A current waveform optimization output method based on adaptive predictive control

Abstract

The application discloses a current waveform optimization output method based on adaptive predictive control and relates to the technical field of adaptive control. The method senses the load state change by online identification of equivalent parameters of the load and generation of a load mutation gating factor, predicts a natural output current waveform based on a series RL equivalent model, compares the natural output current waveform with an original expected current waveform to construct a deviation injection reference current waveform segment, and then coordinates the deviation injection close action and waveform curvature maintaining action in the same closed frame through the load mutation gating factor and its complementary factor, so that the reconstructed target current waveform segment is output to a subsequent current controller. The method can enhance compensation and inhibit overshoot when the load mutates, and can enhance smooth maintaining and inhibit sharp peaks and distortion when the load is stable, so that the dynamic response speed and the steady-state output quality are considered.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology, and in particular to a current waveform optimization output method based on adaptive predictive control. Background Technology

[0002] In applications such as programmable current sources, power electronic test power supplies, inverter output current control systems, and load current rapid change tests, current output devices typically need to output continuous, low-distortion, and dynamically responsive current signals according to a pre-defined desired current waveform. When the load condition is relatively stable, methods such as proportional-integral control, sliding mode control, adaptive control, or model predictive control can achieve basic current tracking. However, once the load's equivalent impedance, equivalent inductance, or dynamic characteristics change rapidly, the output current is prone to problems such as response lag, overshoot, local spikes, waveform bending, or harmonic distortion.

[0003] To improve current output performance under load disturbances, a common approach is to use fixed-proportion feedforward compensation. This type of solution directly adds the prediction error, load disturbance, or current error to the control input, PWM duty cycle, compensation voltage, or reference current at a fixed ratio. It is relatively simple to implement and can provide some compensation when the load changes slowly. However, fixed-proportion compensation is difficult to adapt to the different needs of load abrupt changes and load steady-state periods simultaneously: during load abrupt changes, fixed-proportion compensation is often insufficient or results in overshoot; while during load steady-state periods, fixed-proportion compensation can easily lead to overcompensation, causing jitter and spikes in the target current waveform.

[0004] Another common approach is to use disturbance observers, adaptive control, or predictive control. These approaches focus on online estimation of disturbances or optimization of control inputs, which can improve current tracking accuracy to some extent. However, the target current tracked by the controller is usually not reconstructed for sudden load changes; that is, the target current waveform remains the original desired waveform. When a sudden load change occurs, the control action mainly focuses on corrections at the drive signal level, making it difficult to compensate and adjust in advance at the waveform level. Building on this, some approaches further introduce smoothing filters or regularization to reduce reference current jitter. However, simple smoothing, while suppressing jitter, often weakens the rapid compensation capability required during sudden load changes.

[0005] In summary, existing current waveform control schemes struggle to simultaneously address the demands for rapid compensation during load abrupt changes and low distortion during load stabilization. Prioritizing rapid response can easily introduce jitter and distortion during stabilization; conversely, focusing on smooth maintenance can lead to response lag and overshoot during abrupt changes. Therefore, providing a current waveform optimization method that enhances compensation during load abrupt changes and suppresses jitter and distortion during load stabilization has become a pressing technical problem in this field. Summary of the Invention

[0006] This invention proposes a current waveform optimization output method based on adaptive predictive control. Its purpose is to solve the technical problems of existing current output systems, such as slow response and large overshoot during load changes, and the tendency to cause jitter and distortion of the target current waveform due to overcompensation during stable load conditions. The goal is to achieve a balance between rapid compensation during load changes and low distortion output during stable load conditions.

[0007] The technical solution of this invention is as follows:

[0008] A current waveform optimization method based on adaptive predictive control performs the following steps in each control cycle:

[0009] Step S1: Based on the acquired output voltage and output current, identify the load equivalent resistance and load equivalent inductance through the series RL equivalent model, and generate the load mutation gating factor according to the relative change of the load equivalent parameters between adjacent control cycles.

[0010] Step S2: Based on the current output current, current output voltage, and the load equivalent resistance and load equivalent inductance identified in step S1, predict the natural output current waveform segment within the future short window control steps using the series RL equivalent model, and construct the deviation injection reference current waveform segment according to the deviation between the original expected current waveform segment and the natural output current waveform segment.

[0011] Step S3: Based on the load mutation gating factor generated in step S1, the deviation injection reference current waveform segment constructed in step S2, the original desired current waveform segment, and the short-window second-order difference matrix, the reconstructed target current waveform segment is obtained through closed-loop reconstruction calculation, and the first sampling point of the reconstructed target current waveform segment is used as the target current reference value of the current control cycle and rolled out to the subsequent current controller.

[0012] As a further improvement to the current waveform optimization output method based on adaptive predictive control, step S1 includes: utilizing the nearest short window control step number Based on historical sampling data from each control cycle, the equivalent load resistance for the current control cycle is identified using the series RL equivalent model and the least squares method. Equivalent inductance of load ;right and The baseline quantity is normalized to obtain a normalized load parameter vector; based on the relative change between the normalized load parameter vectors of the current control cycle and the previous control cycle, the load mutation gating factor is generated. .

[0013] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the equivalent load resistance of the current control cycle is identified based on the series RL equivalent model and the least squares method. Equivalent inductance of load This includes: acquiring the output voltage for each control cycle. It is expressed as the sum of the voltage drop across the resistor and the voltage drop across the inductor. This forms a relationship with the load equivalent resistance. Equivalent inductance of load The system of overdetermined constraint equations is solved using the least squares method. and ;in, For the first Output current per control cycle This represents the change in current between adjacent control cycles. To control the cycle length.

[0014] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the... and Perform baseline normalization to obtain a normalized load parameter vector, including: according to The normalized load parameter vector is obtained. ;in, Rated load resistance, For rated load inductance, This indicates the matrix transpose.

[0015] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the generation of the load mutation gating factor... This includes: calculating relative changes. and to Amplitude limiting is performed to obtain , ;in, This is the normalized load parameter vector from the previous control cycle. express Norm, To prevent positive numbers with a denominator of zero.

[0016] As a further improvement to the current waveform optimization output method based on adaptive predictive control, step S2 includes: using the output current of the current control cycle as the initial current value for short-window prediction, and based on the output voltage of the current control cycle and the equivalent resistance of the load... The equivalent inductance of the load and control cycle length Based on the discretized form of the cascaded RL equivalent model, the number of future short-window control steps is calculated cycle by cycle. The natural output current of each control cycle is obtained and arranged in chronological order to form the natural output current waveform segment. ; Acquisition and The corresponding original desired current waveform segment ;according to and The deviation between them is used to construct the waveform segment of the reference current injected into the deviation. .

[0017] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the deviation injection reference current waveform segment is constructed. Including: according to structure .

[0018] As a further improvement to the current waveform optimization output method based on adaptive predictive control, step S3 includes: according to the number of short window control steps Construct the short-window second-order difference matrix According to the load mutation gating factor mentioned above. and The closed-form calculation formula, combined with the original expected current waveform segment. The waveform segment of the deviation injection reference current as well as The reconstructed target current waveform segment is calculated. ;Will The first sampling point is used as the target current reference value for the current control cycle and is continuously output to the subsequent current controller.

[0019] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the closed-loop calculation formula is:

[0020]

[0021] in, To and The corresponding identity matrix, for The transpose of .

[0022] As a further improvement to the current waveform optimization output method based on adaptive predictive control, the short-window second-order difference matrix The dimension is For lengths of Arbitrary current waveform segment ,satisfy , ;matrix The line in The column values ​​are , No. The column values ​​are , No. The column values ​​are The values ​​of the remaining columns are... .

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. This method coordinates the proximity effect of deviation injection and the waveform curvature preservation effect within the same closed reconstruction framework by using a load change gating factor and its complementary factor. When the load change is strong, the gating factor increases, enhancing the influence of the deviation injection reference current waveform segment on the reconstruction result, causing the target current waveform to include the compensation component in advance, improving the response speed and suppressing overshoot. When the load change is weak, the complementary factor increases, strengthening the waveform curvature preservation effect, ensuring that the reconstructed target current waveform maintains a smooth change trend similar to the original desired current waveform, suppressing unnecessary jitter and distortion. Thus, this method can simultaneously meet the requirements of rapid compensation during load change phases and low distortion during load stability phases, solving the technical problem that existing solutions cannot adapt to both load states simultaneously.

[0025] 2. Online identification of load equivalent parameters and generation of load change gating factors enable the target current waveform reconstruction process to possess real-time load state perception and adaptive adjustment capabilities. When load parameters change significantly, the gating factor automatically increases to enhance compensation; when load parameters change slightly, the gating factor automatically decreases to maintain smoothness. This allows for dynamic adjustment of the reconstruction strategy according to different load states, improving the response speed of the current output system under load change scenarios.

[0026] 3. The construction method of the deviation injection reference current waveform segment is as follows: when the predicted natural output current is lower than the original expected current, the deviation injection reference current waveform segment is higher than the original expected current, thus exhibiting an early compensation trend; when the predicted natural output current is higher than the original expected current, the deviation injection reference current waveform segment is lower than the original expected current, thus exhibiting an early overshoot suppression trend. This deviation injection is completed at the target current waveform level, rather than being directly superimposed on the subsequent control quantity or drive signal, thus reducing control impact and overshoot risks.

[0027] 4. A short-window second-order difference matrix is ​​introduced in the calculation of the reconstructed target current waveform segment to constrain the local bending trend of the waveform segment. Utilizing the characteristic of second-order difference to reflect the local curvature change of the waveform, this matrix can suppress spikes, jumps, and local distortions in the target current waveform during the reconstruction process. Simultaneously, the reconstruction calculation adopts a closed-loop solution method, integrating the deviation injection proximity term, waveform curvature preservation term, and original expectation preservation term into the same linear equation system. This is suitable for rolling real-time execution within each control cycle, balancing waveform quality optimization with the real-time requirements of engineering implementation. Attached Figure Description

[0028] Figure 1 This is a flowchart of a current waveform optimization output method based on adaptive predictive control. Detailed Implementation

[0029] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0030] The present invention provides a current waveform optimization output method based on adaptive predictive control, which is applicable to scenarios such as programmable current sources, power electronic test power supplies, inverter output current control systems, and load current rapid change tests, where the current output device needs to output a continuous, low-distortion, and dynamically responsive current signal according to the set desired current waveform.

[0031] Reference Figure 1 The diagram illustrates a flowchart of a current waveform optimization output method based on adaptive predictive control, provided by an embodiment of the present invention. Within each control cycle, this method sequentially completes a closed-loop rolling optimization process from sampling to target current reference value update through three steps: load equivalent parameter identification and abrupt gating factor generation, prediction of the natural output current waveform and deviation injection reference construction, and adaptive reconstruction of the target current waveform and rolling output. For ease of description, the following will refer to the method as follows: Taking a control cycle as an example, the explanation is as follows: A positive integer, representing the control period index, where the length of the control period is the sampling period. .

[0032] Step S1: Based on the acquired output voltage and output current, identify the load equivalent resistance and load equivalent inductance through the series RL equivalent model and the least squares method, and generate the load mutation gating factor according to the relative change of the load equivalent parameters between adjacent control cycles.

[0033] Specifically, in the current control cycle First, the output current of the current output system is collected. and output voltage And construct the system state vector ,in This represents the matrix transpose. The system state vector is used to characterize the operating state of the current output system within the current control cycle.

[0034] Subsequently, based on the output voltage and output current within the current and historical control cycles, using the most recent The load equivalent parameters are identified using historical sampling data from each control cycle. The number of steps for short window control ranges from 3 to 20. (This refers to setting the most recent...) The output voltage collected in each control cycle within a control cycle Output current and the current change between adjacent control cycles Time alignment is performed, and based on the voltage relationship of the series RL equivalent circuit, the output voltage is expressed as the sum of the resistor voltage drop and the inductor voltage drop:

[0035]

[0036] In the above formula, The load equivalent resistance, The load equivalent inductance, Sampling period. Resistance voltage drop. Proportional to the current, used to characterize the steady-state energy loss; inductor voltage drop It is proportional to the change in current and is used to characterize the inertial part of the dynamic response.

[0037] Recently The voltage relationships within each control cycle are uniformly organized to form a statement regarding the equivalent resistance of the load. Equivalent inductance of load The over-constrained equation set ensures that each control cycle corresponds to a voltage relationship used to fit the equivalent load parameters. The least squares method is then employed. From the historical data of each control cycle, find a set of equivalent load resistance and equivalent load inductance to minimize the cumulative error between the output voltage calculated by the series RL load equivalent model and the actual acquired output voltage. This is used to obtain the equivalent load resistance for the current control cycle. Equivalent inductance of load This forms the load equivalent parameter vector for the current control cycle. .

[0038] Next, the load equivalent parameters are normalized using reference quantities, with reference resistors used for each. and reference inductor Normalization is performed to obtain the normalized load parameter vector. :

[0039]

[0040] In the above formula, The rated load resistance of the current output system. This is the rated load inductance of the current output system.

[0041] Then, based on the normalized load parameter vector of the current control cycle Normalized load parameter vector from the previous control cycle The relative changes between them are used to calculate the load mutation gating factor. Specifically, first calculate the relative change. :

[0042]

[0043] In the above formula, Representing vectors Norm, To prevent positive numbers with a denominator of zero, the range of values ​​is [value missing]. This is used to prevent the relative change in the load parameter vector from being abnormally amplified due to an excessively small denominator. Then, the relative change is further limited to obtain the load mutation gating factor. :

[0044]

[0045] in, .when When the value is close to 0, it indicates that the load equivalent parameter changes little between adjacent control cycles, and the load state is relatively stable; when... A value close to 1 indicates a significant change in the load equivalent parameters and a large load mutation intensity. Load mutation gating factor. This provides a mutation criterion for the adaptive target current waveform reconstruction in the subsequent step S3.

[0046] Step S2: Based on the current output current, output voltage, and the equivalent load resistance identified in step S1. Equivalent inductance of load The natural output current waveform segment within the future short-window control steps is predicted using a series RL equivalent model. And based on the original desired current waveform segment With natural output current waveform segment The deviation between them is used to construct a deviation injection reference current waveform segment. .

[0047] Specifically, in obtaining the load equivalent resistance identified in step S1 Equivalent inductance of load Then, the output current of the current control cycle is used as the initial current value for short-window prediction, and the output voltage and equivalent load resistance of the current control cycle are used as the initial current value. Equivalent inductance of load and sampling period As the prediction input, the natural output current for each control cycle within the future short window is calculated cycle by cycle according to the discretized form of the series RL load equivalent model. The series RL load equivalent model treats the current output process as a dynamic response process determined by both resistance and inductance. The resistance component describes the steady-state voltage drop effect of the current at the current moment, while the inductance component describes the inertial effect of the current change over time. This completes the continuous... After predicting for one control cycle, the resulting multiple natural output currents are arranged in time sequence to form a predicted natural output current waveform segment. :

[0048]

[0049] In the above formula, Indicates the first The future number predicted within the control cycle is... The natural output current of each control cycle. , for A column vector representing the future current output system without additional compensation control. The current waveform segment that will be naturally output within each control cycle.

[0050] Next, the number of short-window control steps is read from the upper-level waveform setting module or external control commands. Corresponding original desired current waveform segment :

[0051]

[0052] In the above formula, Indicates the first The future number set within each control cycle The initial desired current value for each control cycle. For the initial desired current waveform segment. and predicting natural output current waveform segments Perform a consistency check on the control cycle and current unit. If the two dimensions are inconsistent, re-truncate or resample according to the task configuration until the two dimensions are consistent. .

[0053] Subsequently, based on the original desired current waveform segment With the predicted natural output current waveform segment The deviation between them is used to construct a deviation injection reference current waveform segment. :

[0054]

[0055] In the above formula, for A dimensional column vector, representing the 3D column vector, represents the 3D column vector. The number of future short-window control steps constructed within each control cycle The deviation of each control cycle is injected into the reference current waveform segment. The purpose of this construction is to: predict the natural output current. Lower than the original expected current When, deviation Positive, deviation injected reference current waveform segment Current waveform segment higher than the original expected current This results in a tendency to compensate in advance; when predicting the natural output current. Higher than the original expected current When, deviation Negative, deviation injected reference current waveform segment Current waveform segment below the original expected value This allows for early suppression of overshoot. (Deviation injection reference current waveform segment) This provides compensation direction guidance for the adaptive reconstruction in the subsequent step S3.

[0056] Step S3: Based on the load mutation gating factor generated in step S1 The deviation injection reference current waveform segment constructed in step S2 Original desired current waveform segment and the short-window second-order difference matrix The target current waveform segment is obtained through closed-loop reconstruction calculation. The first sampling point of the reconstructed target current waveform segment is used as the target current reference value for the current control cycle and is rolled out to the subsequent current controller.

[0057] Specifically, firstly, control the number of steps based on the short window. Constructing a short-window second-order difference matrix Short-window second-order difference matrix The dimension is Used for lengths of Perform a sliding three-point second-order difference operation on a current waveform segment. Let the length be... The arbitrary current waveform segment is ,but and The product of yields a A dimensional vector, the vector whose dimensional vector is the first dimensional vector. The elements are:

[0058]

[0059] Accordingly, matrix The line in The column values ​​are , No. The column values ​​are , No. The column values ​​are The values ​​of the remaining columns are... ,Right now Each row takes a value only in three consecutive positions. The second-order difference reflects the local curvature changes of the waveform: when there is a linear change between three adjacent sampling points, the second-order difference is zero, indicating local smoothness of the waveform; when there are abrupt jumps or peaks between three adjacent sampling points, the absolute value of the second-order difference is large, indicating severe local bending of the waveform. Therefore, Waveform segments can be measured The overall degree of bending.

[0060] Next, a closed-form calculation formula for reconstructing the target current waveform segment is constructed according to the three functions of "preserving the original expectation, ensuring the deviation injection is close, and preserving the waveform curvature". Among them, "preserving the original expectation" is used to ensure that the reconstructed target current waveform segment does not deviate from the original expected current waveform segment. "Offset Injection Closer" is used to inject a reference current waveform segment into the reconstructed target current waveform segment when a sudden increase in load occurs. The waveform is brought closer to the target current waveform to introduce compensation components in advance; "waveform curvature preservation" is used to keep the reconstructed target current waveform segment close to the original desired current waveform segment when the load change is weak. Similar smooth change trends suppress unwanted spikes, jitter, and local bends.

[0061] The coordination of the above three functions is achieved through the load mutation gating factor. and its complementary factors To achieve: load mutation gating factor Used to adjust for deviations in injection. The larger the value, the stronger the influence of the bias injection reference current waveform segment on the reconstruction result; complementarity factor Used to adjust the waveform curvature retention function. The larger the value, the more likely the reconstructed target current waveform segment will maintain the smooth change trend of the original desired current waveform segment. Therefore, when the load change is strong ( near The reconstructed target current waveform segment includes a compensation component in advance; when the load change is weak ( near The target current waveform segment is reconstructed to suppress unnecessary spikes, jitter, and local bends.

[0062] The closed-form calculation formula for reconstructing the target current waveform segment is as follows:

[0063]

[0064] In the above formula, for A dimensional column vector, representing the 3D column vector, represents the 3D column vector. The number of future short-window control steps calculated within each control cycle The reconstructed target current waveform segment for each control cycle; To control the number of steps in the short window corresponding identity matrix; Represents the second-order difference matrix of the short window The transpose of .

[0065] The functions of each term in this closed-form calculation are as follows: The original expectation is preserved, and a reference anchor point is provided for the reconstructed target current waveform segment to prevent the reconstruction result from deviating from the original set waveform without constraints. To inject a proximity term for the deviation, the influence of the deviation injection reference current waveform segment on the reconstruction result is increased when there is a significant load change, so that the reconstruction target current waveform segment includes the compensation component in advance. This represents the local curvature preservation information extracted from the original desired current waveform segment, used to prevent unnecessary spikes, jumps, or local bends from appearing in the reconstructed target current waveform segment; This is a waveform curvature preservation term, designed to maintain the smooth variation trend of the reconstructed target current waveform segment, closely resembling the original desired current waveform segment, when the load is stable or its variation is weak. In the matrix inversion term, This is used as an amplitude constraint weight term to ensure that the reconstructed target current waveform segment is constrained at the amplitude level. This is a curvature constraint weight term used to constrain the degree of local bending when solving for the reconstructed target current waveform segment. The closed-form calculation is as follows: Linear equation systems can be solved efficiently through matrix decomposition or direct inversion, making them suitable for rolling real-time execution within each control cycle.

[0066] After obtaining the reconstructed target current waveform segment Then, its first sampling point The target current reference value for the current control cycle is input to the subsequent current controller. The subsequent current controller generates a drive signal based on the deviation between the target current reference value and the actual output current to drive the current output actuator to output the current waveform.

[0067] In the next control cycle Internally, re-collect the output current. and output voltage Steps S1 to S3 are repeated to recalculate the load equivalent parameters, load mutation gating factor, predicted natural output current waveform segment, deviation injection reference current waveform segment, and reconstructed target current waveform segment. The first sampling point of the newly reconstructed target current waveform segment is then used as the target current reference value and input to the subsequent current controller. Through this rolling update process, the current output system completes a short-window target current waveform segment reconstruction once in each control cycle, thereby achieving continuous optimized current waveform output.

[0068] In summary, this embodiment provides a current waveform optimization output method based on adaptive predictive control. It generates a load mutation gating factor online by identifying equivalent load parameters, predicts the natural output current using a series RL equivalent model, and constructs a deviation injection reference current waveform segment. Finally, within the same closed frame, it coordinates the deviation injection proximity effect and waveform curvature preservation effect to obtain a reconstructed target current waveform segment and output it in a rolling manner. This method can automatically enhance compensation and suppress overshoot in advance during load mutations, and automatically enhance smoothness preservation and suppress spikes and distortions when the load is stable, thus balancing dynamic response speed and steady-state output quality.

[0069] It should be noted that, as will be apparent to those skilled in the art, the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics thereof. The scope of the present invention is defined by the claims rather than the foregoing description.

Claims

1. A current waveform optimization output method based on adaptive predictive control, characterized in that, The following steps are performed in each control cycle: Step S1: Based on the acquired output voltage and output current, identify the load equivalent resistance and load equivalent inductance through the series RL equivalent model, and generate the load mutation gating factor according to the relative change of the load equivalent parameters between adjacent control cycles. Step S2: Based on the current output current, current output voltage, and the load equivalent resistance and load equivalent inductance identified in step S1, predict the natural output current waveform segment within the future short window control steps using the series RL equivalent model, and construct the deviation injection reference current waveform segment according to the deviation between the original expected current waveform segment and the natural output current waveform segment. Step S3: Based on the load mutation gating factor generated in step S1, the deviation injection reference current waveform segment constructed in step S2, the original desired current waveform segment, and the short-window second-order difference matrix, the reconstructed target current waveform segment is obtained through closed-loop reconstruction calculation, and the first sampling point of the reconstructed target current waveform segment is used as the target current reference value of the current control cycle and rolled out to the subsequent current controller.

2. The current waveform optimization output method based on adaptive predictive control according to claim 1, characterized in that, Step S1 includes: controlling the number of steps using the nearest short window. Based on historical sampling data from each control cycle, the equivalent load resistance for the current control cycle is identified using the series RL equivalent model and the least squares method. Equivalent inductance of load ;right and The baseline quantity is normalized to obtain a normalized load parameter vector; based on the relative change between the normalized load parameter vectors of the current control cycle and the previous control cycle, the load mutation gating factor is generated. .

3. The current waveform optimization output method based on adaptive predictive control according to claim 2, characterized in that, The equivalent load resistance for the current control cycle is identified based on the series RL equivalent model and the least squares method. Equivalent inductance of load This includes: acquiring the output voltage for each control cycle. It is expressed as the sum of the voltage drop across the resistor and the voltage drop across the inductor. This forms a relationship with respect to the load equivalent resistance. Equivalent inductance of load The system of overdetermined constraint equations is solved using the least squares method. and ;in, For the first Output current per control cycle This represents the change in current between adjacent control cycles. To control the cycle length.

4. The current waveform optimization output method based on adaptive predictive control according to claim 3, characterized in that, The pair and Perform baseline normalization to obtain a normalized load parameter vector, including: according to The normalized load parameter vector is obtained. ;in, Rated load resistance, For rated load inductance, This indicates the matrix transpose.

5. The current waveform optimization output method based on adaptive predictive control according to claim 4, characterized in that, The generation of the load mutation gating factor This includes: calculating relative changes. and to Amplitude limiting is performed to obtain , ;in, This is the normalized load parameter vector from the previous control cycle. express Norm, To prevent positive numbers with a denominator of zero.

6. The current waveform optimization output method based on adaptive predictive control according to claim 1, characterized in that, Step S2 includes: using the output current of the current control cycle as the initial current value for short-window prediction, based on the output voltage of the current control cycle and the equivalent resistance of the load. The equivalent inductance of the load and control cycle length Based on the discretized form of the cascaded RL equivalent model, the number of future short-window control steps is calculated cycle by cycle. The natural output current of each control cycle is obtained and arranged in chronological order to form the natural output current waveform segment. ; Acquisition and The corresponding original desired current waveform segment ;according to and The deviation between them is used to construct the waveform segment of the reference current injected into the deviation. .

7. The current waveform optimization output method based on adaptive predictive control according to claim 6, characterized in that, The construction of the deviation injection reference current waveform segment Including: according to structure .

8. The current waveform optimization output method based on adaptive predictive control according to claim 1, characterized in that, Step S3 includes: controlling the number of steps according to the short window. Construct the short-window second-order difference matrix According to the load mutation gating factor mentioned above. and The closed-form calculation formula, combined with the original expected current waveform segment. The deviation injection reference current waveform segment as well as The reconstructed target current waveform segment is calculated. ;Will The first sampling point is used as the target current reference value for the current control cycle and is continuously output to the subsequent current controller.

9. The current waveform optimization output method based on adaptive predictive control according to claim 8, characterized in that, The closed-form calculation formula is: ； in, To and The corresponding identity matrix, for The transpose of .

10. The current waveform optimization output method based on adaptive predictive control according to claim 9, characterized in that, The short-window second-order difference matrix The dimension is For lengths of Arbitrary current waveform segment ,satisfy , ;matrix The line in The column values ​​are , No. The column values ​​are , No. The column values ​​are The values ​​of the remaining columns are... .

Images