Three-phase grid-connected inverter current-sensor-free control method based on model identification perception

Abstract

The invention discloses a three-phase grid-connected inverter current sensor-free control method based on model identification perception. And the inverter control link modulating wave and the alternating current side power grid voltage are used as input to reconstruct alternating current side current. According to the implementation method, a current reconstruction function model on the alpha-beta axis is identified through the voltage and current of the filter inductor on the alpha-beta axis of a two-phase static coordinate system, the precision of a current reconstruction function is ensured by adopting an online identification method, and effective observation of the inverter alternating current side current is realized. In addition, in the current feedback control of the grid-connected inverter, the observation current can replace the sensor sampling current, so that the control without the current sensor is realized. According to the method, the application range of a grid-connected inverter current reconstruction model is expanded, the method can be applied to various inductance filtering application scenes, timely and accurate current reconstruction and non-current sensor control are achieved after the sensor is aged or breaks down, and the operation safety and reliability of the grid-connected inverter are improved.

Description

Technical Field

[0001] This invention belongs to the field of grid-connected inverter current closed-loop control, specifically a sensorless control method for three-phase grid-connected inverters based on model identification and sensing. Background Technology

[0002] As a crucial interface device between renewable energy generation and the power grid, the continuous and reliable operation of grid-connected inverters is paramount. Inverters require sampling the AC-side inductor current to complete dual closed-loop control. Because inverters operate under harsh conditions of electrothermal stress and strong electromagnetic interference, current sensor failure can easily lead to system collapse. Therefore, current sensor fault diagnosis and fault-tolerant control have significant research and application value.

[0003] Current sensor failures can be categorized into jamming, gain failure, and offset failure. In traditional balanced three-phase grid-connected systems, after a single current sensor failure, the faulty phase current can be calculated from the currents of the remaining two phases, thus achieving fault-tolerant control. However, when faced with failures of two or more sensors, existing methods have improved system reliability by increasing the number of current sensors and utilizing hardware backups to achieve redundant control under fault conditions. However, hardware backups increase system complexity and cost. Therefore, sensorless control offers significant advantages.

[0004] Previous research and applications of sensorless inverter control have primarily focused on fault current reconstruction, current loop-free control, and predictive current control. Predictive control methods offer advantages such as fast response and strong interference suppression, but the observer setup process is complex and requires high model accuracy. Current loop-free control eliminates the current control loop, thus eliminating the need for a current sensor, but its dynamic performance is poor. Fault current reconstruction methods can be implemented using sliding mode observers to reconstruct the fault current and dq-axis current functions. Sliding mode observer current reconstruction requires a high-quality model, while dq-axis current reconstruction is complex due to dq-axis coupling and requires high accuracy of filter parameters, limiting its application.

[0005] The key to improving the observation accuracy of the output current of a current sensor is to obtain a current observation function that is simple, easy to derive, and more accurate. Summary of the Invention

[0006] The voltage and current along the αβ axis of the inductor filter are directly related. Therefore, by using the voltage along the αβ axis of the filter inductor and combining it with the inductor model, the αβ axis current can be obtained. The dq axis current can then be obtained through Park transformation. The reconstruction function model for the αβ axis current along the inductor satisfies the following formula:

[0007]

[0008] L1 is the AC side filter inductor, r1 is the parasitic resistance of the filter inductor, and i gα i gβ It is the αβ axis component of the inductor current, u tα u tβ It is the αβ axis component of the inverter's AC side output voltage, u gα u gβ α is the component of the grid voltage along the αβ axis, and s is the differential operator.

[0009] Based on the αβ-axis voltage and inductance model, the αβ-axis current of the inductor can be obtained. After being converted to the dq-axis current through Park transformation, it is used for current closed-loop feedback control. Inverter output voltage u tα u tβ Containing a large number of switching frequency harmonics, the control loop modulates the waveform with the inverter's DC voltage, and u is observed. Nα u Nβ Voltage within the inverter's control bandwidth range.

[0010]

[0011] In the formula: u tαref with u tβref These are the modulated waves along the α and β axes generated in the inverter control loop, respectively. The abc axis modulated waves are obtained from the control loop via Clark transform. dc This is the DC side voltage of the inverter.

[0012] Due to the influence of ambient temperature and the magnitude of the current flowing through it, the actual parameters of the filter inductor L will change. Continuing to use the current reconstruction function model with a fixed inductor model after these parameter changes will significantly affect the accuracy of the reconstructed current. Therefore, the previously used current reconstruction function model is updated and replaced with the identified observation function model.

[0013] Using the inductor voltage and current along the αβ axis as input signals, the model order was determined by solving the current reconstruction function using the Levenberg-Marquardt (LM) algorithm. The current identification transfer function model used is as follows:

[0014]

[0015] The current reconstruction function model of the previous fixed inductance model was replaced by an identification transfer function model with a good fit of greater than 85%.

[0016]

[0017] Based on the above formula, the αβ axis reconstructed current is obtained. The αβ axis reconstructed current is then transformed by Park to obtain the dq axis reconstructed current, which replaces the sampling current. This enables sensorless fault-tolerant control and identifies the current reconstructed by the transfer function to eliminate errors caused by inductor aging. Attached Figure Description

[0018] To clearly demonstrate the technical solutions of the embodiments of the present invention and the prior art, the drawings used in the embodiments will be introduced below. The drawings described below are embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the grid-connected inverter system and control structure of the present invention;

[0020] Figure 2 This is a block diagram of the current closed-loop feedback control used in the grid-connected inverter system of this invention;

[0021] Figure 3 This is the implementation process of the current reconstruction function system identification in this invention;

[0022] Figure 4 This is a block diagram of the current closed-loop control of the present invention, which uses the αβ axis to identify the reconstructed current;

[0023] Figure 5 These are simulation waveforms before and after the replacement of the identification and reconstruction function in this invention;

[0024] Figure 6 These are simulation waveforms before and after the current sensorless control switching of this invention. Detailed Implementation

[0025] The purpose of this invention is to provide a sensorless control method for three-phase grid-connected inverters based on model recognition and perception, in order to solve the problems of low accuracy and adaptability of the current reconstruction function, and to achieve accurate current reconstruction and sensorless closed-loop control.

[0026] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Figure 1 This is a structural block diagram of the grid-connected inverter system of the present invention, as shown below. Figure 1 As shown, it includes a DC / AC grid-connected inverter and an AC power grid, U dc The voltage source voltage is L1, and the AC side filter inductors and their parasitic resistances are R1 and R2, respectively. The voltage inverter uses a three-phase bridge inverter circuit composed of six IGBTs. g The inductance is the equivalent impedance of the power grid, ei (i = a / b / c) is the grid-side voltage, i gx u gx (x = a / b / c) represent the measured three-phase current and voltage, respectively.

[0028] Figure 2 This is a block diagram of the current feedback control used in the grid-connected inverter system of this invention, as shown below. Figure 2 As shown, i dref and i qref These are the dq-axis components of the grid-connected current setpoint, i d and i q These are the dq-axis components of the actual grid-connected current, u gd and u gq These are the dq-axis components of the grid-connected voltage, u tdref and u tqref These are the dq-axis components of the inverter output voltage reference value, respectively. To achieve unity power factor grid connection, i... qref Set it to 0.

[0029] Figure 3 The current reconstruction function system identification process of this invention uses the inductor voltage and inductor current on the αβ axis as identification input and output data, confirms the model order, and obtains the current reconstruction function by solving the Levenberg-Marquardt (LM) algorithm.

[0030] Figure 4 This is an equivalent block diagram of the sensorless current closed-loop control system for reconstructing current using the αβ axis, as described in this invention. In the diagram: G P (s) is the equivalent transfer function of the PI element; G d (s) represents the PWM modulation and digital control delay stage; G L (s) is the equivalent transfer function of the AC-side filter inductor; G αest (s), G βest (s) is the αβ-axis inductor current identification and reconstruction function; i dest i qest These are the reconstructed currents along the d-axis and q-axis, respectively.

[0031] The voltage and current along the αβ axis of the inductor filter are directly related. Therefore, by using the voltage along the αβ axis of the filter inductor and combining it with the inductor model, the αβ axis current can be obtained. The dq axis current can then be obtained through Park transformation. The reconstruction function model for the αβ axis current along the inductor satisfies the following formula:

[0032]

[0033] L1 is the AC side filter inductor, r1 is the parasitic resistance of the filter inductor, and i gα i gβIt is the αβ axis component of the inductor current, u tα u tβ It is the αβ axis component of the inverter's AC side output voltage, u gα u gβ α is the component of the grid voltage along the αβ axis, and s is the differential operator.

[0034] Based on the αβ-axis voltage and inductance model, the αβ-axis current of the inductor can be obtained. After being converted to the dq-axis current through Park transformation, it is used for current closed-loop feedback control. Inverter output voltage u tα u tβ Containing a large number of switching frequency harmonics, the control loop modulates the waveform with the inverter's DC voltage, and u is observed. Nα u Nβ Voltage within the inverter's control bandwidth range.

[0035]

[0036] In the formula: u tαref with u tβref These are the modulated waves along the α and β axes generated in the inverter control loop, respectively. The abc axis modulated waves are obtained from the control loop via Clark transform. dc This is the DC side voltage of the inverter.

[0037] Due to the influence of ambient temperature and the magnitude of the current flowing through it, the actual parameters of the filter inductor L will change. Continuing to use the current reconstruction function model with a fixed inductor model after these parameter changes will significantly affect the accuracy of the reconstructed current. Therefore, the previously used current reconstruction function model is updated and replaced with the identified observation function model.

[0038] Using the inductor voltage and current along the αβ axis as input signals, the model order was determined by solving the current reconstruction function using the Levenberg-Marquardt (LM) algorithm. The current identification transfer function model used is as follows:

[0039]

[0040] The current reconstruction function model of the previous fixed inductance model was replaced by an identification transfer function model with a good fit of greater than 85%.

[0041]

[0042] Based on the above formula, the αβ axis reconstructed current is obtained. The αβ axis reconstructed current is then transformed by Park to obtain the dq axis reconstructed current, which replaces the sampling current. This enables sensorless fault-tolerant control and identifies the current reconstructed by the transfer function to eliminate errors caused by inductor aging.

[0043] Figure 5These are simulation waveforms before and after replacing the identification and reconstruction function according to the present invention. The reconstructed current obtained using the identification and reconstruction function has a small steady-state error and good dynamic performance, and can quickly track the response of the current step, thus proving the effectiveness and accuracy of the identification and reconstruction current.

[0044] Figure 6 This is a simulation waveform diagram before and after the sensorless control switching of this invention. At 0.25s, the d-axis current reference value decreases from 50A to 25A. At 0.5s, the switch to sensorless control occurs, and at 0.75s, the d-axis current reference value jumps from 25A to 50A. Analysis of the three-phase current waveforms and dq-axis currents before and after the sensorless control switch shows no significant change in either the three-phase current or the dq-axis current. This demonstrates the effectiveness and accuracy of the sensorless inverter control proposed in this invention. In summary, the above analysis shows that when using the model recognition-based sensing-based sensorless control method for three-phase grid-connected inverters proposed in this invention, the safe and reliable operation of the system can be guaranteed even when all current sensors are unavailable.

Claims

1. A model-based identification perception-based current sensorless control method for a three-phase grid-connected inverter, applied to a current closed-loop controlled grid-connected inverter, the grid-connected inverter system comprising: A voltage source converter (VSC), a DC line and a DC side voltage stabilizing capacitor, an AC line, an AC side filter inductor and a current closed-loop feedback control method, characterized in that: Step 1, obtain the two-phase rotating coordinate system dq axis modulation wave through the inverter controller output, perform park inverse transformation to obtain the two-phase static coordinate system αβ axis modulation wave, multiply the modulation wave and the DC voltage to obtain the inverter side voltage, perform clark transformation on the collected grid side voltage to obtain the two-phase static coordinate system αβ axis voltage, and then obtain the αβ axis inductance voltage drop. Step 2, perform clark transformation on the collected grid side current to obtain the two-phase static coordinate system αβ axis current, use the αβ axis voltage drop on the inductor and the current as the input of model identification, and identify the inverter AC side filter inductor current reconstruction function by using Levenberg-Marquardt (LM) algorithm. Step 3, use the αβ axis voltage drop on the inductor as the input of the current reconstruction function, and then obtain the two-phase static coordinate system αβ axis observed current, perform park transformation to obtain the two-phase static coordinate system dq axis observed current. Step 4, subtract the dq axis observed current from the given value, output the modulation wave through the PI regulator, so that the actual inductor current follows the given value, and realize zero-error regulation.

2. The three-phase grid-connected inverter current sensorless control method based on model identification perception according to claim 1, characterized in that: The voltage and current on the αβ axis of the inductor filter are directly related, so the αβ axis current can be obtained by using the voltage on the filter inductor and combining the inductor model, and the dq axis current can be obtained by park transformation. The αβ axis current reconstruction function model satisfies the following formula: L1 is the AC side filter inductor, r1 is the parasitic resistance of the filter inductor, and i gα i gβ It is the αβ axis component of the inductor current, u tα u tβ It is the αβ axis component of the inverter's AC side output voltage, u gα u gβ α is the component of the grid voltage along the αβ axis, and s is the differential operator. The inductance αβ axis current can be obtained based on the inductance αβ axis voltage and inductance model, and is converted into dq axis current by park transformation and used for current closed loop feedback control. The inverter output voltage u tα , u tβ contains a large number of switching frequency harmonics, so the control link modulation wave and inverter DC voltage are used to observe u Nα , u Nβ voltage in the inverter control bandwidth range. In the formula: u tαref u tβref u dc is the modulation wave under the αβ axis generated in the inverter control link, the abc axis modulation wave is obtained from the control link through the Clark transformation, u dc is the inverter DC side voltage. Due to the influence of environmental temperature and circulating current size, the actual parameters of the filter inductor L will change. After the parameters change, the current reconstruction function model using the fixed inductor model will significantly affect the accuracy of the reconstructed current. Therefore, the observed function model obtained by identification is used to update and replace the previous current reconstruction function model. The inductor voltage and current on the αβ axis are used as input signals, and the input model order is solved by Levenberg-Marquardt (LM) algorithm to obtain the current reconstruction function. The current identification transfer function model used is as follows: Select the identification transfer function model with a fitting degree greater than 85% to replace the current reconstruction function model of the previous fixed inductor model. According to the above formula, the αβ axis reconstructed current is obtained, and the dq axis reconstructed current is obtained by park transformation of the αβ axis reconstructed current.

Images