A grid-connected inverter harmonic suppression model-free robust control method

Abstract

The application provides a grid-connected inverter harmonic suppression model-free robust control method, and belongs to the technical field of photovoltaic grid control. The application comprises the following steps: establishing an LCL inverter hyperlocal model; designing an adaptive gain expansion state observer for estimating grid-side current and lumped disturbance items in the grid-connected inverter hyperlocal model; designing a double second-order generalized integrator for filtering a sampling current and extracting a pure fundamental positive sequence component; and using the current and disturbance information predicted by the observer, calculating a reference voltage based on the zero-error principle, and driving the inverter through space vector modulation. The application only needs to sample the grid-side current, can reduce the hardware cost and dependence on an accurate model, and thus can significantly improve the control accuracy, dynamic response and robustness of the system under parameter perturbation and grid disturbance.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic grid control technology and relates to a model-free robust control method for harmonic suppression of grid-connected inverters. Background Technology

[0002] Inductor-capacitor-inductor (LCL) grid-connected inverters have become one of the core devices in distributed generation systems due to their low switching losses and high harmonic suppression capabilities. With the rapid development of new energy power grids and power electronics technology, the control requirements for grid-connected inverters are increasing rapidly. Traditional dual-loop proportional-integral (PI) control methods are no longer sufficient to meet the requirements of fast response, high precision, and strong robustness.

[0003] In addition, the industry currently generally adopts two sampling schemes: sampling the inverter side current and using it as the control target, or sampling the grid side current and capacitor branch voltage simultaneously with the grid side current as the control target. The disadvantage of the former is that the steady-state accuracy of the grid side current is only guaranteed by the LCL filter, while the disadvantage of the latter is that it increases the hardware cost due to the acquisition of two electrical quantities.

[0004] Model-Free Deadbeat Current Predictive Control (MF-DPCC) treats all unknown or difficult-to-model components, such as capacitor branch voltage, LCL higher-order dynamics, and electrical parameter errors, as lumped disturbances and estimates these lumped disturbances online. It only needs to sample the grid-side current for control, significantly reducing its dependence on parameters and sampling sensors. Currently, mainstream methods for estimating lumped disturbance terms in hyperlocal models include observer-based, least squares, and data-driven methods. Although these methods can achieve effective control, their drawbacks include complex design and high computational load, making them difficult to apply to LCL grid-connected inverter control.

[0005] In addition, using an Extended State Observer (ESO) to estimate the unknown parts in the hyperlocal model and compensate for them to design the controller has the advantages of low computational cost and simple results. However, ESO has the phenomenon of initial differential peaks, and a fixed ESO gain cannot adapt to LCL resonance phenomena. The contradiction is that if the disturbance caused by resonance is to be included in the lumped disturbance, the bandwidth needs to be higher than the resonance frequency. However, an excessively large bandwidth will lead to insufficient ESO phase margin. If an additional active damping element is added to handle the resonance phenomenon, it will increase the hardware computational burden.

[0006] Therefore, there is an urgent need for a model-free control method for LCL grid-connected inverters that can adaptively adjust the gain to balance tracking performance and noise immunity, and effectively handle current harmonic interference. Summary of the Invention

[0007] This invention provides a model-free robust control method for harmonic suppression in grid-connected inverters, which solves the problems of insufficient dynamic response, weak environmental adaptability, and reliance on experience or offline optimization for parameter tuning in existing photovoltaic power generation MPPT control technology, as well as the poor online adjustment capability and instability of traditional optimization algorithms.

[0008] Based on the above technical concept, the present invention adopts the following technical solution:

[0009] A model-free robust control method for harmonic suppression in a grid-connected inverter includes the following steps:

[0010] Step 1: Establish a hyperlocal model of the LCL-type grid-connected inverter.

[0011] The hyperlocal model of the LCL grid-connected inverter in the dq rotating coordinate system is expressed as follows:

[0012]

[0013] In the formula, for The derivative with respect to time; This refers to the current on the d-axis grid side. Input gain parameter; This refers to the output voltage of the d-axis inverter. This is the lumped disturbance term along the d-axis; for The derivative with respect to time; This refers to the q-axis network-side current. This refers to the output voltage of the q-axis inverter. This is the lumped disturbance term along the q-axis.

[0014] The aforementioned LCL-type grid-connected inverter hyperlocal model reduces all high-order dynamics, parameter uncertainties, and grid voltage disturbances of the grid-connected inverter system to lumped disturbance terms, thereby avoiding dependence on precise physical parameters.

[0015] Step 2: Design an Adaptive Gain Extended State Observer (AGESO) to estimate the grid-side current and lumped disturbance terms in the hyperlocal model of the grid-connected inverter.

[0016] The expression for AGESO on the q-axis is:

[0017]

[0018] In the formula, for The derivative with respect to time; for The estimated value; As the first auxiliary variable; The first parameter; for The estimated value; As the second auxiliary variable; For time; For time-varying gain; The first gain; for The derivative with respect to time; This is the second gain; , Both are adaptive and time-varying gains, and can be automatically adjusted according to different operating conditions;

[0019] in:

[0020]

[0021]

[0022] In the formula, For the second parameter, The third parameter is a positive number.

[0023] The adaptive laws for the first and second gains in the AGESO are designed, and their expressions are as follows:

[0024]

[0025] In the formula, The time-varying adaptive bandwidth is used to reduce the number of parameters to be tuned in AGESO from two to one. To save computational resources, it is updated every 100 current sampling cycles. hour, Fixed at a constant In order to eliminate the initial differential peak; hour, It automatically adjusts between the set upper and lower limits based on different situations faced by the current loop.

[0026] in:

[0027]

[0028] In the formula, This represents the maximum bandwidth for AGESO. This is the lower limit of AGESO's bandwidth. It is the fourth parameter; This is an estimate of the q-axis current on the grid side.

[0029] The AGESO is discretized using the first-order forward Euler discretization method to obtain the th... The q-axis stator current at time t is estimated to be the same as the hyperlocal model parameter estimate.

[0030]

[0031] In the formula, For the first Predicted output of the q-axis network current at any given time; For the first Moment The sampled values; To control the cycle; For the first The inverter voltage that constantly changes to the q-axis; For the first Moment The sampled values; For the discretized first The time-varying observer gain at any given moment; For the first Moment The sampled values; For the first Moment The sampled values; For the first Always The estimated value; For the first Moment The sampled values.

[0032] Step 3: Design a dual second-order generalized integrator (DSOGI) to extract the pure fundamental positive-sequence current from the sampled grid-side current containing harmonics, and use the extracted current as the input of the adaptive gain extended state observer.

[0033] The expression for DSOGI is:

[0034]

[0035] In the formula, To extract Axis-pure positive-sequence components; This is the α-axis output signal for DSOGI; The quadrature of the output signal along the β axis of DSOGI; To extract Axial pure positive-sequence components; This is the output signal of the β axis of DSOGI; Let be the quadrature of the output signal along the α axis of DSOGI, which is related to... Lagging behind , Angle 90°;

[0036] in:

[0037]

[0038] In the formula, for The derivative with respect to time; The fundamental angular frequency; The resonant coefficient; The sampled current of the power grid is transformed to the α-axis; for The derivative with respect to time; for The derivative with respect to time; The sampled current of the power grid is transformed to the β-axis; for The derivative with respect to time.

[0039] DSOGI is not started within the first 10ms of the controller's operation, pending stable tracking by the phase-locked loop (PLL). Then start DSOGI.

[0040] The DSOGI designed in this invention has the following functions:

[0041] 1. The DSOGI bandpass filter of this invention, taking the q-axis as an example, when a fixed value is taken... Bode plots showing the filtering effects of signals at different multiples of frequency after passing through DSOGI; regardless of... Whether the value is large or small, it can achieve the effect of moving away from the frequency. The amplitude decreases rapidly, and SOGI can decay far away. High-frequency harmonics;

[0042] 2. The present invention, DSOGI, extracts the orthogonal components of the fundamental frequency and uses these orthogonal components to extract a pure positive-sequence component. and By using specific linear combinations, it is possible to construct operations that enhance positive-order components and cancel negative-order components, thereby extracting pure positive-order components from the signal.

[0043] When the sampled abc network-side three-phase current i ga i gb i gc After Clark transformation, the two-phase currents i on the αβ axis grid side are obtained. gα with i gβ Before performing the Park transform to obtain the dq-axis network current, the pure fundamental positive-sequence current i is extracted using DSOGI. fα with i fβ Then use ifα with i fβ Replace i gα with i gβ A Park transformation is performed; thus, DSOGI can provide AGESO with a high-quality data source, making the latter unaffected by current harmonics, and the final output of DSOGI is the pure fundamental positive-sequence current required for MF-PFCC, which is then filtered and purified. and Transforming to the dq axis will yield the i required for AGESO and MF-DPCC. gd and i gq .

[0044] Step 4: Using the grid-side current prediction value and lumped disturbance term prediction value obtained from AGESO observations, calculate the inverter-side reference voltage using MF-DPCC.

[0045] The expression for calculating the inverter-side reference voltage using the MF-DPCC is as follows:

[0046]

[0047] In the formula, This is the inverter-side reference voltage on the d-axis; This is the reference current on the d-axis grid side; The d-axis current at time (k+1) is obtained from the AGESO prediction of the d-axis. The sampling period; The d-axis hyperlocal model lumped perturbation parameters predicted by AGESO for the d-axis; This is the inverter-side reference voltage for the q-axis; This is the reference current on the q-axis network side.

[0048] Step 5: Generate a switching signal using space vector modulation technology to drive the grid-connected inverter, causing the inverter to output an inverter-side reference voltage; specifically:

[0049] In obtaining , Subsequently, space vector modulation technology decomposes the inverter's output voltage vector into a combination of basic voltage vectors. The inverter's switching state determines the direction and magnitude of the basic voltage vectors. By appropriately selecting the combination and timing of the basic voltage vectors, the desired voltage vector is synthesized. and This controls the grid-side current.

[0050] The beneficial effects of this invention are as follows:

[0051] This invention can adaptively adjust the observer gain, effectively suppress initial peak values, improve the tracking ability and noise immunity of lumped disturbances, and simultaneously filter out harmonic interference through current preprocessing, thereby enhancing the robustness and control accuracy of the entire control system under parameter perturbations and grid disturbances; specifically:

[0052] 1. This invention addresses the problem that traditional predictive control requires sampling two state variables in a grid-connected system or using inverter-side current as an indirect control target. The MF-DPCC only needs to sample grid-side current to directly complete the control, which can save sensor hardware resources while ensuring control accuracy.

[0053] 2. In view of the problems of traditional ESO, such as initial differential peak, difficulty in balancing tracking performance and noise immunity, and difficulty in dealing with LCL filter resonance, this invention designs an adaptive gain AGESO to replace the conventional ESO.

[0054] 3. To address the issue that harmonics and negative sequence current in the sampled current can affect ESO performance, this invention designs a dual second-order generalized integrator to clean the sampled current and extract the pure fundamental positive sequence current, which can effectively reduce the impact of harmonics on the estimation effect. In addition, the pure current signal allows AGESO to increase the maximum bandwidth without excessively amplifying noise, and the higher the bandwidth, the better AGESO can include resonance in the lumped disturbance. Therefore, DSOGI indirectly improves the ability to suppress resonance phenomena. Attached Figure Description

[0055] Figure 1 This is a structural block diagram of the dual second-order generalized integrator in an embodiment of the present invention;

[0056] Figure 2 This is a Bode plot of the filtering effect of the dual second-order generalized integrator under different resonance coefficients in an embodiment of the present invention;

[0057] Figure 3 This is a block diagram illustrating the control principle of an embodiment of the present invention;

[0058] Figure 4 This is a control flowchart of an embodiment of the present invention;

[0059] Figure 5 This is a waveform diagram of the grid-side phase a current and voltage under steady-state conditions using this embodiment.

[0060] Figure 6 This is a comparison of the grid-side current waveforms of three different control systems under the reference current sudden change condition.

[0061] Figure 7 yes Figure 6 A magnified view of a portion of the image;

[0062] Figure 8 This is a comparison of the grid-side current waveforms of three different control systems under grid voltage fluctuation conditions.

[0063] Figure 9 yes Figure 8 A magnified view of a portion of the image;

[0064] Figure 10 This is a comparison of the grid-side current waveforms of three different control systems under filter parameter perturbation conditions.

[0065] Figure 11 yes Figure 10 A magnified view of a portion of the image;

[0066] Figure 12 This is a comparison diagram of the experimental waveforms of the grid-side a-phase current under steady-state conditions between a traditional quasi-proportional resonant controller and the embodiment of this invention.

[0067] Figure 13 This is a comparison diagram of the experimental waveforms of the grid-side a-phase current under the reference current sudden change condition, between the traditional quasi-proportional resonant controller and the embodiment of the present invention. Detailed Implementation

[0068] The technical solution of the present invention will be described below with reference to the accompanying drawings and implementation methods.

[0069] Example

[0070] This embodiment provides a model-free robust control method for harmonic suppression in grid-connected inverters, including the following steps:

[0071] Step 1: Establish the hyperlocal model of the LCL-type grid-connected inverter. Its expression in the dq rotating coordinate system is as follows:

[0072]

[0073] In the formula, for The derivative with respect to time; This refers to the current on the d-axis grid side. Input gain parameter; This refers to the output voltage of the d-axis inverter. This is the lumped disturbance term along the d-axis; for The derivative with respect to time; This refers to the q-axis network-side current. This refers to the output voltage of the q-axis inverter. This is the lumped disturbance term along the q-axis.

[0074] Step 2: Design an adaptive gain extended state observer to estimate the grid-side current and lumped disturbance terms in the hyperlocal model of the grid-connected inverter.

[0075] The expression for the adaptive gain extended state observer on the q-axis is:

[0076]

[0077] In the formula, for The derivative with respect to time; for The estimated value; As the first auxiliary variable; The first parameter; for The estimated value; As the second auxiliary variable; For time; For time-varying gain; The first gain; for The derivative with respect to time; This is the second gain; , Both are adaptive and time-varying gains, and can be automatically adjusted according to different operating conditions;

[0078] in:

[0079]

[0080]

[0081] In the formula, For the second parameter, The third parameter is a positive number.

[0082] The adaptive laws for the first and second gains in the adaptive gain extended state observer are designed as follows:

[0083]

[0084] In the formula, The time-varying adaptive bandwidth is used to reduce the number of parameters to be tuned in AGESO from two to one. To save computational resources, it is updated every 100 current sampling cycles. hour, Fixed at a constant In order to eliminate the initial differential peak; hour, It automatically adjusts between the set upper and lower limits based on different situations faced by the current loop.

[0085] in:

[0086]

[0087] In the formula, This represents the maximum bandwidth for AGESO. This is the lower limit of AGESO's bandwidth. It is the fourth parameter; This is an estimate of the q-axis current on the grid side.

[0088] The adaptive gain extended state observer is discretized using the first-order forward Euler discretization method to obtain the... The q-axis stator current at time t is estimated to be the same as the hyperlocal model parameter estimate.

[0089]

[0090] In the formula, For the first Predicted output of the q-axis network current at any given time; For the first Moment The sampled values; To control the cycle; For the first The inverter voltage that constantly changes to the q-axis; For the first Moment The sampled values; For the discretized first The time-varying observer gain at any given moment; For the first Moment The sampled values; For the first Moment The sampled values; For the first Always The estimated value; For the first Moment The sampled values.

[0091] Step 3, as follows Figure 1 and 2 As shown, a dual second-order generalized integrator is designed to extract the pure fundamental positive-sequence current from the sampled grid-side current containing harmonics, and the extracted current is used as the input of an adaptive gain extended state observer.

[0092] The expression for the dual second-order generalized integrator is:

[0093]

[0094] In the formula, To extract Axis-pure positive-sequence components; This is the α-axis output signal for DSOGI; The quadrature of the output signal along the β axis of DSOGI; To extract Axial pure positive-sequence components; This is the output signal of the β axis of DSOGI; This is the quadrature of the output signal along the α axis of DSOGI, and it is related to... Lagging behind , Angle 90°;

[0095] in:

[0096]

[0097] In the formula, for The derivative with respect to time; The fundamental angular frequency; The resonant coefficient; The sampled current of the power grid is transformed to the α-axis; for The derivative with respect to time; for The derivative with respect to time; The sampled current of the power grid is transformed to the β-axis; for The derivative with respect to time.

[0098] DSOGI is not started within the first 10ms of the controller's operation, pending stable tracking by the phase-locked loop (PLL). Then start DSOGI.

[0099] Step 4: Using the grid-side current prediction value and lumped disturbance term prediction value obtained by the adaptive gain extended state observer, calculate the inverter-side reference voltage using the model-free, deadbeat-free current prediction method.

[0100] The expression for calculating the inverter-side reference voltage using the model-free, deadbeat-free current prediction method is as follows:

[0101]

[0102] In the formula, This is the inverter-side reference voltage on the d-axis; This is the reference current on the d-axis grid side; The d-axis current at time (k+1) is obtained from the AGESO prediction of the d-axis. The sampling period; The d-axis hyperlocal model lumped perturbation parameters predicted by AGESO for the d-axis; This is the inverter-side reference voltage for the q-axis; This is the reference current on the q-axis network side.

[0103] Step 5: Generate a switching signal using space vector modulation technology to drive the grid-connected inverter, so that the inverter outputs the inverter-side reference voltage.

[0104] The control system in this embodiment is as follows: Figure 3 As shown, the process is as follows: Figure 4 As shown.

[0105] To verify the effectiveness of the strategy proposed in this embodiment, three control systems were built for the current control of the LCL filter grid-connected inverter system: 1. A conventional quasi-PR controller, referred to as System 1 below; 2. An MF-DPCC based on conventional ESO without DSIGO stage, referred to as System 2 below; 3. An MF-DPCC based on AGESO proposed in this embodiment, referred to as System 3 below. Table 1 shows the main simulation parameters of the three controllers.

[0106] Table 1. Main simulation parameters of the three controllers

[0107]

[0108] This embodiment uses a three-phase LCL grid-connected inverter as the controlled object, and its system parameters are shown in Table 2.

[0109] Table 2 Simulation Parameter Settings

[0110]

[0111] The simulation results are analyzed as follows:

[0112] Depend on Figure 5 It can be seen that under steady state, the grid-side current of system 3 has good sinusoidal properties and is in phase with the grid voltage. The total harmonic distortion rate is only 0.66%, which is far lower than the grid connection standard of 5%.

[0113] Depend on Figure 6 and 7 It can be seen that when the current reference value jumps from 45A to 60A, the response speed of System 2 and System 3 is significantly faster than that of System 1, and the overshoot is small and the recovery time is short; the total harmonic distortion (THD) of System 3 remains the lowest in all stages.

[0114] Depend on Figure 8 and 9 It can be seen that when the grid voltage drops suddenly, the current of System 2 and System 3 hardly fluctuates and quickly recovers to a steady state, demonstrating a stronger grid voltage feedforward compensation capability.

[0115] Depend on Figure 10 and 11 It can be seen that when the LCL filter parameters change abruptly, the current waveforms of System 2 and System 3 can still remain stable with very small THD changes, while the waveform distortion and THD of System 1 increase significantly, thus proving the parameter robustness of the model-free control method.

[0116] HIL Experimental Results Analysis:

[0117] Depend on Figure 12 It can be seen that the steady-state current waveform of system 3 is smoother, has a higher sinusoidal degree, and a shorter settling time than that of system 1.

[0118] Depend on Figure 13 It can be seen that when the reference current changes abruptly, System 3 has a faster tracking speed, a smoother transition process, and a smaller overshoot, which once again verifies its superior dynamic performance.

[0119] In summary, the method of this invention, through gain adaptation and current preprocessing, can effectively solve the problems of strong model dependence, contradiction between anti-disturbance and anti-noise, and harmonic interference in traditional methods. While ensuring hardware simplicity, it significantly improves the dynamic and static performance and robustness of the system under various operating conditions.

Claims

1. A model-free robust control method for harmonic suppression in a grid-connected inverter, characterized in that, Includes the following steps: Establish a hyperlocal model of the LCL-type grid-connected inverter; Design an adaptive gain extended state observer to estimate the grid-side current and lumped disturbance terms in the hyperlocal model of a grid-connected inverter; Design a dual second-order generalized integrator to extract the pure fundamental positive-sequence current from the sampled grid-side current containing harmonics, and use the extracted current as the input of an adaptive gain extended state observer. The inverter-side reference voltage is calculated using the grid-side current prediction value and the lumped disturbance term prediction value obtained by the adaptive gain extended state observer and the model-free, deadbeat-free current prediction method. The expression for the inverter-side reference voltage is: ； In the formula, This is the inverter-side reference voltage on the d-axis; This is the reference current on the d-axis grid side; The d-axis current at time (k+1) is obtained from the AGESO prediction of the d-axis. The sampling period; The d-axis hyperlocal model lumped perturbation parameters predicted by AGESO for the d-axis; This is the inverter-side reference voltage for the q-axis; This is the reference current on the q-axis grid side; Input gain parameter; A switching signal is generated using space vector modulation technology to drive the grid-connected inverter, enabling the inverter to output a reference voltage on the inverter side.

2. The model-free robust control method for harmonic suppression of a grid-connected inverter according to claim 1, characterized in that, The hyperlocal model of the LCL grid-connected inverter in the dq rotating coordinate system is expressed as follows: ； In the formula, for The derivative with respect to time; This refers to the current on the d-axis grid side. Input gain parameter; This refers to the output voltage of the d-axis inverter. This is the lumped disturbance term along the d-axis; for The derivative with respect to time; This refers to the q-axis network-side current. This refers to the output voltage of the q-axis inverter. This is the lumped disturbance term along the q-axis.

3. The model-free robust control method for harmonic suppression of a grid-connected inverter according to claim 1, characterized in that, The expression for the adaptive gain extended state observer on the q-axis is: ； In the formula, for The derivative with respect to time; for The estimated value; As the first auxiliary variable; The first parameter; for The estimated value; As the second auxiliary variable; For time; For time-varying gain; The first gain; for The derivative with respect to time; This is the second gain; in: ； ； In the formula, The second parameter; This is the third parameter.

4. The model-free robust control method for harmonic suppression of a grid-connected inverter according to claim 3, characterized in that, The adaptive laws for the first and second gains in the adaptive gain extended state observer are designed as follows: ； In the formula, Time-varying adaptive bandwidth; in: ； In the formula, This is the upper limit of the bandwidth for the adaptive gain-expanded state observer; This is the lower bandwidth limit for the adaptive gain-expanded state observer; This is the fourth parameter, used to adjust the impact of errors on bandwidth; To The estimated value; The adaptive gain extended state observer is discretized using the first-order forward Euler discretization method to obtain the... The q-axis stator current at time t is estimated to be the same as the hyperlocal model parameter estimate. ； In the formula, For the first Predicted output of the q-axis network current at any given time; For the first Moment The sampled values; To control the cycle; For the first The inverter voltage that constantly changes to the q-axis; For the first Moment The sampled values; For the discretized first The time-varying observer gain at any given moment; For the first Moment The sampled values; For the first Moment The sampled values; For the first Always The estimated value; For the first Moment The sampled values.

5. The model-free robust control method for harmonic suppression of a grid-connected inverter according to claim 1, characterized in that, The expression for the dual second-order generalized integrator is: ； In the formula, To extract Axial pure positive-sequence components; For a dual second-order generalized integrator Axis output signal; For a dual second-order generalized integrator The quadrature of the axis output signal; To extract Axial pure positive-sequence components; For a dual second-order generalized integrator The output signal of the shaft; For a dual second-order generalized integrator The quadrature of the output signals of the axis; in: ； In the formula, for The derivative with respect to time; The fundamental angular frequency; The resonant coefficient; To transform the grid sampling current to the α axis; for The derivative with respect to time; for The derivative with respect to time; The grid sampling current is transformed to the β axis; for The derivative with respect to time.

Images