraints, steady-state frequency deviation constraints, settling time constraints, peak support power constraints, and support energy constraints are set. In the parameter space, combinations of inertial support parameters that satisfy the small-signal stability constraint, the maximum absolute frequency deviation constraint, the maximum absolute frequency change rate constraint, the steady-state frequency deviation constraint, the settling time constraint, the peak support power constraint, and the support energy constraint at all representative operating points are selected to form the initial discrete frequency shaping domain. The initial discrete frequency shaping domain is subjected to a two-stage discrete scan to obtain the final discrete frequency shaping domain; The final discrete frequency shaping domain is conservatively continuous by using the orthogonal unit reconstruction method based on the full vertex consistency criterion to obtain the conservative continuous frequency shaping domain. The frequency shaping domain is constructed based on the final discrete frequency shaping domain and the conservative continuous frequency shaping domain.
7. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 6, characterized in that, The dynamic state information of the target grid-connected system is represented as follows: in, This refers to the dynamic status information of the target grid-connected system. for Frequency deviation at time, for Rate of change of frequency at time, Indicates the control time step; The adjustment amount of the inertia support parameter is expressed as follows: in, This refers to the adjustment amount of the inertia support parameter. for The adjustment amount of the virtual inertia parameter at any given time. for The adjustment amount of the virtual damping parameter at any given time. for The adjustment amount of the active power imbalance filter bandwidth at all times, and the adjusted inertia support parameter combination are as follows: ; The step of correcting the adjusted inertia support parameter combination using the frequency shaping domain includes: Perform motion correction mapping using the frequency shaping domain: in, The corrected inertia support parameter combination, This refers to the frequency shaping domain. Indicates motion correction mapping; when Located in the frequency shaping domain hour, ;when Located in the frequency shaping domain When outside, Set as frequency shaping domain Internal and The nearest combination of inertia support parameters.
8. The adaptive control method for the grid-type main photovoltaic virtual synchronous machine according to claim 7, characterized in that, The data-driven parameter adaptive adjustment strategy employs a dual-delay deep deterministic gradient algorithm.
9. An adaptive control device for a grid-type main photovoltaic virtual synchronous machine, characterized in that, include: The model building module is used to build a grid-connected main photovoltaic virtual synchronous machine model of the target grid-connected system at multiple operating points and to determine the candidate control parameters to be analyzed. The parameter identification module is used to determine the equivalent inertia transfer function based on the grid-type main photovoltaic virtual synchronous machine model, and to identify the inertia support parameters from the candidate control parameters according to the equivalent inertia transfer function. The inertia support parameters include virtual inertia parameters, virtual damping parameters, and active power imbalance filtering bandwidth. The frequency shaping domain construction module is used to construct a frequency shaping domain within the parameter space composed of the inertia support parameters, based on small-signal stability constraints, frequency response characteristics, and frequency modulation resource constraints at multiple operating points. The status acquisition and correction module is used to acquire the dynamic status information of the target grid-connected system, generate the adjustment amount of the inertia support parameters based on the data-driven parameter adaptive adjustment strategy and obtain the adjusted inertia support parameter combination, and use the frequency shaping domain to correct the adjusted inertia support parameter combination in order to update the control parameters of the grid-type main photovoltaic virtual synchronous machine.
10. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the adaptive control method of the grid-type main photovoltaic virtual synchronous machine as described in any one of claims 1 to 8.
