Inverter multi-objective optimization method based on sequential model predictive control

Abstract

The invention provides an inverter multi-objective optimization method based on sequential model predictive control, which comprises the following steps: adopting a hierarchical prediction and screening mechanism, calculating voltage tracking costs only for 27 states in the first layer, and sorting the voltage tracking costs; the second layer only needs to calculate midpoint balance cost for partial states and sort the midpoint balance cost; and the third layer only needs to calculate the common-mode voltage cost of partial states. The method has the advantages that weight coefficients are completely abandoned, and controller design is carried out by adopting candidate set number parameters (N1, N2) with clear physical significance. And a three-step layered optimization structure of output voltage tracking, neutral-point potential balancing and common-mode voltage suppression is constructed, the logic is rigorous, and the engineering safety and performance requirements are met. And high-performance cooperative control is realized: the output voltage quality, the operation safety and the electromagnetic compatibility of the system are synchronously improved on the premise of ensuring the real-time performance of the algorithm.

Description

Technical Field

[0001] This invention relates to the field of power electronics, and specifically to a multi-objective optimization method and system for a T-type three-level inverter based on sequential model predictive control. Background Technology

[0002] With the rapid development of new energy power generation, industrial transmission, and power quality control, the performance requirements for power electronic converters are increasing. While traditional two-level voltage source inverters are simple in structure and easy to control, they suffer from problems such as high switching losses, high output voltage harmonic content, and severe electromagnetic interference in medium- and high-voltage, high-power applications. To overcome these limitations, multi-level inverter technology has emerged.

[0003] Three-level inverters, as a typical representative of multi-level technology, significantly improve the output voltage waveform quality, reduce the voltage stress on switching devices, and decrease the size and cost of output filters by increasing the number of output levels. Among them, the T-type three-level inverter, due to its unique topology, combines the advantages of diode-clamped three-level inverters and flying capacitor three-level inverters: while maintaining the same number of output levels, it requires fewer power devices, has lower conduction losses, and eliminates the need for complex capacitor voltage balancing circuits, showing promising application prospects in fields such as photovoltaic grid connection, uninterruptible power supplies, and motor drives.

[0004] Currently, the main control strategies for three-level inverters include classic proportional-integral control, hysteresis control, space vector pulse width modulation, and model predictive control, which has emerged in recent years.

[0005] The traditional approach combining linear control with SVPWM typically employs a dual-closed-loop PI control structure with outer-loop voltage and inner-loop current, coupled with a complex space vector pulse width modulation algorithm. Its advantage lies in the fixed switching frequency, facilitating filter design. However, PI controllers are sensitive to parameter changes and exhibit slow dynamic response; the SVPWM algorithm itself is computationally complex and requires an additional independent midpoint potential balance control algorithm, further increasing system complexity and parameter tuning difficulty. Furthermore, this method struggles to directly and flexibly incorporate objectives such as common-mode voltage suppression into the control framework.

[0006] Traditional FCS-MPC (Classic FCS-MPC controller, C-MPC) typically employs a weighted summation of a single cost function. The fundamental limitation of this method lies in the design of the weighting coefficients. Optimal weighting coefficients heavily depend on specific system parameters, load conditions, and performance preferences. Engineers must rely on experience and extensive trial-and-error simulations to tune them, a time-consuming and laborious process lacking a systematic approach. Furthermore, the optimal weights tuned for a specific operating condition may no longer be optimal when system parameters change or operating points shift, leading to a degraded controller performance.

[0007] Disadvantages of existing technology: When the traditional FCS-MPC (Classic FCS-MPC controller, C-MPC) is applied to a T-type three-level inverter, it mainly traverses 27 basic voltage vectors in each cycle and selects the optimal voltage vector according to the cost function. The weighting coefficient design is complicated, debugging is difficult, and engineering practicality is poor. Summary of the Invention

[0008] The objective of this invention is achieved through the following technical solutions.

[0009] The present invention aims to propose a three-step sequential model predictive voltage control method specifically designed for T-type three-level inverters.

[0010] The innovations of this invention are: (1) It adopts a three-layer sequential optimization structure, completely abandoning the complex method of processing multiple objectives through a single weighted cost function in traditional model predictive control, and innovatively constructing a step-by-step, layered optimization structure. (2) It adopts a layered prediction and screening mechanism. In each control cycle, the algorithm does not completely calculate all three objective functions for all 27 states. The first layer only calculates and sorts the voltage tracking cost for the 27 states; the second layer only calculates and sorts the midpoint balance cost for N1 states; and the third layer only calculates the common-mode voltage cost for N2 states. This "screening-focusing" mechanism makes the computational load decrease step by step as the optimization level deepens.

[0011] Specifically, this invention provides a multi-objective optimization method for inverters based on sequential model predictive control, including: A hierarchical prediction and screening mechanism is adopted. The first layer calculates and sorts the voltage tracking cost for only 27 states of the inverter; the second layer calculates and sorts the midpoint balance cost for only some states; and the third layer calculates the common-mode voltage cost for only some states.

[0012] The inverter is a three-phase T-type three-level grid-connected inverter.

[0013] The advantages of this invention are: It completely abandons weighting coefficients and uses the physically meaningful "candidate set size" parameters (N1, N2) for controller design. It establishes clear control priorities: constructing a three-step hierarchical optimization structure of "output voltage tracking → midpoint potential balance → common-mode voltage suppression," which is logically rigorous and meets engineering safety and performance requirements. It achieves high-performance collaborative control: while ensuring the real-time performance of the algorithm, it simultaneously improves the system's output voltage quality, operational safety, and electromagnetic compatibility. Attached Figure Description

[0014] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Appendix Figure 1 A structural diagram of a T-type three-level inverter according to an embodiment of the present invention is shown.

[0015] Appendix Figure 2 A flowchart of a multi-objective optimization method for a T-type three-level inverter based on sequential model predictive control according to an embodiment of the present invention is shown.

[0016] Appendix Figure 3 The diagram shows the current and output voltage waveforms under a resistive load according to an embodiment of the present invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] Terminology Explanation: T-Type Three-Level Inverter: A high-performance three-level topology with four power switches per phase arm, capable of outputting +V... DC / 2, 0, -V DC / 2 offers three voltage levels. Compared to traditional two-level and neutral-point clamped three-level topologies, it has advantages such as low conduction loss, good switching characteristics, and low output voltage harmonic content. However, its control requires simultaneously handling multiple coupled objectives such as output voltage tracking, neutral point potential balance, and common-mode voltage suppression, resulting in high control complexity.

[0019] Finite Control Set Model Predictive Control (FCS-MPC) is an advanced control strategy particularly suitable for power electronic converters. Its core principle lies in leveraging the finite number of switching states of the converter. Within each control cycle, it predicts the future system behavior under all possible combinations of switching states using a discrete model of the system, and directly selects the optimal switching state to apply to the converter based on a predefined cost function. This method offers fast dynamic response and is naturally suited for digital control.

[0020] Classical Model Predictive Control (C-MPC): The most traditional implementation strategy under the FCS-MPC framework. It handles multiple control objectives through a weighted summation of a single cost function. The main drawback of this strategy is that the design of the weight coefficients is complex and lacks physical intuitiveness, requiring extensive trial-and-error simulations, resulting in a high barrier to engineering application.

[0021] Sequential Model Predictive Control (S-MPC) is an advanced multi-objective predictive control strategy. Its core lies in abandoning the complex and difficult-to-adjust weight coefficients of traditional model predictive control, replacing them with a clear hierarchical optimization structure. Control objectives are prioritized, and in each level, only a limited number of candidate switching states selected from the previous level are evaluated, progressively approaching the optimal solution. This method is intuitive in design and can effectively coordinate conflicts between multiple control objectives.

[0022] A. System Model The controlled object of this invention is a three-phase T-type three-level grid-connected inverter, whose topology is as follows: Figure 1 As shown. The system mainly consists of a DC power supply and DC-side voltage divider support capacitors ( , The topology consists of a T-type three-level power bridge arm, an LC-type filter, and the power grid. Each phase arm of the T-type three-level topology consists of four IGBTs (S1, S2, S3, S4, S5, S6, S7, S8, S9, S1 ...1, S1, S1, x1 S x2 S x3 S x4 The four transistors (x=a, b, c) and anti-parallel diodes form a "T"-shaped structure. By controlling the on and off combinations of these four transistors, the voltage at each phase output terminal relative to the DC side midpoint O can present three levels: positive level... Zero level 0, negative level . This indicates the DC power supply voltage. A three-phase inverter has a total of 3... 3 =27 effective switching vector combinations.

[0023] Compared to traditional two-level inverters, three-level T-type topologies have lower switching stress, better harmonic characteristics, and higher efficiency, making them widely used in medium-power applications.

[0024] To simplify modeling and analysis, the following assumptions are made for the three-level T-type inverter: (1) Power switching devices are ideal devices; (2) DC side capacitor And the voltage remains balanced. ); (3) The load current changes slowly over two consecutive sampling periods and can be replaced by an estimated value.

[0025] (4) The grid voltage is a three-phase symmetrical sine wave.

[0026] B. Sequential Model Predictive Control Algorithm Control Method To simplify analysis and facilitate controller design, the model is mapped to a two-phase stationary state using an amplitude-invariant Clarke transform. Coordinate system, where Clark transformation matrix for: At rest In the coordinate system, the forward Euler method is used to discretize the system (sampling period). The core equation used for prediction is obtained as follows: In the formula: This is the inverter-side inductor current; therefore... for time Inverter-side inductor current in coordinate system, subscripts in other parameters of the formula The same applies; This is the voltage of the filter capacitor (i.e., the output load voltage). This is the load current; The inverter output voltage vector is determined by the switching state. For filtering inductors, This is a filter capacitor.

[0027] Since the load current changes slowly over two consecutive sampling periods, an estimation method based on equation (3) is obtained: To compensate for the computational delay in digital control, the control algorithm needs to predict... The output voltage at time t. This needs to be achieved through an iterative model: First, use equations (1)(2)(3) to predict The state at any given moment: Then, based on the possible switching states Predict the corresponding Current at any moment: Ultimately, we obtained Predicted output voltage at any given time: C. Control Process S-MPC completely abandons complex weighting coefficients and adopts a step-by-step, layered filtering mechanism to optimize multiple objectives. For example... Figure 2 As shown, the algorithm proceeds in the following three stages: The first stage involves iterating through all 27 switch vectors. The corresponding output voltage can be predicted using formulas (6) and (7). Calculate the voltage tracking cost function for each vector. : In the formula: , for shaft and Reference output voltage on the axis; Predicted when using the l-th switch vector Shaft output voltage; Predicted when using the l-th switch vector Shaft output voltage.

[0028] according to Sort all 27 switch vectors by value from smallest to largest, and select... The smallest value The switch vectors constitute the first-level candidate set. .in It is an adjustable parameter of the algorithm that determines the number of candidates that enter the next round of screening; For the first A switch vector.

[0029] In the second stage, the algorithm uses the first candidate set that already meets the voltage tracking performance requirements. Within, a secondary screening is conducted with the goal of achieving midpoint potential balance. For For each candidate state, calculate the midpoint potential balance cost. : In the formula: Capacitor on the DC side The voltage; Capacitor on the DC side The voltage; Then, according to value pairs Sort the states in ascending order and select the ones with the lowest cost. states ( This constitutes the second-level candidate set. .parameter This determines the number of states to be allocated for common-mode optimization in the final stage while taking into account the midpoint balance.

[0030] In the third stage, the algorithm selects a second-level candidate set that simultaneously satisfies the first two objectives. The final selection is made with common-mode voltage rejection as the objective. For For each candidate state, calculate its common-mode voltage cost. : In the formula: For the first The output voltage of A relative to the DC side midpoint O under each switching vector; For the first The output voltage of B relative to the DC side midpoint O under each switching vector; For the first The output voltage of C relative to the DC side midpoint O under a switching vector.

[0031] Ultimately, from Select to make The switch state with the smallest value is taken as the global optimal solution. And applied to inverters.

[0032] This invention utilizes a simulation model built on the Matlab platform to verify the correctness of the theory. To verify the effectiveness of the method, simulation tests were conducted under balanced resistive load conditions.

[0033] Figure 3 The current and output voltage waveforms of this invention under resistive load are shown. As can be seen from the output voltage waveform, the waveform quality meets general requirements in steady state.

[0034] S-CMP offers a slightly lower computational burden while meeting real-time performance requirements, provided the THD condition is met. Compared to C-MPC, it eliminates the need for cumbersome weight parameter tuning, significantly improving its practicality in engineering applications.

[0035] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-objective optimization method for inverters based on sequential model predictive control, characterized in that, include: Calculate the voltage tracking cost for the switching vectors corresponding to all states of the inverter, and sort all switching vectors in ascending order based on the voltage tracking cost, selecting the first one with the lowest voltage tracking cost. The switch vectors constitute the first-level candidate set; Calculate the midpoint equilibrium cost for the switch vectors in the first-level candidate set. Sort the switch vectors in the first-level candidate set from smallest to largest according to the midpoint equilibrium cost, and select the one with the lowest cost. A switch vector, This constitutes the second-level candidate set; The common-mode voltage cost is calculated for the switching vectors in the second-level candidate set, and the switching vector with the minimum common-mode voltage cost is selected as the global optimal solution and applied to the inverter.

2. The multi-objective optimization method for inverters based on sequential model predictive control according to claim 1, characterized in that, The inverter is a three-phase T-type three-level grid-connected inverter.

3. A multi-objective optimization method for inverters based on sequential model predictive control according to claim 1 or 2, characterized in that, Before calculating the voltage tracking cost, at rest In the coordinate system, the forward Euler method is used to discretize the system, with a sampling period of... The core equation used for prediction is obtained as follows: In the formula: This refers to the inverter-side inductor current. This is the voltage across the filter capacitor, i.e., the output load voltage. This is the load current; The inverter output voltage vector is determined by the switching state. For filtering inductors, This is a filter capacitor.

4. The multi-objective optimization method for inverters based on sequential model predictive control according to claim 3, characterized in that, After obtaining the core equation for prediction, the process further includes: Based on equation (3), the load current is estimated as follows: A multi-objective optimization method for inverters based on sequential model predictive control according to claim 4, characterized in that, Following the estimation of the load current, the process further includes: predict The state at any given moment: A multi-objective optimization method for inverters based on sequential model predictive control according to claim 5, characterized in that, In the prediction Following the state at a given moment, it further includes: Based on inverter switching state Predict the corresponding Current at any moment: get Predicted output voltage at any given time: 。 5. The multi-objective optimization method for inverters based on sequential model predictive control according to claim 6, characterized in that, The voltage tracking cost is calculated for the switching vectors corresponding to all states of the inverter, and all switching vectors are sorted from smallest to largest according to the voltage tracking cost, selecting the first one with the lowest voltage tracking cost. A set of switch vectors constitutes the first-level candidate set, including: Traverse all 27 switch vectors The corresponding output voltage can be predicted using formulas (6) and (7). For each vector, calculate its voltage tracking cost function. : In the formula: , for shaft and Reference output voltage on the axis; Predicted when using the l-th switch vector Shaft output voltage; Predicted when using the l-th switch vector Shaft output voltage; according to Sort all 27 switch vectors by value from smallest to largest, and select... The smallest value The switch vectors constitute the first-level candidate set. ;in It is an adjustable parameter that determines the number of candidates that will proceed to the next round of screening; For the first A switch vector.

6. The multi-objective optimization method for inverters based on sequential model predictive control according to claim 7, characterized in that, The process involves calculating the midpoint balance cost for the switch vectors within the first-level candidate set, sorting the switch vectors in the first-level candidate set from smallest to largest based on the midpoint balance cost, and selecting the top vectors with the lowest cost. A switch vector, This constitutes the second-level candidate set, including: In the first-level candidate set that already meets voltage tracking performance Within, a secondary screening is conducted with the goal of achieving midpoint potential balance; for For each candidate state, calculate the midpoint potential balance cost. : In the formula: Capacitor on the DC side The voltage; Capacitor on the DC side The voltage; Then, according to value pairs Sort the states in ascending order and select the ones with the lowest cost. A state, This constitutes the second-level candidate set. ;parameter The decision is made to allocate the number of states for common-mode optimization in the final stage while taking into account the midpoint balance.

7. The multi-objective optimization method for inverters based on sequential model predictive control according to claim 8, characterized in that, The process of calculating the common-mode voltage cost for the switching vectors in the second-level candidate set, selecting the switching vector with the minimum common-mode voltage cost as the global optimal solution, and applying it to the inverter includes: The second-level candidate set that simultaneously satisfies the first two objectives is superior. Internally, the final selection is made with common-mode voltage rejection as the objective; for For each candidate state, calculate its common-mode voltage cost. : In the formula: For the first The output voltage of A relative to the DC side midpoint O under each switching vector; For the first The output voltage of B relative to the DC side midpoint O under each switching vector; For the first The output voltage of C relative to the DC side midpoint O under a switching vector.

8. Finally, from Select to make The switch state with the smallest value is taken as the global optimal solution. And applied to inverters.

Images