A method, system and medium for online identification and predictive control of a two-level inverter
By reconstructing the inverter output voltage and identifying the filter inductor and equivalent resistance online, the prediction model is updated, solving the prediction model deviation problem caused by the dynamic changes of filter parameters in two-level grid-connected inverters, realizing adaptive predictive control, and improving control accuracy.
Patent Information
- Application Number
- CN202611455319.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-18
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies have failed to effectively address the parameter deviation problem in model predictive control when the filter inductance and equivalent resistance change with the operating state in two-level grid-connected inverters, resulting in a mismatch between the predictive model and the actual characteristics of the filter branch.
By collecting inverter output current, grid voltage, and DC bus voltage, the filter branch voltage is reconstructed, a filter branch parameter identification model is established, and the filter inductance and equivalent resistance are identified online and updated to the prediction model of model predictive control to achieve adaptive predictive control.
Adaptive predictive control of a two-level inverter is achieved when the filter parameters change dynamically, effectively solving the problem that the predictive model is difficult to continuously match the actual filter branch parameters and improving the control accuracy.
Smart Images

Figure CN123122283A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of inverter control technology, and in particular to an online identification and predictive control method, system and medium for a two-level inverter. Background Technology
[0002] Model predictive control (MDI) predicts the output current under different candidate switching states by establishing a discrete predictive model of the inverter, and selects the target switching state based on the evaluation results. Its control performance is highly dependent on the accuracy of the predictive model parameters. Existing technology CN113285481A estimates the filter inductance of the grid-connected converter online and uses the estimation results for predictive control to reduce the impact of inductor parameter deviations on predictive performance. CN113517822A, on the other hand, uses the least squares method to identify inverter load parameters online to improve the parameter mismatch problem in finite control set MDI.
[0003] However, existing solutions typically focus on estimating a single inductor parameter or identifying general inverter load parameters. They fail to address the situation in two-level grid-connected inverters where the filter inductance and equivalent resistance change together with the operating conditions. They fail to continuously correct the filter branch parameters in the predictive model using actual operating signals, leading to discrepancies between the predictive model and the actual filter branch characteristics, thus affecting the accuracy of model predictive control. Therefore, how to enable the filter branch parameters used in model predictive control to be dynamically updated with actual operating conditions has become a pressing problem to be solved in existing technologies.
[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention
[0005] This invention provides an online identification and predictive control method, system, and medium for two-level inverters, which can effectively solve the problems in the background art.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An online identification and predictive control method for a two-level inverter, the method comprising: Collect the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter; The inverter output voltage is reconstructed based on the DC bus voltage and the switch state, and the filter branch voltage is determined based on the inverter output voltage and the grid voltage. A filter branch parameter identification model is established based on the filter branch voltage, the output current, and the rate of change of the output current. The filter inductor and equivalent resistance are then identified online based on the filter branch parameter identification model. The filter inductor and equivalent resistance obtained by joint online identification are updated to the prediction model of model predictive control, and the predicted current corresponding to each candidate switch state is obtained based on the updated prediction model. The target switching state is determined based on the predicted current, and the two-level inverter is controlled based on the target switching state.
[0007] Furthermore, the two-level inverter is a three-phase two-level voltage source inverter, and the AC side of the two-level inverter is connected to the power grid through an L-type filter having the filter inductor and the equivalent resistance.
[0008] Furthermore, the determination of the filter branch voltage includes: The output current and the grid voltage are transformed by coordinate transformation to obtain the first current component, the second current component, the first grid voltage component and the second grid voltage component in a stationary two-axis coordinate system. The inverter output voltage is reconstructed based on the DC bus voltage and the switching state, and the first output voltage component and the second output voltage component of the inverter output voltage in the stationary two-axis coordinate system are obtained. The first filter branch voltage component is determined based on the first output voltage component and the first grid voltage component, and the second filter branch voltage component is determined based on the second output voltage component and the second grid voltage component.
[0009] Furthermore, the online identification includes: The rate of change of the output current is determined based on the output current, and a filter branch parameter identification model is established based on the filter branch voltage, the output current, and the rate of change of the output current. The filter branch voltage is determined by the voltage drop across the equivalent resistance and the voltage drop across the filter inductor. Within a preset sliding time window, the voltage of the filter branch, the output current, and the rate of change of the output current at multiple sampling times are obtained, and a parameter estimation relationship is constructed based on the data at multiple sampling times. The filter inductor and the equivalent resistance are jointly determined using the least squares method based on the parameter estimation relationship.
[0010] Further, acquiring the filter branch voltage, the output current, and the rate of change of the output current at multiple sampling times includes: The output current and the voltage of the filter branch are filtered. The rate of change of the output current is determined based on the output current corresponding to the current sampling time and the previous sampling time; The filtered output current, the filtered branch voltage, and the rate of change of the output current are used in the filtered branch parameter identification model.
[0011] Further, updating the prediction model with the online identified filter inductance and equivalent resistance includes: The validity of the parameters of the filter inductor and the equivalent resistance obtained online is determined; When the filter inductor or the equivalent resistance exceeds the corresponding preset physical constraint range, the parameter exceeding the preset physical constraint range is subjected to amplitude limiting, or the effective parameter corresponding to the previous parameter update cycle is used. The filter inductance and equivalent resistance, determined by the validity of the parameters, are updated in the prediction model.
[0012] Further, based on the updated prediction model, the predicted current corresponding to each candidate switch state is obtained, including: The nominal parameters of the filter inductor and the nominal parameters of the equivalent resistance in the prediction model are replaced with the filter inductor and the equivalent resistance obtained online. Based on the current output current, the grid voltage, the filter inductor, the equivalent resistance, and the inverter output voltage corresponding to each candidate switch state, the predicted current for the next control cycle corresponding to each candidate switch state is determined.
[0013] Further, determining the target switching state based on the predicted current includes: An evaluation function is constructed based on the error between the predicted current and the reference value of the output current in the next control cycle corresponding to each candidate switch state. The evaluation result corresponding to each candidate switch state is determined based on the evaluation function, wherein the evaluation function further includes a switch action constraint term or a switch frequency constraint term; The candidate switch state with the smallest evaluation function value is determined as the target switch state.
[0014] An online identification and predictive control system for a two-level inverter, the system comprising: The signal acquisition module collects the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter. The voltage reconstruction module reconstructs the inverter output voltage based on the DC bus voltage and switch status, and determines the filter branch voltage based on the inverter output voltage and the grid voltage. The parameter joint identification module establishes a filter branch parameter identification model based on the filter branch voltage, output current, and rate of change of output current, and jointly identifies the filter inductor and equivalent resistance online based on the filter branch parameter identification model. The prediction model update module updates the filter inductance and equivalent resistance obtained by joint online identification to the prediction model of model prediction control, and obtains the prediction current corresponding to each candidate switch state based on the updated prediction model. The target switch control module determines the target switch state based on the predicted current and controls the two-level inverter based on the target switch state.
[0015] A computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, can implement the online identification and predictive control method for a two-level inverter.
[0016] The technical solution of this invention can achieve the following technical effects: By reconstructing the output voltage of a two-level inverter and determining the voltage of the filter branch, the filter inductor and equivalent resistance are identified online based on the filter branch voltage, output current and their rate of change. The identified parameters are updated in real time to the prediction model of model predictive control, and the candidate switch state is predicted to determine the target switch state. This achieves adaptive predictive control of the two-level inverter, effectively solving the problem that the prediction model of model predictive control is difficult to continuously match the actual filter branch parameters when the filter parameters of a two-level grid-connected inverter change dynamically.
[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 A flowchart illustrating the online identification and predictive control method for a two-level inverter; Figure 2 This is a structural diagram of a two-level inverter system; Figure 3 A block diagram of the control strategy structure; Figure 4 The current waveform is for a traditional parameterless identification strategy; Figure 5 Total harmonic distortion map for traditional parameter-free identification strategies; Figure 6 This is a current waveform diagram under the strategy of this invention; Figure 7 This is the total harmonic distortion diagram under the strategy of this invention; Figure 8 The parameter identification algorithm of this invention outputs the identification result. Detailed Implementation
[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.
[0022] Example 1;
[0023] like Figure 1 As shown, this application provides an online identification and predictive control method for a two-level inverter, the method comprising: S10: Collect the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter; S20: Reconstruct the inverter output voltage based on the DC bus voltage and switch status, and determine the filter branch voltage based on the inverter output voltage and the grid voltage; S30: Establish a filter branch parameter identification model based on the filter branch voltage, output current and the rate of change of output current, and jointly identify the filter inductor and equivalent resistance online based on the filter branch parameter identification model; S40: Update the filter inductance and equivalent resistance obtained by joint online identification to the prediction model of model predictive control, and obtain the predicted current corresponding to each candidate switch state based on the updated prediction model. S50: Determines the target switching state based on the predicted current and controls the two-level inverter based on the target switching state.
[0024] Specifically, this embodiment uses a two-level inverter as the controlled object. The AC side of the inverter is connected to the power grid via a filter branch, which includes a filter inductor and an equivalent resistance. During the control process, the inverter output current, grid voltage, DC bus voltage, and current switching state are collected. The switching state represents the conduction combination of each bridge arm in the current control cycle and is used to characterize the current output state of the inverter. The actual output voltage of the inverter is determined based on the DC bus voltage and the current switching state, and the filter branch voltage is obtained by combining it with the grid voltage. The filter branch voltage refers to the voltage across the two ends of the filter branch between the inverter output terminal and the power grid, and is used to characterize the actual voltage excitation received by the filter branch. A filter branch parameter identification model is established based on the filter branch voltage, output current, and rate of change of output current. The filter inductor and equivalent resistance are used as parameters to be identified for joint online identification. The rate of change of output current represents the degree of change of output current over time and is used to characterize the dynamic current response corresponding to the filter inductor. During the online identification process, the filter branch parameters are continuously updated using the operating data obtained by continuous sampling, so that the identification results can reflect the actual parameters of the filter branch under the current operating state. The filter inductance and equivalent resistance obtained online are updated to the prediction model of model predictive control, so that the filter branch parameters in the prediction model are corrected according to the actual operating parameters; the candidate switch state refers to the optional bridge arm switch combination that can be evaluated by model predictive control in the current control cycle, and different candidate switch states correspond to different inverter output states; the updated prediction model is used to obtain the prediction current corresponding to each candidate switch state, so that the prediction result of the candidate state is generated based on the currently identified filter branch parameters; The target switching state is determined based on the predicted current corresponding to each candidate switching state. The target switching state refers to the candidate switching state used to drive the inverter operation after evaluation by model predictive control. The corresponding control signal is generated based on the target switching state and the two-level inverter is controlled to operate. This forms a closed-loop predictive control process that uses the actual operating signal to jointly identify the filter branch parameters online, corrects the prediction model using the identification results, and further completes the selection of the switching state.
[0025] The technical solution of this invention reconstructs the output voltage of a two-level inverter and determines the voltage of the filter branch. Based on the filter branch voltage, output current and their rate of change, the filter inductor and equivalent resistance are identified online. The identified parameters are updated in real time to the prediction model of model predictive control, and the candidate switch state is predicted to determine the target switch state. This achieves adaptive predictive control of the two-level inverter and effectively solves the problem that the prediction model of model predictive control is difficult to continuously match the actual filter branch parameters when the filter parameters of a two-level grid-connected inverter change dynamically.
[0026] Furthermore, the two-level inverter is a three-phase two-level voltage source inverter. The AC side of the two-level inverter is connected to the power grid through an L-type filter with a filter inductor and equivalent resistance.
[0027] As a preferred embodiment of the above, the two-level inverter adopts a three-phase two-level voltage source topology, with the DC side connected to the DC bus and the AC side including three-phase bridge arms a, b, and c; each phase bridge arm includes upper and lower power switching devices, and a total of six power switching devices are set for the three phases, with the midpoint of each phase bridge arm forming the corresponding phase's AC output terminal; the switching state of each phase bridge arm is controlled to generate the corresponding bridge arm output voltage at each phase's AC output terminal, and the AC current is output to the grid through an L-type filter; An L-type filter is connected between the AC output of a two-level inverter and the power grid. Each phase filter branch includes the output filter inductor. and the equivalent resistance corresponding to the output filter inductor Equivalent resistance Used to characterize the internal resistance of the output filter inductor; the inverter output current is injected into the grid after flowing through the filter branch. The filter branch uses the output filter inductor to suppress the high-frequency current component generated by the inverter switching action, and at the same time uses the output filter inductor and its equivalent resistance as the objects of subsequent online identification and prediction model updates. Based on the connection relationship between the three-phase two-level voltage source inverter and the L-type filter, the following conditions must be met between the inverter's AC output terminal, the power grid, and the filter branch: ; in, This is the voltage on the three-phase grid side. This is the output voltage of the three-phase bridge arm. This refers to the output current of the three-phase inverter. This is the equivalent resistance of the L-type filter. For output filter inductor, The rate of change of the output current of the three-phase inverter with respect to time; the DC bus voltage is denoted as... It is used to provide DC-side input voltage for the three-phase bridge arm.
[0028] Furthermore, determining the voltage of the filter branch includes: The output current and the grid voltage are transformed by coordinate transformation to obtain the first current component, the second current component, the first grid voltage component and the second grid voltage component in a stationary two-axis coordinate system. The inverter output voltage is reconstructed based on the DC bus voltage and switching state, and the first and second output voltage components of the inverter output voltage in a stationary two-axis coordinate system are obtained. The voltage component of the first filter branch is determined based on the first output voltage component and the first grid voltage component, and the voltage component of the second filter branch is determined based on the second output voltage component and the second grid voltage component.
[0029] As a preferred embodiment of the above, the three-phase output current is... and three-phase grid voltage Perform a Clark transformation to convert the electrical quantities in the three-phase stationary coordinate system to the αβ stationary two-axis coordinate system, and obtain the α-axis current. β-axis current α-axis grid voltage and β-axis grid voltage The first current component corresponds to the α-axis current. The second current component corresponds to the β-axis current. The first grid voltage component corresponds to the α-axis grid voltage. The second grid voltage component corresponds to the β-axis grid voltage. The αβ stationary two-axis coordinate system is a two-dimensional stationary coordinate system that uses mutually orthogonal α-axis and β-axis to represent three-phase electrical quantities. After Clark transformation of the three-phase electrical quantities, the electrical relationship of the filter branch is expressed as follows: ; in, Represents the grid voltage along the α axis and β-axis grid voltage The voltage of the power grid constituted This indicates the output voltage of the α-axis inverter. and β-axis inverter output voltage The output voltage of the inverter. Represented by the α-axis current and β-axis current The resulting output current This represents the rate of change of the output current in the αβ stationary two-axis coordinate system; Obtain DC bus voltage and the current status of the three-phase bridge arm switches. ,according to and Determine the inverter output voltage vector corresponding to the current switching combination, and obtain the α-axis output voltage of this output voltage vector in the αβ stationary two-axis coordinate system. and β-axis output voltage ;in, These represent the switching states of phases a, b, and c, respectively; the finite control set of the two-level inverter includes two zero-voltage vectors and six non-zero active voltage vectors, with different three-phase bridge arm switching states corresponding to the corresponding output voltage vectors; the first output voltage component corresponds to the α-axis output voltage. The second output voltage component corresponds to the β-axis output voltage. ; get Then, calculate the filter branch voltages on the α-axis and β-axis respectively: ; in, This refers to the voltage component of the first filter branch, i.e., the component of the filter branch voltage on the α axis. This is the voltage component of the second filter branch, i.e., the component of the filter branch voltage on the β axis. Thus, the three-phase electrical relationship between the inverter output and the grid is converted into voltage and current relationships in two mutually orthogonal directions, the α axis and the β axis, providing input for subsequent parameter identification based on the filter branch voltage, output current and output current change rate.
[0030] Furthermore, online identification includes: The rate of change of the output current is determined based on the output current, and a filter branch parameter identification model is established based on the filter branch voltage, output current and the rate of change of the output current. The filter branch voltage is determined by the voltage drop across the equivalent resistance and the voltage drop across the filter inductor. Within a preset sliding time window, the filter branch voltage, output current, and rate of change of output current at multiple sampling times are obtained, and parameter estimation relationships are constructed based on the data at multiple sampling times. The filter inductor and equivalent resistance are determined jointly using the least squares method based on the parameter estimation relationship.
[0031] As a preferred embodiment of the above, the equivalent resistance of the L-type filter is determined based on the output current and filter branch voltage in the αβ stationary two-axis coordinate system. and filter inductor Joint online identification; joint online identification refers to simultaneously obtaining the parameters during the same parameter estimation process. and This allows the two parameters to be updated synchronously based on the inverter's current operating data. Based on the voltage balance relationship of the filter branch, the voltage of the filter branch is composed of the voltage drop across the resistance generated by the equivalent resistance and the voltage drop across the inductance generated by the filter inductance, which can be expressed as follows in the αβ stationary two-axis coordinate system: ; in, and These are the α-axis and β-axis components of the filter branch voltage, respectively. and These are the α-axis and β-axis components of the output current, respectively. and These represent the rate of change of the α-axis output current and the rate of change of the β-axis output current, respectively. The output current change rate is determined based on the continuously sampled output current changes and is used to characterize the degree of change in output current per unit time. Each time the parameters are updated, data from multiple consecutive sampling moments are selected within a preset sliding time window. The sliding time window refers to a data interval that moves forward with the current sampling moment and includes the most recent consecutive sampled data. Let the time window contain... Each sampling time, The integer is greater than 1; at each sampling time, a set of branch voltage, output current, and output current change rate data along the α-axis and β-axis are provided respectively; Based on the above formula, the α-axis data at each sampling time is... and As a parameter coefficient, As the corresponding output; β-axis data is... and As a parameter coefficient, As the corresponding output, the data from each sampling moment within the sliding time window are sequentially combined to form an observation matrix. and output vector Among them, the observation matrix The output vector consists of the output current components and the corresponding rate of change of current at each sampling time. It consists of the voltage components of the filtered branch at the corresponding sampling time, thus yielding: ; in: ; The parameter vector to be estimated includes the equivalent resistance. and filter inductor ; The observation matrix; This is the output vector; Based on multiple sets of sampled data contained in the sliding time window, the parameters to be estimated are jointly obtained using the least squares method, and the calculation relationship is as follows: ; in, This is a parameter estimation vector, whose two elements correspond to the estimated equivalent resistance values. and filter inductance estimate ; Represents the observation matrix The transpose of the matrix; Representation matrix The inverse matrix; At each parameter update time, the latest sampled data obtained within the current sliding time window is used to update the parameter. and Then recalculate according to the above formula. and By continuously moving the sliding time window and repeating parameter estimation, the identification results can be updated as the filter branch parameters change, providing the filter inductance and equivalent resistance under the current operating state for subsequent prediction model parameter correction.
[0032] Furthermore, the filter branch voltage, output current, and rate of change of output current are obtained at multiple sampling times, including: The output current and the voltage of the filter branch are filtered. The rate of change of the output current is determined based on the output current corresponding to the current sampling time and the previous sampling time. The filtered output current, the voltage of the filtered branch, and the rate of change of the output current are used in the filter branch parameter identification model.
[0033] As a preferred embodiment of the above, before constructing the parameter identification data for each sampling moment within the sliding time window, signal preprocessing is performed on the output current and the voltage of the filter branch; specifically, for the α-axis output current... β-axis output current α-axis filter branch voltage and β-axis filter branch voltage Low-pass filtering is performed separately to reduce the impact of sampling noise and high-frequency ripple generated by inverter switching on the parameter identification results. The current components and voltage components of the filtered branches after low-pass filtering are retained according to the corresponding sampling time for subsequent parameter estimation. Within each sampling period, the α-axis output current and β-axis output current at the current sampling time and the previous sampling time are acquired respectively. The previous sampling time refers to the sampling time that is adjacent to the current sampling time and located before the current sampling time in the sampling sequence. The output current of two adjacent sampling times on the same coordinate axis is differentially calculated, and combined with the sampling time interval between adjacent sampling times, the α-axis output current change rate and β-axis output current change rate corresponding to the current sampling time are obtained. The output current change rate characterizes the degree of change of the corresponding axis output current over time. Subsequently, the filtered α-axis output current, filtered β-axis output current, filtered α-axis filter branch voltage, filtered β-axis filter branch voltage, and the α-axis and β-axis output current change rates corresponding to the same sampling time are combined one-to-one according to the sampling time and used in the filter branch parameter identification model established above. Thus, each sampling time within the sliding time window forms a set of data for parameter estimation, so that the current components and current change rates in the observation matrix correspond to the filter branch voltage in the output vector in time, and are then used to jointly identify the filter inductance and equivalent resistance.
[0034] Furthermore, the filter inductance and equivalent resistance obtained online are updated in the prediction model, including: The validity of the parameters of the filter inductor and equivalent resistance obtained online is judged; When the filter inductance or equivalent resistance exceeds the corresponding preset physical constraint range, the parameter exceeding the preset physical constraint range is limited, or the effective parameter corresponding to the previous parameter update cycle is used. The filter inductance and equivalent resistance, determined by the validity of the parameters, are updated in the prediction model.
[0035] As a preferred embodiment of the above, after each online identification of the filter inductor and equivalent resistance is completed, the parameter estimation results obtained this time are first judged for parameter validity, and then it is decided whether to update the prediction model. The parameter validity judgment refers to comparing the currently identified filter inductor estimation value and equivalent resistance estimation value with the corresponding preset physical constraint range, so as to avoid parameters that deviate significantly from the actual physical characteristics from abnormal sampling, calculation abnormality or parameter estimation sudden change from entering the prediction model. The preset physical constraint range is the upper and lower limit range of the parameter range that the filter inductor and equivalent resistance are allowed to occur under the actual operating conditions of the inverter. The filter inductor and equivalent resistance are set with independent physical constraint ranges, and the validity of the two types of parameters is judged separately. The parameter update cycle refers to the time interval between two consecutive online identification results being used to update the parameters of the prediction model. The effective parameter refers to the filter inductor or equivalent resistance that is allowed to participate in the calculation of the prediction model after the parameter validity judgment. When the currently identified filter inductance and equivalent resistance are both within their respective preset physical constraints, the current identification result is determined as the effective parameter for this parameter update cycle and used to replace the corresponding parameter in the prediction model. When the filter inductance or equivalent resistance exceeds the corresponding preset physical constraint range, the parameters that exceed the range are either subjected to amplitude limiting or the valid parameters of the previous parameter update cycle are maintained. Amplitude limiting means adjusting parameters that are below the lower limit of the corresponding physical constraint range to the lower limit and adjusting parameters that are above the upper limit of the corresponding physical constraint range to the upper limit, so that the processed parameters are kept within the allowable physical range. When using the valid parameters of the previous parameter update cycle, the current identification result that exceeds the range is not written into the prediction model, but the corresponding parameter that has been most recently judged for validity is continued to be called. After the filter inductor and equivalent resistance have completed the above judgments, the effective parameters obtained through parameter validity judgment or after amplitude limiting are updated to the prediction model of model predictive control. In this way, the abnormal identification results can be prevented from directly changing the prediction model parameters, and the prediction model can still maintain usable filter branch parameters when the online identification results change abruptly.
[0036] Furthermore, based on the updated prediction model, the predicted current corresponding to each candidate switch state is obtained, including: The nominal parameters of the filter inductance and the nominal parameters of the equivalent resistance in the prediction model are replaced with the filter inductance and the equivalent resistance obtained online. Based on the current output current, grid voltage, filter inductance, equivalent resistance, and inverter output voltage corresponding to each candidate switch state, the predicted current for the next control cycle corresponding to each candidate switch state is determined.
[0037] As a preferred embodiment of the above, after completing the online identification and parameter validity judgment of the filter inductor and equivalent resistance, the currently valid filter inductor and equivalent resistance are written into the prediction model of the model predictive control, respectively replacing the original nominal parameters of the filter inductor and equivalent resistance in the prediction model; the nominal parameters are the filter branch parameters that are pre-configured and used for prediction calculation when the prediction model is established; through the above parameter replacement, the filter branch parameters used by the prediction model are updated with the current operating state; The continuous model of the two-level inverter in the αβ stationary two-axis coordinate system is discretized using the forward Euler method, resulting in: ; in, For sampling control period, Indicates the first Each sampling time, Indicates the first The above formula is used to predict the α-axis output current and β-axis output current at the next sampling time based on the output current, grid voltage, filter inductance, equivalent resistance, and inverter output voltage corresponding to the candidate switch state at the current sampling time. To illustrate the relationship between the filter branch parameters and the predicted current, let's assume... , These are the actual filter inductance and actual resistance in the actual system, respectively. , These are the filter inductor and resistance parameters used in model predictive control, respectively; the filter inductor current of the actual system satisfies: ; Among them, subscript Represents the actual value in the real system. This is the actual filter inductor. This is the actual resistance; The predictive model in model predictive control satisfies: ; Among them, subscript This represents the predicted value in the control system. The filter inductor parameters used in the prediction model. The resistance parameters used in the prediction model; , and They represent the first The output current, inverter output voltage, and grid voltage at each sampling moment in the αβ stationary two-axis coordinate system; The actual filter inductor current and the predicted current of the control system satisfy the following relationship: ; The above equation shows that when the filter inductor or resistor parameters used in the prediction model deviate from the actual parameters, it will result in a current prediction error at the next sampling time. The filter inductor directly affects the rate of change of the output current. When the prediction model underestimates the rate of change of current; when At this time, the prediction model overestimates the rate of change of current; the mismatch of resistance parameters will change the resistive voltage drop related to the current output current, causing the predicted current to deviate further from the actual current; thus, the deviation between the fixed parameters and the actual filter branch parameters will affect the prediction results of the effect of the candidate switch state. To reduce the prediction bias caused by the aforementioned parameter mismatch, the filter inductance and equivalent resistance currently identified online and judged for parameter validity are respectively used as parameters in the discretized continuous model. and Replace the original nominal parameters; after the parameters are updated, obtain the data within the current control cycle. , , and And for each candidate switch state in the finite control set, determine the corresponding and ; For a candidate switch state, the state corresponding to and Substituting the current output current, grid voltage, filter inductance, and equivalent resistance into the discretized continuous model, we obtain the following results: and Together, they constitute the predicted current for the next control cycle corresponding to the candidate switch state, which is used to characterize the expected output current after the candidate switch state has been applied for one sampling control cycle. The same method is used to traverse all candidate switch states within the finite control set, and the corresponding predicted current for the next control cycle is obtained for each candidate switch state. The prediction results of each candidate switch state are then used for evaluation function calculation and target switch state selection.
[0038] Furthermore, determining the target switching state based on the predicted current includes: An evaluation function is constructed based on the error between the predicted current and the reference value of the output current in the next control cycle corresponding to each candidate switch state. The evaluation results corresponding to each candidate switch state are determined based on the evaluation function, which also includes a switch action constraint term or a switch frequency constraint term. The candidate switch state with the smallest evaluation function value is determined as the target switch state.
[0039] As a preferred embodiment of the above, after obtaining the predicted current for the next control cycle corresponding to each candidate switch state in the finite control set, the error between each predicted current and the output current reference value is calculated, and an evaluation function is constructed based on the error; the output current reference value is the expected output current corresponding to the current control target, which is used to measure the degree to which the predicted current tracks the control target after the candidate switch state is activated; For current control in the αβ stationary two-axis coordinate system, the deviations between the α-axis predicted current and the β-axis predicted current corresponding to each candidate switching state and the corresponding output current reference component are compared. The prediction errors on the two coordinate axes are used together as the current tracking evaluation basis for the candidate switching state. The closer the predicted current is to the output current reference value, the smaller the corresponding current tracking error. To balance output current tracking performance and inverter switching operation status, the evaluation function may also include switching action constraint terms or switching frequency constraint terms. Switching action constraint terms characterize the switching actions caused by candidate switching states, ensuring that candidate state evaluation considers both current tracking error and the number of switching actions. Switching frequency constraint terms characterize the switching action frequency, ensuring that candidate state evaluation considers both current tracking error and switching frequency. When using switching action constraint terms, appropriate constraints are applied to candidate switching states that cause a large number of switching actions. When using switching frequency constraint terms, appropriate constraints are applied to the switching action frequency, thereby avoiding selection of switching states solely based on current prediction error. The evaluation function values of all candidate switch states in the finite control set are calculated sequentially according to the same evaluation rules, and the evaluation function values are compared. The candidate switch state with the smallest evaluation function value is determined as the target switch state for the current control cycle. After the target switch state is determined, the corresponding inverter drive signal is generated according to the switch state, so that the three-phase bridge arm operates according to the target switch state in the next control cycle. At the next sampling time, the current prediction and candidate state evaluation are re-executed, thus forming a closed-loop model predictive control process that is updated cycle by cycle.
[0040] Example 2;
[0041] Based on the same inventive concept as the online identification and predictive control method for a two-level inverter in the foregoing embodiments, the present invention also provides an online identification and predictive control system for a two-level inverter, the system comprising: The signal acquisition module collects the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter. The voltage reconstruction module reconstructs the inverter output voltage based on the DC bus voltage and switch status, and determines the filter branch voltage based on the inverter output voltage and the grid voltage. The parameter joint identification module establishes a filter branch parameter identification model based on the filter branch voltage, output current, and rate of change of output current, and jointly identifies the filter inductor and equivalent resistance online based on the filter branch parameter identification model. The prediction model update module updates the filter inductance and equivalent resistance obtained by joint online identification to the prediction model of model prediction control, and obtains the prediction current corresponding to each candidate switch state based on the updated prediction model. The target switch control module determines the target switch state based on the predicted current and controls the two-level inverter based on the target switch state.
[0042] The system described above in this invention can effectively implement an online identification and predictive control method for a two-level inverter, and the technical effects it can achieve are as described in the above embodiments, which will not be repeated here.
[0043] Example 3;
[0044] Based on the same inventive concept as the online identification and predictive control method for a two-level inverter in the foregoing embodiments, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, can realize the online identification and predictive control method for a two-level inverter.
[0045] Example 4;
[0046] This embodiment verifies the control effect of the online identification and predictive control method under the condition of filter branch parameter mismatch using a three-phase two-level grid-connected inverter simulation system. like Figure 2 As shown, the simulation system uses a three-phase two-level voltage source inverter. The DC side is connected to a DC power supply, and the AC side is connected to a three-phase power grid through an L-type filter. The DC side voltage is set to 1200V, the AC side voltage is set to 690V, and the power grid frequency is set to 50Hz. The filter inductance and its equivalent resistance in the L-type filter are used as the filter branch parameters for model predictive control. like Figure 3 As shown, the control system collects the three-phase output current. Three-phase power grid voltage DC bus voltage and the status of the three-phase bridge arm switch The three-phase output current and three-phase grid voltage are converted to an αβ stationary two-axis coordinate system to obtain... and The inverter output voltage is reconstructed based on the DC bus voltage and the state of the bridge arm switches. Further calculation of the filter branch voltage The output current and filter branch voltage are filtered, and the rate of change of current is calculated based on the output current at adjacent sampling times. Within a sliding time window, the least squares method is used to jointly identify the filter inductor and equivalent resistance; the identified... and After effectiveness assessment, the prediction model is updated to the model predictive control model, and then the updated prediction model is used to complete the prediction of candidate switch states and the selection of target switch states. To simulate the parameter mismatch between the prediction model and the actual filter branch, the filter inductance initially used in the model prediction controller was set to 56μH, and the actual output filter inductance was set to 82μH; thus, there was an initial deviation of 26μH between the filter inductance in the prediction model and the actual filter inductance. First, the first set of comparative tests was conducted. The first set used fixed parameter model predictive control. Online parameter identification was not enabled during the entire test. The predictive model continuously used a 56μH filter inductor parameter to calculate the predicted current for the next control cycle corresponding to the candidate switch state. The results of the first group of tests are as follows: Figure 4 and Figure 5 As shown, where Figure 4 This is a diagram showing the three-phase grid-connected current waveform when using a fixed-parameter model for predictive control. Figure 5 The corresponding total harmonic distortion (THD) test results are shown. Under this operating condition, the total harmonic distortion (THD) of the output current is 1.76%. THD represents the total harmonic distortion rate, which is used to characterize the degree of waveform distortion caused by harmonic components in the grid-connected current. The second set of tests was then conducted. Under the same inverter main circuit parameters, grid parameters, initial filter inductor parameter mismatch conditions and control conditions as the first set, online parameter identification and predictive control was enabled. During the control process, inverter operating data was continuously collected, and the filter inductor and equivalent resistance were identified online within the sliding time window. The prediction model was updated using the effective identified parameters. The results of the second group of tests are as follows: Figure 6 and Figure 7 As shown, where Figure 6 To enable the three-phase grid-connected current waveform after online parameter identification, Figure 7 The corresponding total harmonic distortion (THD) test results are shown; the output current THD is 1.16% at this time; compared with 1.76% in the first group, the THD has decreased by 0.60 percentage points, indicating that the grid-connected current harmonic distortion has been reduced after correcting the prediction model using the online identification results. Further observation of the online identification process of the filter inductor yielded the following results: Figure 8 As shown; Figure 8 The horizontal axis represents time. The vertical axis represents the filter inductance. The unit is H; after the online identification is started, the estimated value of the filter inductance changes from the initial state to the actual filter inductance and converges within about 0.01s, and then stabilizes at around 82μH of the actual filter inductance; this shows that the online identification process can obtain filter inductance parameters close to the actual value when there is a deviation between the nominal parameters and the actual parameters of the filter inductance, and use the identification results to update the parameters of the prediction model. Combination Figure 4-7 The comparison results show that, under the parameter mismatch condition of an actual filter inductance of 82μH and an initial filter inductance of 56μH in the prediction model, the output current THD obtained by fixed parameter model predictive control is 1.76%. After adopting online parameter identification and real-time updating of the prediction model, the output current THD is reduced to 1.16%. Figure 8 It can be seen that during the THD improvement process, the identified value of the filter inductor can converge to the actual parameters, making the filter branch parameters used in the prediction model more consistent with the actual operating parameters, thereby reducing the output current prediction deviation caused by parameter mismatch.
[0047] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.
Claims
1. A two-level inverter online identification predictive control method, characterized in that, The method includes: Collect the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter; The inverter output voltage is reconstructed based on the DC bus voltage and the switch state, and the filter branch voltage is determined based on the inverter output voltage and the grid voltage. A filter branch parameter identification model is established based on the filter branch voltage, the output current, and the rate of change of the output current. The filter inductor and equivalent resistance are then identified online based on the filter branch parameter identification model. The filter inductor and equivalent resistance obtained by joint online identification are updated to the prediction model of model predictive control, and the predicted current corresponding to each candidate switch state is obtained based on the updated prediction model. The target switching state is determined based on the predicted current, and the two-level inverter is controlled based on the target switching state.
2. The two-level inverter online identification predictive control method according to claim 1, characterized in that, The two-level inverter is a three-phase two-level voltage source inverter. The AC side of the two-level inverter is connected to the power grid through an L-type filter with the filter inductor and the equivalent resistance.
3. The two-level inverter online identification predictive control method according to claim 1, characterized in that, The determination of the voltage of the filter branch includes: The output current and the grid voltage are transformed by coordinate transformation to obtain the first current component, the second current component, the first grid voltage component and the second grid voltage component in a stationary two-axis coordinate system. The inverter output voltage is reconstructed based on the DC bus voltage and the switching state, and the first output voltage component and the second output voltage component of the inverter output voltage in the stationary two-axis coordinate system are obtained. The first filter branch voltage component is determined based on the first output voltage component and the first grid voltage component, and the second filter branch voltage component is determined based on the second output voltage component and the second grid voltage component.
4. The two-level inverter online identification predictive control method according to claim 1, characterized in that, The online identification includes: The rate of change of the output current is determined based on the output current, and a filter branch parameter identification model is established based on the filter branch voltage, the output current, and the rate of change of the output current. The filter branch voltage is determined by the voltage drop across the equivalent resistance and the voltage drop across the filter inductor. Within a preset sliding time window, the voltage of the filter branch, the output current, and the rate of change of the output current at multiple sampling times are obtained, and a parameter estimation relationship is constructed based on the data at multiple sampling times. The filter inductor and the equivalent resistance are jointly determined using the least squares method based on the parameter estimation relationship.
5. The two-level inverter online identification predictive control method according to claim 4, characterized in that, Acquiring the filter branch voltage, the output current, and the rate of change of the output current at multiple sampling times includes: The output current and the voltage of the filter branch are filtered. The rate of change of the output current is determined based on the output current corresponding to the current sampling time and the previous sampling time; The filtered output current, the filtered branch voltage, and the rate of change of the output current are used in the filtered branch parameter identification model.
6. The online identification and predictive control method for a two-level inverter according to claim 1, characterized in that, Updating the prediction model with the filter inductance and equivalent resistance obtained online includes: The validity of the parameters of the filter inductor and the equivalent resistance obtained through online identification is determined. When the filter inductor or the equivalent resistance exceeds the corresponding preset physical constraint range, the parameter exceeding the preset physical constraint range is subjected to amplitude limiting, or the effective parameter corresponding to the previous parameter update cycle is used. The filter inductance and equivalent resistance, determined by the validity of the parameters, are updated in the prediction model.
7. The two-level inverter online identification predictive control method according to claim 1, characterized in that, Based on the updated prediction model, the predicted current corresponding to each candidate switch state is obtained, including: The nominal parameters of the filter inductor and the nominal parameters of the equivalent resistance in the prediction model are replaced with the filter inductor and the equivalent resistance obtained online. Based on the current output current, the grid voltage, the filter inductor, the equivalent resistance, and the inverter output voltage corresponding to each candidate switch state, the predicted current for the next control cycle corresponding to each candidate switch state is determined.
8. The two-level inverter online identification predictive control method according to claim 7, characterized in that, Determining the target switching state based on the predicted current includes: An evaluation function is constructed based on the error between the predicted current and the reference value of the output current in the next control cycle corresponding to each candidate switch state. The evaluation result corresponding to each candidate switch state is determined based on the evaluation function, wherein the evaluation function further includes a switch action constraint term or a switch frequency constraint term; The candidate switch state with the smallest evaluation function value is determined as the target switch state.
9. A two-level inverter on-line identification predictive control system, characterized in that, The system includes: The signal acquisition module collects the output current, grid voltage, DC bus voltage, and switch status of the two-level inverter. The voltage reconstruction module reconstructs the inverter output voltage based on the DC bus voltage and switch status, and determines the filter branch voltage based on the inverter output voltage and the grid voltage. The parameter joint identification module establishes a filter branch parameter identification model based on the filter branch voltage, output current, and rate of change of output current, and jointly identifies the filter inductor and equivalent resistance online based on the filter branch parameter identification model. The prediction model update module updates the filter inductance and equivalent resistance obtained by joint online identification to the prediction model of model prediction control, and obtains the prediction current corresponding to each candidate switch state based on the updated prediction model. The target switch control module determines the target switch state based on the predicted current and controls the two-level inverter based on the target switch state.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which includes program instructions that, when executed by a processor, implement the online identification and predictive control method for a two-level inverter as described in any one of claims 1-8.
Citation Information
Patent Citations
Grid-connected converter inductance parameter online estimation method, prediction control method and system
CN113285481A
Method for quickly identifying inverter parameters under finite set model predictive control
CN113517822A
