Inverter predictive control method based on dynamic linearization

Abstract

The invention discloses an inverter predictive control method based on dynamic linearization. The method comprises the following steps: representing an inverter system as a local partial format dynamic linearization linear data model; estimating a linear part of the local partial format dynamic linearization linear data model through an improved projection algorithm to obtain pseudo partial derivative matrix estimation; constructing a partial format dynamic linearization disturbance observer, estimating the gain of the partial format dynamic linearization disturbance observer based on pseudo derivative matrix estimation to obtain an observer gain matrix, and updating the partial format dynamic linearization disturbance observer; and based on a partial format dynamic linearization disturbance observer, in combination with a finite set model prediction framework, predicting the output of an inverter system in a switching tube state, and selecting the switching tube state with the minimum cost function to control the inverter. According to the invention, the method can achieve the precise prediction control of an inverter system, reduces the dependence on the load parameters of the inverter, and effectively resists the negative impact caused by the drift of the load parameters.

Description

Technical Field

[0001] This application relates to the technical field of inverter control, and particularly relates to a predictive control method for an inverter based on dynamic linearization. Background Art

[0002] With the exponential growth of the computing performance of microprocessors, the engineering implementation of advanced control strategies in power electronic devices has become feasible. It is worth noting that FCS-MPC (Finite Control-Set Model Predictive Control) technology has become a research hotspot in the field of power electronic control in recent years due to its unique multi-objective optimization ability, fast dynamic response, and multi-variable coordinated control advantages, and has received extensive attention in both academia and industry.

[0003] As a control strategy that depends on the model of the controlled object, the performance of FCS-MPC highly depends on the accuracy of parameters. However, under the influence of environmental factors such as temperature and humidity, parameters such as the resistance and inductance of the controlled object may drift near their nominal values, resulting in a mismatch between the nominal values stored in the controller and the actual values. This mismatch phenomenon has a significant impact on the control effect of FCS-MPC, thereby leading to the deterioration of control performance. Summary of the Invention

[0004] To solve the deficiencies of the prior art, the following technical solutions are adopted in this application: A predictive control method for an inverter based on dynamic linearization provided by this application, the predictive control method includes the following steps: Represent the inverter system as a local partial format dynamic linearization linear data model; Based on the output of the inverter system, estimate the linear part in the local partial format dynamic linearization linear data model through an improved projection algorithm to obtain a pseudo partial derivative matrix estimate; Construct a partial format dynamic linearization disturbance observer, which is used to estimate the non-linear term in the local partial format dynamic linearization linear data model, and estimate the gain of the partial format dynamic linearization disturbance observer based on the pseudo partial derivative matrix estimate to obtain an observer gain matrix; Update the partial format dynamic linearization disturbance observer based on the pseudo partial derivative matrix estimate and the observer gain matrix; Based on the updated partial format dynamic linearization disturbance observer, combined with the finite set model prediction framework, predict the output of the inverter system under all switch tube states; Based on the output of the inverter system predicted, with the goal of minimizing the cost function, select the switching tube states that minimize the cost function to control the inverter.

[0005] In summary, a predictive control method for an inverter based on dynamic linearization provided by this application constructs the inverter system into a local partial-form dynamic linearization linear data model. Based on the historical output data of the inverter system, the linear part of the local partial-form dynamic linearization linear data model is estimated by an improved projection algorithm, and the nonlinear part of the local partial-form dynamic linearization linear data model is estimated by a partial-form dynamic linearization disturbance observer. The changes of the inverter system are completely captured based on the combination of the improved projection algorithm and the partial-form dynamic linearization disturbance observer, and the partial-form dynamic linearization disturbance observer is updated. Based on the updated partial-form dynamic linearization disturbance observer, combined with the finite set model predictive framework, the output prediction of the inverter system under all switching tube states is completed. The switching tube states of the inverter are evaluated using the cost function, and the switching tube states that minimize the cost function are selected to control the inverter, realizing precise predictive control of the inverter system, avoiding the dependence on inverter load parameters of traditional control methods, and effectively resisting the negative impact caused by load parameter drift.

[0006] Further, the predictive control method further includes constructing the local partial-form dynamic linearization linear data model through the following steps: Define the part of the inverter system that is not modeled as dynamic and external disturbances , and perform an inner product process on the pseudo-partial derivative matrix the data matrix input to the inverter system and then add it to the dynamic and external disturbances to obtain the local partial-form dynamic linearization linear data model.

[0007] Further, the predictive control method further includes: Design an ideal partial-form dynamic linearization disturbance observer; Convert the ideal partial-form dynamic linearization disturbance observer into an equivalent DL data model; Assume that the prediction error at the next moment is 0 and substitute it into the equivalent DL data model to obtain a practical partial-form dynamic linearization disturbance observer, which is used to estimate the nonlinear part in the local partial-form dynamic linearization linear data model.

[0008] Further, the ideal partial-form dynamic linearization disturbance observer is represented by the following formula: ; In the formula, h$k_1$ represents the gain of the ideal partial-format dynamic linearization disturbance observer, e p ( k ) represents the estimation error of the output quantity of the inverter system at the $k$-th moment, e p ( k + 1) represents the observation error of the ideal partial-format dynamic linearization disturbance observer at the $(k + 1)$-th moment, y ( k ) represents the output quantity of the inverter system at the $k$-th moment, z $\hat{y}_1( k ) represents the estimated value of the output quantity of the inverter system at the $k$-th moment, z $\hat{d}_2( k ) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the $k$-th moment, z $\hat{y}_1( k + 1)$ represents the estimated value of the output quantity of the inverter system at the $(k + 1)$-th moment, z $\hat{d}_2( k + 1)$ represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the $(k + 1)$-th moment, n z represents the end of historical estimation, n e represents the historical error order, O $f(·)$ represents an unknown nonlinear function.

[0009] Furthermore, the equivalent DL data model includes: the calculation of the change in the estimated value of the output quantity of the inverter system at the $(k + 1)$-th moment $\Delta\hat{y}_1$, and the calculation of the change in the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the $(k + 1)$-th moment $\Delta\hat{d}_2$; Among them, according to the ideal partial-format dynamic linearization disturbance observer, the change $\Delta\hat{y}_1$ is obtained in the following way: after taking the inner product of the pseudo partial derivative matrix and the data matrix input by the inverter system, adding it to the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the $k$-th moment z $\hat{d}_2( k )$, and then adding the result of the product of the observer gain $k_1$ and the estimation error e p ( k ) of the output quantity of the inverter system at the $k$-th moment; The change $\Delta\hat{d}_2$ is obtained by multiplying the gain matrix of the partial-format dynamic linearization disturbance observer by the change in the estimation error of the output quantity of the inverter system Obtained by the inner product processing of the composed matrix.

[0010] Further, assume that the prediction error at the next moment is 0 and substitute it into the equivalent DL data model to obtain a practical partial-format dynamic linearization disturbance observer, including the following steps: Make the observation error of the ideal partial-format dynamic linearization disturbance observer at the (k + 1)-th moment e p ( k +1) be 0, so as to convert the change amount of the inverter system output quantity estimation error composed of the matrix into a correction error matrix , and substitute the correction error matrix into the equivalent DL data model to obtain the practical partial-format dynamic linearization disturbance observer; where, the correction error matrix is represented in the following form: ; In the formula, L e represents the length of the correction error matrix, e p ( k ) represents the observation error of the ideal partial-format dynamic linearization disturbance observer at the k-th moment.

[0011] Further, the gain of the partial-format dynamic linearization disturbance observer is estimated by the following formula: ; In the formula, represents the gain matrix estimation of the partial-format dynamic linearization disturbance observer at the k-th moment, represents the penalty factor.

[0012] Further, the improved projection algorithm is represented by the following formula: ; In the formula, represents the estimated value of the pseudo partial derivative matrix at the k-th moment, is the initial value; η represents the step size factor, and η ∈ (0, 2); μ represents the penalty factor for the change of the pseudo partial derivative matrix estimation, and μ > 0.

[0013] Further, the predictive control method further includes: based on the pseudo partial derivative matrix estimation and the observer gain matrix, updating the practical partial-format dynamic linearization disturbance observer, and based on the updated practical partial-format dynamic linearization disturbance observer, combining with the finite set model prediction framework, predicting the output of the inverter system under all switch tube states.

[0014] Further, the cost function is used to characterize the reference output current at the (k + 1)-th moment and the predicted value of the output of the inverter system at the (k + 1)-th moment The deviation between them, where the predicted value of the output of the inverter system at the (k + 1)-th moment is obtained by predicting the output of the inverter system under all switch states through the updated practical partial form dynamic linearization disturbance observer in combination with the finite set model prediction framework. Description of the Drawings

[0015] Figure 1 is a flowchart of the steps of the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application; Figure 2 is a flowchart of the steps of estimating the nonlinear part in the local partial form dynamic linearization linear data model in the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application; Figure 3 is a schematic diagram of the control process of the inverter system within one control cycle in the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application; Figure 4a is a schematic diagram of the simulation experiment results of the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application under the condition that the filter inductance L L = 10 mH and other system parameters are matched; Figure 4b is a schematic diagram of the simulation experiment results of the traditional control method provided as a comparative example under the condition that the filter inductance L L = 10 mH and other system parameters are matched; Figure 5a is a schematic diagram of the simulation experiment results of the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application under the condition that the filter inductance L L = 8 mH and other system parameters are mismatched; Figure 5b is a schematic diagram of the simulation experiment results of the traditional control method provided as a comparative example under the condition that the filter inductance L L = 8 mH and other system parameters are mismatched; Figure 6 is a control block diagram of the inverter predictive control method based on dynamic linearization provided by an embodiment of the present application. Detailed Embodiments

[0016] The following will describe the present application in detail in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application, and any structural, method, or functional transformation made by those of ordinary skill in the art based on these embodiments is included within the protection scope of the present application.

[0017] To address the deficiencies of the prior art, as Figure 1 shown, the embodiment of the present application provides a predictive control method for an inverter based on dynamic linearization. The predictive control method includes the following steps: Step S101: Represent the inverter system as a locally partial-form dynamic linearized linear data model.

[0018] Step S102: Based on the output of the inverter system, estimate the linear part in the locally partial-form dynamic linearized linear data model through an improved projection algorithm to obtain a pseudo-partial derivative matrix estimate.

[0019] Step S103: Construct a partial-form dynamic linearized disturbance observer, which is used to estimate the non-linear term in the locally partial-form dynamic linearized linear data model. Estimate the gain of the partial-form dynamic linearized disturbance observer based on the pseudo-derivative matrix estimate to obtain an observer gain matrix.

[0020] Step S104: Update the partial-form dynamic linearized disturbance observer based on the pseudo-partial derivative matrix estimate and the observer gain matrix.

[0021] Step S105: Based on the updated partial-form dynamic linearized disturbance observer, combined with the finite set model predictive framework, predict the output of the inverter system under all switch states.

[0022] Step S106: Based on the predicted output of the inverter system, with the goal of minimizing the cost function, select the switch state that minimizes the cost function to control the inverter.

[0023] Specifically, in step S101, the inverter system is represented as a locally partial-form dynamic linearized linear data model. The locally partial-form dynamic linearized linear data model dynamically describes the behavior of the inverter system through historical input-output data, reducing the dependence on the physical parameters of the inverter. The locally partial-form dynamic linearized linear data model decomposes the output change of the inverter system into a linear part and a non-linear part. The linear part is represented by the product of the input data matrix and the pseudo-partial reciprocal matrix of the inverter system, and the non-linear part includes unmodeled dynamics and external disturbances. Through the locally partial-form dynamic linearized linear data model, the inverter system can be abstracted into a linear dynamic structure that only depends on historical data, thus avoiding the dependence on parameter accuracy in traditional physical modeling.

[0024] After the construction of the local partial format dynamic linearization linear data model, in step S102, based on the output data of the inverter system, the linear part in the local partial format dynamic linearization linear data model is estimated by an improved projection algorithm. The estimation process can adopt an adaptive update algorithm. By comparing the difference between the actual output change and the model prediction value, the matrix parameters are dynamically adjusted using the output error at the current moment, so as to obtain the pseudo partial derivative matrix estimation. The pseudo partial derivative matrix can characterize the linear influence degree of the input signal of the inverter system on the output change. Further, the improved projection algorithm can update the pseudo partial derivative matrix through iterative calculation of historical input and output data, providing a reliable linear part estimation for subsequent control.

[0025] In step S103, a partial format dynamic linearization disturbance observer is constructed. The partial format dynamic linearization disturbance observer is configured to estimate the non - linear terms in the local partial format dynamic linearization linear data model. The partial format dynamic linearization disturbance observer unifies the unmodeled dynamics and external disturbances of the inverter system as collective disturbances, and estimates the collective disturbances through the observer structure. Further, based on the pseudo partial derivative matrix, the gain of the partial format dynamic linearization disturbance observer is estimated to obtain the observer gain matrix, avoiding empirical parameter tuning methods such as pole placement, and simplifying the preliminary preparation work in the actual application process.

[0026] After obtaining the pseudo partial derivative matrix estimation and the observer gain matrix, in step S104, based on the above - mentioned pseudo partial derivative matrix estimation and observer gain matrix, the partial format dynamic linearization disturbance observer is iteratively updated, integrating the latest linear part estimation and gain matrix into the observer model, so as to calculate the observer state at the next moment, enabling the partial format dynamic linearization disturbance observer to accurately reflect the current state characteristics of the inverter system.

[0027] In step S105, based on the updated partial format dynamic linearization disturbance observer, combined with the finite - set model prediction framework, all possible switching - tube states of the inverter system are traversed, and the output prediction values of the inverter system at the next moment for each switching - tube state are calculated. The prediction process is based on the current observer state, the pseudo partial derivative matrix, and the input data of the inverter system, and is realized through linear superposition and disturbance compensation. Through prediction calculation, the dynamic evaluation of all candidate switching - tube states can be completed within a single control cycle, and the output predictions corresponding to all switching - tube states can be obtained in advance, providing data support for the optimal control decision.

[0028] After completing the output prediction of the inverter system for all switch states, in step S106, based on the predicted output of the inverter system, the deviation between the predicted output corresponding to the inverter system in each switch state and the reference value is evaluated through a cost function. By traversing all candidate switch states and comparing their cost function values, the switch state that minimizes the cost function value is quickly selected as the control input to control the inverter, ensuring the dynamic response speed of the control strategy. By real-time optimization, the harmonic distortion and switching losses of the inverter are reduced, and the efficient and stable operation of the inverter is achieved.

[0029] According to the above description, a predictive control method for an inverter based on dynamic linearization provided by an embodiment of the present application constructs the inverter system as a locally partial-form dynamic linearized linear data model. Based on the historical output data of the inverter system, the linear part of the locally partial-form dynamic linearized linear data model is estimated through an improved projection algorithm, and the nonlinear part of the locally partial-form dynamic linearized linear data model is estimated through a partial-form dynamic linearized disturbance observer. The changes of the inverter system are completely captured based on the combination of the improved projection algorithm and the partial-form dynamic linearized disturbance observer, and the partial-form dynamic linearized disturbance observer is updated. Based on the updated partial-form dynamic linearized disturbance observer, combined with the finite set model prediction framework, the output prediction of the inverter system for all switch states is completed. The switch states of the inverter are evaluated using a cost function, and the switch state that minimizes the cost function is selected to control the inverter, realizing precise predictive control of the inverter system, avoiding the dependence of traditional control methods on inverter load parameters, and effectively resisting the negative impact caused by load parameter drift.

[0030] As an optional implementation manner, in step S101, the predictive control method further includes: defining the unmodeled part of the inverter system as dynamic and external disturbances, performing an inner product operation on the pseudo-partial derivative matrix and the data matrix input to the inverter system, and then comparing it with the dynamic and external disturbances to obtain a locally partial-form dynamic linear data model. In one embodiment, the locally partial-form dynamic linearized linear data model can be represented by the following formula: ; In the formula, , representing the output change amount of the inverter system; , representing the data matrix input to the inverter system, L is a positive constant, representing the linearization length; , representing the pseudo-partial derivative matrix, which is the linear part of the locally partial-form dynamic linearized linear data model; k represents the time; represents the unmodeled dynamic and external disturbances of the inverter system, which is the nonlinear part of the locally partial-form dynamic linearized linear data model.

[0031] As an alternative implementation, in step S102, the anti-interference ability and estimation ability of the control system are improved by modifying the projection algorithm. The modified projection algorithm can be expressed by the following formula: ; In the formula, represents the estimated value of the pseudo partial derivative matrix at the k-th moment, is the initial value; η represents the step size factor, and η ∈ (0, 2); μ represents the penalty factor for the change of the estimated pseudo partial derivative matrix, and μ is greater than 0.

[0032] As an alternative implementation, the predictive control method estimates the non-linear part in the local partial format dynamic linearized linear data model through the following steps, as Figure 2 shown, specifically including: Step S201, design an ideal partial format dynamic linearized disturbance observer.

[0033] Step S202, convert the ideal partial format dynamic linearized disturbance observer into an equivalent DL data model.

[0034] Step S203, assume that the prediction error at the next moment is 0, and substitute it into the equivalent DL data model to obtain a practical partial format dynamic linearized disturbance observer, which is used to estimate the non-linear part in the local partial format dynamic linearized linear data model.

[0035] Specifically, for the local partial format dynamic linearized linear data model, an ideal partial format dynamic linear disturbance observer is designed, which can accurately estimate the non-linear part in the local partial format dynamic linearized linear data model. For example, during the inverter control process, when the load inductance drifts due to temperature change, the ideal partial format dynamic linear disturbance observer can capture the resulting unmodeled dynamics and provide a compensation basis for subsequent control. By using the input and output data of the inverter system to construct the observer, the dependence on the physical parameters of the inverter system is reduced, so as to be applicable to the inverter system with time-varying parameters and non-linear characteristics.

[0036] Based on the ideal partial format dynamic linear disturbance observer, by replacing the non-linear function in the ideal partial format dynamic linear disturbance observer with a time-varying vector, the non-linear dynamics of the ideal partial format dynamic linear disturbance observer is transformed into a linear parameterized form, thereby converting the ideal partial format dynamic linear disturbance observer into an equivalent DL data model. By converting the ideal partial format dynamic linearized disturbance observer into an equivalent DL data model, the complex non-linear observation problem is transformed into a linear parameter optimization model, reducing the computational complexity.

[0037] Set the prediction error at the next moment to 0, substitute the prediction error at the next moment into the equivalent DL data model for calculation, and derive a practical partial-format dynamic linearization disturbance observer. The practical partial-format dynamic linearization disturbance observer is configured to estimate the non-linear part in the local partial-format dynamic linearization linear data model. The practical partial-format dynamic linearization disturbance observer can achieve the disturbance estimation of the non-linear part by using historical error data to solve the causality constraint problem of the inverter system. Exemplarily, during the inverter control process, when the load parameters suddenly change and cause model mismatch, the practical partial-format dynamic linearization disturbance observer can comprehensively capture the disturbance dynamics by expanding the weighted combination of historical errors, improve the estimation accuracy, provide accurate compensation information for predictive control, and effectively improve the robustness and control accuracy of the system.

[0038] As an alternative implementation, in the above step S201, the designed ideal partial-format dynamic linearization disturbance observer can be expressed by the following formula: ; In the formula, h 1 represents the gain of the ideal partial-format dynamic linearization disturbance observer, e p ( k ) represents the estimation error of the output quantity of the inverter system at the k-th moment, e p ( k + 1) represents the observation error of the ideal partial-format dynamic linearization disturbance observer at the (k + 1)-th moment, y ( k ) represents the output quantity of the inverter system at the k-th moment, z 1( k ) represents the estimated value of the output quantity of the inverter system at the k-th moment, z 2( k ) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the k-th moment, z 1( k + 1) represents the estimated value of the output quantity of the inverter system at the (k + 1)-th moment, z 2( k + 1) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment, n z indicates the end of historical estimation, n e represents the order of historical error, O (·) represents an unknown non-linear function.

[0039] As an alternative implementation, in the above step S202, the ideal partial-format dynamic linearization disturbance observer is converted into an equivalent DL data model, and the equivalent DL data model includes: the change in the estimated value of the output quantity of the inverter system at the (k + 1)-th moment calculation, and the calculation of the change in the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment calculation.

[0040] Among them, according to the ideal partial-format dynamic linearization disturbance observer, the change is obtained in the following way: the pseudo partial derivative matrix is inner product processed with the data matrix input to the inverter system, and then added to the estimated value z 2( k ) of the unmodeled dynamics and external disturbances of the inverter system at the k-th moment, and then added to the result of the product between the observer gain k1 and the estimated error e p ( k ) of the output quantity of the inverter system at the k-th moment. The observer gain k1 can be set according to empirical values.

[0041] The change is obtained by inner product processing of the matrix composed of the gain matrix of the partial-format dynamic linearization disturbance observer and the change of the estimated error of the output quantity of the inverter system.

[0042] In one embodiment, the equivalent DL data model can be represented by the following formula: ; In the formula, , represents the gain matrix of the partial-format dynamic linearization disturbance observer; represents the change in the estimated value of the output quantity of the inverter system at the (k + 1)-th moment; represents the change in the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment; represents the matrix composed of the change in the estimated error of the output quantity of the inverter system, , ; represents the gain of the partial-format dynamic linearization disturbance observer.

[0043] As an alternative implementation, in the above step S203, the predicted error at the next moment is set to 0, that is , and can be obtained, and substituting it into the above equivalent DL data model, a practical partial-format dynamic linearization disturbance observer can be obtained, including the following steps: Make the observation error of the ideal partial format dynamic linearization disturbance observer at the (k + 1)-th moment e p ( k +1) be 0, so as to convert the change of the estimation error of the output quantity of the inverter system The matrix composed of into a corrected error matrix Substitute the corrected error matrix into the equivalent DL data model to obtain a practical partial format dynamic linearization disturbance observer; among them, the corrected error matrix ; In the formula, represents the length of the corrected error matrix, e p ( k ) represents the observation error of the ideal partial format dynamic linearization disturbance observer at the k-th moment.

[0044] Furthermore, the practical partial format dynamic linearization disturbance observer can be expressed by the following formula: ; In the formula, represents the corrected error matrix; represents the change of the estimated value of the output quantity of the inverter system at the (k + 1)-th moment; represents the change of the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment; represents the gain of the partial format dynamic linearization disturbance observer; z 2( k ) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the k-th moment.

[0045] Furthermore, for the gain matrix of the partial format dynamic linearization disturbance observer, consider the following criterion function: ; In the formula, represents the gain matrix of the partial format dynamic linearization disturbance observer, represents the penalty factor. The first term of the criterion function enables the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment to track the actual unmodeled dynamics and external disturbances of the inverter system, and the second term is the penalty term for the change of the gain matrix of the partial format dynamic linearization disturbance observer.

[0046] As an alternative implementation, in step S103, the above criterion function expression is optimized such that , the following formula can be obtained to estimate the gain of the partial-format dynamic linearization disturbance observer: ; In the formula, represents the estimated gain matrix of the partial-format dynamic linearization disturbance observer at the k-th moment, represents the penalty factor.

[0047] As an alternative implementation, after obtaining the estimated pseudo-partial derivative matrix and the gain matrix of the partial-format dynamic linearization disturbance observer, in step S104, based on the estimated pseudo-partial derivative matrix and the gain matrix of the observer, the practical partial-format dynamic linearization disturbance observer is updated, and the updated practical partial-format dynamic linearization disturbance observer can be expressed by the following formula: ; In the formula, represents the estimated value of the pseudo-partial derivative matrix at the k-th moment.

[0048] As an alternative implementation, in step S105, the predictive control method further includes: based on the estimated pseudo-partial derivative matrix and the observer gain matrix, updating the practical partial-format dynamic linearization disturbance observer, and based on the updated practical partial-format dynamic linearization disturbance observer, combining the finite set model predictive framework to predict the output of the inverter system under all switch states.

[0049] Among them, based on the above updated partial-format dynamic linearization disturbance observer, combining the finite set model predictive framework, the output of the inverter system under all switch states can be predicted by the following formula: ; In the formula, represents the predicted value of the output change of the inverter system, represents the predicted value of the output of the inverter system at the (k + 1)-th moment.

[0050] Based on the predicted output of the inverter system under all switch states, substituting all the prediction results into the cost function to obtain the cost function value of the inverter system under each switch state. The magnitude of the cost function value can reflect the distance between the predicted output voltage and the reference output voltage. With the goal of minimizing the cost function, the switch state that minimizes the cost function is selected to control the inverter system, thereby completing the control work of the inverter system within one control cycle.

[0051] As an alternative implementation, the cost function is used to characterize the reference output current at the (k + 1)-th moment The deviation from the predicted value of the output of the inverter system at the (k + 1)-th moment wherein, the predicted value of the output of the inverter system at the (k + 1)-th moment is obtained by predicting the output of the inverter system under all switch states by combining a finite set model predictive framework with an updated practical partial format dynamic linearization disturbance observer; the reference output current at the (k + 1)-th moment can be calculated by extrapolation. The extrapolation calculation method is as follows: ; wherein, represents the reference output current at the k-th moment, and and so on.

[0052] In one embodiment, the cost function can be expressed by the following formula: ; In the formula, represents the reference output current at the (k + 1)-th moment, represents the cost function value.

[0053] According to the above description, based on a predictive control method for an inverter based on dynamic linearization provided by an embodiment of the present application, the control process of the inverter system within one control cycle is as Figure 3 shown. Step 1: First, initialize the control system parameters, including initializing the historical input data matrix, historical error matrix, observer gain matrix, and pseudo partial derivative matrix. Step 2: Detect the output of the inverter system, compare it with the observer output, obtain the prediction error to update the historical error matrix; and record the input voltage of the inverter system to update the historical input data matrix. Step 3: Detect the output current of the inverter system and update the pseudo partial derivative matrix. Step 4: Based on the updated pseudo partial derivative matrix, update the observer gain matrix. Step 5: Based on the updated observer gain matrix, update the observer to obtain the estimated value of the updated output of the inverter system and the estimated value of the unmodeled dynamics and external disturbances of the inverter system. Step 6: Predict the future behavior of the updated estimated value of the unmodeled dynamics and external disturbances of the inverter system, the updated historical input data matrix, the updated pseudo partial derivative matrix, and the control input voltage to obtain the output prediction of the inverter system under all switch states. Step 7: Calculate through the cost function to obtain the inverter switch state that minimizes the cost function, and apply the control input voltage corresponding to this switch state to the control system to complete the control of the inverter system within one control cycle. Repeat steps 2 - 7 in each subsequent control cycle to complete the control management of the inverter.

[0054] As an alternative implementation, in the embodiments of the present application, the proposed predictive control method can be replaced with continuous control set predictive control. In continuous control set predictive control, the control quantity can be selected within a continuous range, no longer limited to finite discrete values, further improving the smoothness and accuracy of the system output, and reducing switching losses and harmonic distortion. In continuous set predictive control, steps 1 to 6 above remain unchanged. In step 7, the control strategy is improved by changing the original method of finitely discretely selecting the cost function value to directly performing space vector modulation on the control variable within the continuous space range to achieve predictive control of the inverter.

[0055] To further illustrate a predictive control method for an inverter based on dynamic linearization provided by the embodiments of the present application, the following simulation experiments are conducted to verify the effectiveness of the predictive control method. In the simulation experiment, a neutral point clamped three-level inverter is used, and the simulation parameters are shown in Table 1:

[0056] Table 1 In the case where the filter inductance L L0 = 10 mH and other system parameters are matched, the experimental results of the predictive control method provided by the embodiments of the present application are as Figure 4a shown, and the experimental results of the traditional control method are as Figure 4b shown. Under the traditional control method, the total harmonic distortion (THD) of the output current of the inverter is 1.76%. Under the predictive control method provided by the embodiments of the present application, the THD of the output current of the inverter is 1.65%. By comparing the THD index of the output current of the inverter, it can be seen that the relative reduction of the predictive control method provided by the embodiments of the present application compared to the traditional control method reaches 6.25%. It can be known that the predictive control method provided by the embodiments of the present application has better harmonic suppression ability than the traditional control method. In the case where the filter inductance L L0 = 8 mH and other system parameters are mismatched, the experimental results of the predictive control method provided by the embodiments of the present application are as Figure 5a shown, and the experimental results of the traditional control method are as Figure 5b shown. Under the traditional control method, the THD of the output current of the inverter rises to 3.47%. Under the predictive control method provided by the embodiments of the present application, the THD of the output current of the inverter remains at 2.31%. By comparing the THD index of the output current of the inverter, it can be seen that the predictive control method provided by the embodiments of the present application achieves a 33.43% performance improvement compared to the traditional control method.

[0057] The comparative analysis of the implementation results shows that under the traditional control method, the current quality of the inverter has a significant dependence on the matching degree of the controller parameters, and its THD index shows an obvious deterioration trend with the increase of the parameter adaptation degree. The predictive control method provided by the embodiments of the present application effectively overcomes the defect of parameter sensitivity by introducing an adaptive compensation mechanism, making the predictive control method have stronger robustness and engineering applicability, and being more suitable for application scenarios with parameter perturbation or measurement errors.

[0058] According to the above description, as Figure 6 shown, based on the predictive control method provided by the embodiments of the present application, based on the output of the inverter system, first, the linear part in the local partial format dynamic linearization linear data model is estimated by an improved projection algorithm, and the nonlinear part of the local partial format dynamic linearization linear data model is estimated by a partial format dynamic linearization disturbance observer. The changes of the inverter system are completely captured based on the combination of the improved projection algorithm and the partial format dynamic linearization disturbance observer, and the partial format dynamic linearization disturbance observer is updated. Based on the updated partial format dynamic linearization disturbance observer, combined with the finite set model predictive framework, the output prediction of the inverter system under all switch tube states is completed. The cost function is used to evaluate the switch tube states of the inverter, and the switch tube state that minimizes the cost function is selected to control the inverter, realizing precise predictive control of the inverter system, avoiding the dependence of the traditional control method on the inverter load parameters, and effectively resisting the negative impact caused by the load parameter drift.

[0059] It can be understood that the term "exemplary" used herein means "as an example, illustration, or explanation". Any embodiment described as "exemplary" is not necessarily superior to or better than other embodiments and / or does not exclude combining the features of other embodiments. It should be understood that certain features of the present application described in the context of separate embodiments may also be provided in combination in a single embodiment. Conversely, the various features of the present application described in the context of a single embodiment may also be provided separately or in any suitable combination or as any other described embodiment of the present application.

[0060] In the description of the present application, unless otherwise specified, " / " means "or". For example, A / B may represent A or B. The "and / or" herein is merely a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, "at least one" means one or more, and "a plurality" means two or more. The words such as "first" and "second" do not limit the quantity and execution order, and the words such as "first" and "second" do not necessarily limit to be different.

[0061] The above-disclosed are only the preferred embodiments of the present application, but they are not intended to limit the scope of the rights of the present application. Those of ordinary skill in the art can understand that within the spirit and scope of the present application and the appended claims, changes, modifications, substitutions, combinations, and simplifications should all be equivalent replacement methods and still fall within the scope covered by the invention.

Claims

1. A predictive control method for an inverter based on dynamic linearization, characterized in that, The predictive control method includes the following steps: Represent the inverter system as a locally partial-form dynamic linearized linear data model; Based on the output of the inverter system, estimate the linear part in the locally partial-form dynamic linearized linear data model through an improved projection algorithm to obtain a pseudo partial derivative matrix estimate; Construct a partial-form dynamic linearized disturbance observer, which is used to estimate the non-linear term in the locally partial-form dynamic linearized linear data model, and estimate the gain of the partial-form dynamic linearized disturbance observer based on the pseudo partial derivative matrix estimate to obtain an observer gain matrix; Update the partial-form dynamic linearized disturbance observer based on the pseudo partial derivative matrix estimate and the observer gain matrix; Based on the updated partial-form dynamic linearized disturbance observer, combined with the finite set model prediction framework, predict the output of the inverter system under all switch states; Based on the predicted output of the inverter system, with the goal of minimizing the cost function, select the switch state that minimizes the cost function to control the inverter.

2. The predictive control method for an inverter based on dynamic linearization according to claim 1, characterized in that The predictive control method further includes constructing the locally partial-form dynamic linearized linear data model through the following steps: Define the unmodeled part of the inverter system as the dynamics and external disturbances , and perform an inner product operation on the pseudo-partial derivative matrix and the data matrix of the input of the inverter system , and then add the result to the dynamics and external disturbances to obtain the local partial-format dynamic linearization linear data model.

3. The predictive control method for an inverter based on dynamic linearization according to claim 2, wherein The predictive control method further includes: Design an ideal partial-form dynamic linearized disturbance observer; Convert the ideal partial-form dynamic linearized disturbance observer into an equivalent DL data model; Assume that the prediction error at the next moment is 0 and substitute it into the equivalent DL data model to obtain a practical partial-form dynamic linearized disturbance observer, which is used to estimate the non-linear part in the locally partial-form dynamic linearized linear data model.

4. The predictive control method for an inverter based on dynamic linearization according to claim 3, characterized in that The ideal partial-form dynamic linearized disturbance observer is represented by the following formula: ； Wherein, h 1 represents the gain of the ideal partial-format dynamic linearization disturbance observer, e p ( k ) represents the estimation error of the output of the inverter system at the k-th moment, e p ( k + 1) represents the observation error of the ideal partial-format dynamic linearization disturbance observer at the (k + 1)-th moment, y ( k ) represents the output of the inverter system at the k-th moment, z 1( k ) represents the estimated value of the output of the inverter system at the k-th moment, z 2( k ) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the k-th moment, z 1( k + 1) represents the estimated value of the output of the inverter system at the (k + 1)-th moment, z 2( k + 1) represents the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment, n z represents the historical estimation order, n e represents the historical error order, O (·) represents an unknown nonlinear function.

5. The predictive control method for an inverter based on dynamic linearization according to claim 4, characterized in that The equivalent DL data model includes: the calculation of the variation of the estimated value of the output quantity of the inverter system at the (k + 1)-th moment, and the calculation of the variation of the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment; and the calculation of the variation of the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the (k + 1)-th moment; and the calculation; Among them, according to the ideal partial format dynamic linearization disturbance observer, the variation is obtained in the following way: after performing an inner product operation on the pseudo partial derivative matrix and the data matrix of the inverter system input , adding the result to the estimated value of the unmodeled dynamics and external disturbances of the inverter system at the k-th moment z 2( k ), and then adding the product of the observer gain k1 and the estimated error of the output of the inverter system at the k-th moment e p ( k ); Variation By performing inner product processing on the matrix composed of the gain matrix of the partial-format dynamic linearization disturbance observer and the variation of the estimation error of the output quantity of the inverter system 6. The predictive control method of the inverter based on dynamic linearization according to claim 5, wherein The step of assuming that the prediction error at the next moment is 0 and substituting it into the equivalent DL data model to obtain a practical partial-form dynamic linearized disturbance observer includes the following steps: Make the observation error of the ideal partial format dynamic linearization disturbance observer at the (k + 1)-th moment e p ( k + 1) be 0, so as to convert the change amount of the estimation error of the output quantity of the inverter system The matrix composed of is converted into a correction error matrix , and substitute the correction error matrix into the equivalent DL data model to obtain the practical partial format dynamic linearization disturbance observer; wherein, the correction error matrix is represented in the following form: ； In the formula, L e represents the length of the correction error matrix, e p ( k ) represents the observation error of the ideal partial format dynamic linearization disturbance observer at the k-th moment.

7. The predictive control method of the inverter based on dynamic linearization according to claim 6, characterized in that Estimate the gain of the partial-form dynamic linearized disturbance observer through the following formula: ； In the formula, represents the gain matrix estimation of the partial format dynamic linearization disturbance observer at the k-th moment, represents the penalty factor.

8. The predictive control method for an inverter based on dynamic linearization according to claim 4, characterized in that The improved projection algorithm is represented by the following formula: ； wherein, represents the estimated value of the pseudo partial derivative matrix at the k-th moment, is the initial value; η represents the step size factor, and η ∈ (0, 2); μ represents the penalty factor for the change in the estimated pseudo partial derivative matrix, and μ is greater than 0.

9. The predictive control method for an inverter based on dynamic linearization according to claim 3, wherein The predictive control method further includes: Based on the pseudo partial derivative matrix estimate and the observer gain matrix, update the practical partial-form dynamic linearized disturbance observer, and based on the updated practical partial-form dynamic linearized disturbance observer, combined with the finite set model prediction framework, predict the output of the inverter system under all switch states.

10. The inverter predictive control method based on dynamic linearization according to claim 9, characterized in that The cost function is used to characterize the reference output current at the (k + 1)-th moment and the predicted value of the output of the inverter system at the (k + 1)-th moment The deviation between them, where the predicted value of the output of the inverter system at the (k + 1)-th moment is obtained by predicting the output of the inverter system under all switch states through the updated practical partial form dynamic linearization disturbance observer in combination with the finite set model prediction framework.

Images