A model-free predictive control method for parallel T-type three-level rectifier

Abstract

The application provides a model-free predictive control method for a parallel T-type three-level rectifier. The method adopts double closed-loop control, the outer loop adopts adaptive droop control with voltage feedforward to realize the stability of DC voltage, high power factor and load power distribution; the inner loop adopts model-free predictive control, first, discrete mathematical models of the inner loop control target grid current, midpoint voltage and circulating current are established; on this basis, the model-free predictive control implementation methods of the three control targets are designed, wherein the synchronous updating method is adopted to update the current difference; finally, the multi-target optimization of the inner loop model-free predictive control is realized. The model-free predictive control of the application is based on data driving and does not depend on any hardware parameters and circuit models, and has strong parameter robustness; the multi-target model-free predictive control of the inner loop grid current, midpoint voltage and circulating current of the parallel system is realized; the synchronous updating method is adopted to update the current difference data, and simultaneously, due to the relationship between the circulating current and the common-mode voltage, the updating frequency of the current difference is faster than that of a single rectifier. The overall scheme realizes the multi-target optimization control of the parallel system and greatly improves the parameter robustness of the system.

Description

Technical Field

[0001] This invention belongs to the field of power electronic converter control technology, specifically relating to a model-free predictive control method for parallel T-type three-level rectifiers. Background Technology

[0002] Currently, with the widespread application of new energy vehicles, large-scale DC charging technology is increasingly becoming the development trend of new energy vehicle charging. Furthermore, high-power rectifiers are needed in railway locomotive traction, mine hoisting systems, and wind power generation systems. The superior power density and current harmonics of T-type three-level rectifiers have led to their wider application in high-power applications. When a single T-type three-level rectifier cannot meet the load power demand, a parallel structure is usually used to further increase the power level. However, parallel T-type three-level rectifiers face challenges related to power distribution, DC voltage stability, power factor, midpoint voltage balance, and circulating current suppression.

[0003] Model predictive control (MDI) predicts state variables using a system model and selects the optimal solution by designing a cost function. It offers significant advantages in complex, multi-objective control systems. However, the effectiveness of MDI depends on the accuracy of the system model, and it struggles to function effectively in situations of parameter mismatch or unknown circuit relationships. Existing methods for improving the robustness of MDI parameters are computationally complex and often focus on model-free control methods based on current tracking, neglecting the multi-objective optimization capabilities of model-free control. Therefore, it is necessary to research model-free predictive control methods applicable to parallel rectifier systems to achieve multi-objective control of parallel systems and improve the parameter robustness of MDI.

[0004] Droop control is a common method for achieving rectifier power distribution, DC voltage stability, and a high power factor. Its basic principle is to linearly reduce the voltage as the output current increases. However, the VI droop control curve dictates that droop control will inevitably lead to DC voltage droop. Therefore, it is necessary to study an adaptive droop control method to reduce DC voltage droop. Summary of the Invention

[0005] To address the aforementioned shortcomings in the prior art, this invention provides a model-free predictive control method for parallel T-type three-level rectifiers.

[0006] To achieve the above-mentioned objectives, the overall technical solution adopted in this invention is a dual closed-loop control. The outer loop uses adaptive droop control with voltage feedforward to achieve power distribution, high power factor, and stable DC voltage. The inner loop uses model-free predictive control to achieve midpoint voltage balance, circulating current suppression, and grid current tracking.

[0007] A model-free predictive control method for parallel T-type three-level rectifiers, the specific details of which include the adaptive droop control with outer loop voltage feedforward are as follows:

[0008] DC voltage droop control controls the active power input to the DC network by detecting the difference between the DC voltage and a reference voltage, thereby achieving power balance and voltage stability. This invention employs VI characteristic droop control, which can be expressed as:

[0009]

[0010] In the formula U dc , For DC voltage and its reference value, I dc , K represents the direct current and its reference value. d Let K be the droop coefficient of the VI characteristic. d The droop factor is changed to an adaptive one, typically setting the system operating reference point to: when the transmission power is 0, the DC bus voltage is the system rated voltage; that is, setting... The rated voltage is used. After adding an adaptive droop factor, the VI characteristic droop control is changed to:

[0011]

[0012] In the formula, P N Representing the rated power of the rectifier, the droop factor changes with the DC current, achieving adaptive droop control. When two rectifiers are connected in parallel, the principle of droop control for power distribution is as follows: Figure 1 As shown.

[0013] Due to inherent flaws in the droop control principle, the DC voltage will exhibit a certain degree of droop. To improve this droop characteristic, a feedforward is added based on the difference between the DC voltage and the reference value to compensate for the impact of droop control on the DC voltage.

[0014] A model-free predictive control method for parallel T-type three-level rectifiers, wherein the inner loop model-free predictive control comprises the following steps:

[0015] S1. Establish a discretized mathematical model of the grid current, midpoint voltage, and circulating current of the inner loop control target of the parallel T-type three-level rectifier, and determine the model-free predictive control method for the inner loop control target.

[0016] S2. Sample the grid current i at time k. n1 (k), Switch state S n1(k), the DC side capacitor voltage is calculated and the midpoint voltage deviation Δv1(k) is obtained. The sampled grid current is transformed from the three-phase coordinate system to the two-phase rotating coordinate system, and the current difference Δi at time k is calculated using the grid current in the two-phase rotating coordinate system and the delay element. 1m |V(k);

[0017] S3. Based on the current difference sampled at time k and the switching state, update the current difference data table. According to the model-free predictive control method of current tracking, update the current difference data caused by the switching state of the coaxial line in the phase tables of the 27 switching vectors.

[0018] S4. Based on the collected data and the updated current difference data table, predict the current difference Δi caused by different switching vectors at time k+1. 1m (k+1), Substitute into the current tracking cost function g 1j Calculate and sort the current error in the two-phase rotating coordinate system caused by different switching vectors. Based on the sorted switching vectors, select the optimal switching vector that simultaneously takes into account current tracking, midpoint voltage balance, and circulating current suppression.

[0019] Furthermore, the establishment of a discretized mathematical model of the grid current for the parallel T-type three-level rectifier to achieve model-free control for grid current tracking specifically includes the following steps:

[0020] The two rectifiers are identical in model structure except for their transmission power and hardware parameters; therefore, the first rectifier will be used as an example for analysis. The mathematical model of the parallel T-type three-level rectifier in a two-phase rotating coordinate system is as follows:

[0021]

[0022] In the formula i 1d i 1q The current is the dq-axis grid current of the first rectifier, L1 is the filter inductance, R1 is the filter parasitic resistance, and e d e q For the dq axis grid voltage, u 1d u 1q ω is the input voltage of the rectifier, and ω is the angular velocity of the dq axis.

[0023] Discretizing the established mathematical model of the grid current for a T-type three-level rectifier using Euler's formula yields the discrete model of a parallel T-type three-level rectifier as follows:

[0024]

[0025] In the formula, (k) and (k+1) are the sampling times, and T sThe sampling period is [value]. Model-free predictive control for grid current tracking is established based on current difference data. The formula for calculating the dq-axis grid current difference at time k and time k+1 is [formula].

[0026]

[0027] In the formula Δi 1m (m = d, q) represents the grid current difference along the dq axis, i 1m Let be the dq-axis component of the grid current. The grid current at time k+1 in Equation 5 cannot be sampled, so the current difference at time k+1 cannot be calculated. However, with a sufficiently high sampling frequency, the current difference caused by different switching vectors at time k+1 can be replaced by old data stored in the data table. Therefore, the grid current prediction equation for different switching vectors at time k+1 is expressed as:

[0028] i 1m (k+1)|V j =i 1m (k)+Δi 1m,old |V j V j ∈{V1,...,V 27 Formula Six

[0029] In the formula i 1m (k+1)|V j The switching vector V at time k+1 j The predicted value of the grid current along the dq axis, Δi 1m，old |V j The switch vector V stored in the data table j The resulting current difference, i 1m (k) represents the sampled value of the grid current at time k. The cost function for grid current tracking is:

[0030]

[0031] In the formula Let i be the reference value of the dq-axis grid current at time k+1. 1d (k+1)|V j i 1q (k+1)|V j The switch vector V predicted at time k+1 j This results in the dq-axis grid current. The grid current reference value is obtained based on the adaptive droop control with inner loop voltage feedforward.

[0032] Furthermore, the establishment of a discretized mathematical model of the midpoint voltage of the parallel T-type three-level rectifier to achieve model-free control of the midpoint voltage balance specifically includes the following steps:

[0033] The mathematical model for the DC side midpoint voltage deviation is:

[0034]

[0035] In the formula i np i is the current flowing into the neutral point. n1 (n = A, B, C) represents the three-phase grid current, S n1 In three-phase switch state, C 11 ΔV1 is the DC-side capacitor, and ΔV1 is the DC-side midpoint voltage deviation.

[0036] The discretized midpoint voltage deviation is as follows:

[0037]

[0038] As can be seen from Equation 9, in i np1 When ×ΔV1(k)≤0, the midpoint voltage will be balanced. Based on the relationship between the current flowing into the neutral point and the switching state, model-free predictive control of the midpoint voltage can be achieved.

[0039] Furthermore, the establishment of a mathematical model for the zero-sequence circulating current of the parallel T-type three-level rectifier, and the realization of model-free control for circulating current suppression, specifically includes the following steps:

[0040] According to Kirchhoff's law for zero-sequence circular paths, we can obtain:

[0041]

[0042] In the formula V Nx (x = 1, 2) represents the voltage of the lower DC capacitor, U nxOx R is the voltage between the DC side midpoint and the rectifier input point. x L is the parasitic resistance of the filter inductor of the two rectifiers. x For the filter inductors of the two rectifiers, i nx (n = A, B, C) represents the three-phase grid current of the two rectifiers. V Nx and U nxOx They are represented as follows:

[0043]

[0044] In the formula U cmvx The common-mode voltage of the two rectifiers is ΔV. x This represents the voltage deviation at the midpoint between the two rectifiers. According to the definition of zero-sequence circulating current, zero-sequence circulating current can be expressed as:

[0045]

[0046] In the formula i zThis is a zero-sequence circulation. Combining equations 10, 11, and 12, we can obtain the mathematical model of the circulation as follows:

[0047]

[0048] Equation 13 shows that the main influencing factors of zero-sequence circulating current are the midpoint voltage and common-mode voltage difference between the two rectifiers. Based on the relationship between common-mode voltage and switching state, and the model-free predictive control of midpoint voltage, model-free predictive control of circulating current can be achieved.

[0049] Furthermore, the specific steps for updating the grid current difference are as follows:

[0050] The mathematical model of the current difference shows that it consists of the natural response of the zero-switching vector and the forced response of the non-zero vector.

[0051] in

[0052] In the formula δi 1d0 ,δi 1q0 The natural response caused by the zero vector, δi 1d |V j ,δi 1q |V j The forced response caused by a non-zero vector, Δi 1d |V j ,Δi 1q |V j For the switching vector V j The resulting current difference, u 1d |V j u 1q |V j For the switching vector V j The resulting rectifier input voltage. Combined with the phase diagram of the 27 switch vectors ( Figure 2 By using this method, the current difference relationship of a set of switching vectors under in-phase conditions can be obtained. During the current difference update process, the current difference sampled at time k and the switching state can determine the switching vector current difference data that needs to be updated. Based on the phase relationship of the switching vectors, a set of in-phase switching vectors is updated. Thus, the current difference data updated at each time point is no longer a single value but a set, greatly improving the current difference update frequency. Furthermore, due to the relationship between the common-mode voltage and the switching state, the optimal switching vectors during steady-state operation are basically limited to seven switching vectors that make the common-mode voltage zero. These seven switching vectors belong to four different phase relationship groups, further improving the current difference update frequency.

[0053] Furthermore, the specific steps for selecting the optimal solution using the model-free predictive control are as follows:

[0054] The execution flowchart of the model-free predictive control is as follows: Figure 3 As shown, the current tracking error caused by 27 switching vectors is calculated and sorted according to the cost function of current tracking. The sorted switching vectors are then iterated and selected until the optimal switching vector is found. The specific steps for selecting the optimal switching vector are as follows: First, determine the position of the switching vector in the sorted switching vectors. The position after sorting represents the magnitude of the current tracking error; the earlier the position, the smaller the current tracking error. Under the condition of satisfying current tracking, determine whether the midpoint voltage balance and circulating current suppression are satisfied. If they are satisfied, the optimal solution is selected; if not, the selection condition is relaxed, and the switching vector that satisfies circulating current suppression is selected as the optimal solution. If circulating current suppression is not satisfied, determine whether the current midpoint voltage deviation meets the requirements. If it does, the loop continues; if not, the switching vector that satisfies midpoint balance at the expense of some circulating current suppression is selected as the optimal solution. The selection of the optimal switching vector comprehensively considers the three control objectives of inner-loop current tracking, midpoint voltage balance, and circulating current suppression.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] (1) This invention employs a dual-closed-loop control strategy, allowing voltage control and current control to have different time scales and be implemented in different controllers. DC voltage stabilization, power distribution, and high power factor are implemented in the outer loop, simplifying the control objectives of the inner loop's model-predictive control. The dual-closed-loop control of this invention achieves multi-objective optimization control of a parallel T-type three-level rectifier.

[0057] (2) This invention establishes a model-free predictive control method with three control objectives: inner loop current tracking, midpoint voltage balance, and circulating current suppression. It achieves optimal solution selection for the inner loop without using weighting factors. The model-free predictive control method of this invention does not require any prior knowledge of hardware parameters or circuit models, greatly improving the parameter robustness of the system.

[0058] (3) This invention employs a synchronized current difference update method, which updates the current difference between different switching vectors in the same phase based on the phase relationship of the switching vectors, thereby improving the update frequency of the current difference. In addition, due to the relationship between the common-mode voltage and the switching state, the number of switching vector update groups of the parallel T-type three-level rectifier during steady-state operation is reduced from 7 groups to 4 groups, further improving the update frequency of the current difference. Attached Figure Description

[0059] Figure 1 The principle of droop control power distribution in this invention;

[0060] Figure 2 The 27 switch vector phase diagrams in this invention;

[0061] Figure 3 The flowchart of the model-free predictive control for selecting the optimal solution in this invention;

[0062] Figure 4 The topology of two parallel T-type three-level rectifiers in this invention;

[0063] Figure 5 The control block diagram of the two parallel T-type three-level rectifier system in this invention;

[0064] Figure 6 The present invention contains seven sets of current difference update simulation diagrams; Figure (a) shows the four groups of seven switching vectors that make the common-mode voltage zero, and Figure (b) shows the other three groups that are basically not involved in the selection of the optimal solution under steady state.

[0065] Figure 7 The present invention shows the waveforms of DC voltage and AC power changes when the voltage reference value changes; wherein Figure (a) shows the difference between the DC voltage and the reference value, and Figure (b) shows the AC power.

[0066] Figure 8 Figure 1 shows the experimental comparison between model-free predictive control and existing technologies under parameter adaptation conditions in this invention; Figure 2 shows the grid current control effect and Figure 3 shows the circulating current suppression and midpoint voltage balance control effect. Detailed Implementation

[0067] The technical solution will be clearly and completely described below with reference to preferred embodiments and accompanying drawings. It should be understood that the preferred embodiments are merely illustrative of the invention and not intended to limit the scope of protection of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0068] This invention provides a model-free predictive control method for parallel T-type three-level rectifiers. The embodiment of this invention is used in a parallel system of two T-type three-level rectifiers with different power ratings. The overall technical solution adopted in this invention is a dual-closed-loop control. The outer loop uses adaptive droop control with voltage feedforward to achieve power distribution, high power factor, and stable DC voltage. The inner loop uses model-free predictive control to achieve midpoint voltage balance, circulating current suppression, and grid current tracking.

[0069] A model-free predictive control method for parallel T-type three-level rectifiers, the specific details of which include the adaptive droop control with outer loop voltage feedforward are as follows:

[0070] DC voltage droop control controls the active power input to the DC network by detecting the difference between the DC voltage and a reference voltage, thereby achieving power balance and voltage stability. This invention employs VI characteristic droop control, which can be expressed as:

[0071]

[0072] In the formula U dc , For DC voltage and its reference value, I dc , K represents the direct current and its reference value. d Let K be the droop coefficient of the VI characteristic. d The droop factor is changed to an adaptive one, typically setting the system operating reference point to: when the transmission power is 0, the DC bus voltage is the system rated voltage; that is, setting... The rated voltage is used. After adding an adaptive droop factor, the VI characteristic droop control is changed to:

[0073]

[0074] In the formula, P N Representing the rated power of the rectifier, the droop factor changes with the DC current, achieving adaptive droop control. When two rectifiers are connected in parallel, the principle of droop control for power distribution is as follows: Figure 1 As shown.

[0075] Due to inherent flaws in the droop control principle, the DC voltage will exhibit a certain degree of droop. To improve this droop characteristic, a feedforward is added based on the difference between the DC voltage and the reference value to compensate for the impact of droop control on the DC voltage.

[0076] A model-free predictive control method for parallel T-type three-level rectifiers, wherein the inner loop model-free predictive control comprises the following steps:

[0077] S1. Establish a discretized mathematical model of the grid current, midpoint voltage, and circulating current of the inner loop control target of the parallel T-type three-level rectifier, and determine the model-free predictive control method for the inner loop control target.

[0078] S2. Sample the grid current i at time k. n1 (k), Switch state S n1 (k), the DC side capacitor voltage is calculated and the midpoint voltage deviation ΔV1(k) is obtained. The sampled grid current is transformed from the three-phase coordinate system to the two-phase rotating coordinate system, and the current difference Δi at time k is calculated using the grid current in the two-phase rotating coordinate system and the delay element. 1m |V(k);

[0079] S3. Based on the current difference sampled at time k and the switching state, update the current difference data table. According to the model-free predictive control method of current tracking, update the current difference data caused by the switching state of the coaxial line in the phase tables of the 27 switching vectors.

[0080] S4. Based on the collected data and the updated current difference data table, predict the current difference Δi caused by different switching vectors at time k+1. 1m (k+1), Substitute into the current tracking cost function g 1j Calculate and sort the current error in the two-phase rotating coordinate system caused by different switching vectors. Based on the sorted switching vectors, select the optimal switching vector that simultaneously takes into account current tracking, midpoint voltage balance, and circulating current suppression.

[0081] Furthermore, the establishment of a discretized mathematical model of the grid current for the parallel T-type three-level rectifier to achieve model-free control for grid current tracking specifically includes the following steps:

[0082] The two rectifiers are identical in model structure except for their transmission power and hardware parameters; therefore, the first rectifier will be used as an example for analysis. The mathematical model of the parallel T-type three-level rectifier in a two-phase rotating coordinate system is as follows:

[0083]

[0084] In the formula i 1d i 1q The current is the dq-axis grid current of the first rectifier, L1 is the filter inductance, R1 is the filter parasitic resistance, and e d e q For the dq axis grid voltage, u 1d u 1q ω is the input voltage of the rectifier, and ω is the angular velocity of the dq axis.

[0085] Discretizing the established mathematical model of the grid current for a T-type three-level rectifier using Euler's formula yields the discrete model of a parallel T-type three-level rectifier as follows:

[0086]

[0087] In the formula, (k) and (k+1) are the sampling times, and T s The sampling period is [value]. Model-free predictive control for grid current tracking is established based on current difference data. The formula for calculating the dq-axis grid current difference at time k and time k+1 is [formula].

[0088]

[0089] In the formula Δi 1m (m = d, q) represents the grid current difference along the dq axis, i 1mLet be the dq-axis component of the grid current. The grid current at time k+1 in Equation 5 cannot be sampled, so the current difference at time k+1 cannot be calculated. However, with a sufficiently high sampling frequency, the current difference caused by different switching vectors at time k+1 can be replaced by old data stored in the data table. Therefore, the grid current prediction equation for different switching vectors at time k+1 is expressed as:

[0090] i 1m (k+1)|V j =i 1m (k)+Δi 1m,old |V j V j ∈{V1,...,V 27 Formula Six

[0091] In the formula i 1m (k+1)|V j The switching vector V at time k+1 j The predicted value of the grid current along the dq axis, Δi 1m，old |V j The switch vector V stored in the data table j The resulting current difference, i 1m (k) represents the sampled value of the grid current at time k. The cost function for grid current tracking is:

[0092]

[0093] In the formula Let i be the reference value of the dq-axis grid current at time k+1. 1d (k+1)|V j i 1q (k+1)|V j The switch vector V predicted at time k+1 j This results in the dq-axis grid current. The grid current reference value is obtained based on the adaptive droop control with inner loop voltage feedforward.

[0094] Furthermore, the establishment of a discretized mathematical model of the midpoint voltage of the parallel T-type three-level rectifier to achieve model-free control of the midpoint voltage balance specifically includes the following steps:

[0095] The mathematical model for the DC side midpoint voltage deviation is:

[0096]

[0097] In the formula i np1 i is the current flowing into the neutral point. n1 (n = A, B, C) represents the three-phase grid current, S n1 In three-phase switch state, C11 ΔV1 is the DC-side capacitor, and ΔV1 is the DC-side midpoint voltage deviation.

[0098] The discretized midpoint voltage deviation is as follows:

[0099]

[0100] As can be seen from Equation 9, in i np1 When ×ΔV1(k)≤0, the midpoint voltage will be balanced. Based on the relationship between the current flowing into the neutral point and the switching state, model-free predictive control of the midpoint voltage can be achieved.

[0101] Furthermore, the establishment of a mathematical model for the zero-sequence circulating current of the parallel T-type three-level rectifier, and the realization of model-free control for circulating current suppression, specifically includes the following steps:

[0102] According to Kirchhoff's law for zero-sequence circular paths, we can obtain:

[0103]

[0104] In the formula V Nx (x = 1, 2) represents the voltage of the lower DC capacitor, U nxO R is the voltage between the DC side midpoint and the rectifier input point. x L is the parasitic resistance of the filter inductor of the two rectifiers. x For the filter inductors of the two rectifiers, i nx (n = A, B, C) represents the three-phase grid current of the two rectifiers. V Nx and U nxOx They are represented as follows:

[0105]

[0106] In the formula U cmvx The common-mode voltage of the two rectifiers is ΔV. x This represents the voltage deviation at the midpoint between the two rectifiers. According to the definition of zero-sequence circulating current, zero-sequence circulating current can be expressed as:

[0107]

[0108] In the formula i z This is a zero-sequence circulation. Combining equations 10, 11, and 12, we can obtain the mathematical model of the circulation as follows:

[0109]

[0110] Equation 13 shows that the main influencing factors of zero-sequence circulating current are the midpoint voltage and common-mode voltage difference between the two rectifiers. Based on the relationship between common-mode voltage and switching state, and the model-free predictive control of midpoint voltage, model-free predictive control of circulating current can be achieved.

[0111] Table 1 shows the relationship between switching state and common-mode voltage.

[0112] Furthermore, the specific steps for updating the grid current difference are as follows:

[0113] The mathematical model of the current difference shows that it consists of the natural response of the zero-switching vector and the forced response of the non-zero vector.

[0114] in

[0115] In the formula δi 1d0 ,δi 1q0 The natural response caused by the zero vector, δi 1d |V j ,δi 1q |V j The forced response caused by a non-zero vector, Δi 1d |V j ,Δi 1q |V j For the switching vector V j The resulting current difference, u 1d |V j u 1q |V j For the switching vector V j The resulting rectifier input voltage. Combined with the phase diagram of the 27 switch vectors ( Figure 2 By using this method, the current difference relationship of a set of switching vectors under in-phase conditions can be obtained. During the current difference update process, the current difference sampled at time k and the switching state can determine the switching vector current difference data that needs to be updated. Based on the phase relationship of the switching vectors, a set of in-phase switching vectors is updated. Thus, the current difference data updated at each time point is no longer a single value but a set, greatly improving the current difference update frequency. Furthermore, due to the relationship between the common-mode voltage and the switching state, the optimal switching vectors during steady-state operation are basically limited to seven switching vectors that make the common-mode voltage zero. These seven switching vectors belong to four different phase relationship groups, further improving the current difference update frequency.

[0116] Furthermore, the specific steps for selecting the optimal solution using the model-free predictive control are as follows:

[0117] The execution flowchart of the model-free predictive control is as follows: Figure 3As shown, the current tracking error caused by 27 switching vectors is calculated and sorted according to the cost function of current tracking. The sorted switching vectors are then iterated and selected until the optimal switching vector is found. The specific steps for selecting the optimal switching vector are as follows: First, determine the position of the switching vector in the sorted switching vectors. The position after sorting represents the magnitude of the current tracking error; the earlier the position, the smaller the current tracking error. Under the condition of satisfying current tracking, determine whether the midpoint voltage balance and circulating current suppression are satisfied. If they are satisfied, the optimal solution is selected; if not, the selection condition is relaxed, and the switching vector that satisfies circulating current suppression is selected as the optimal solution. If circulating current suppression is not satisfied, determine whether the current midpoint voltage deviation meets the requirements. If it does, the loop continues; if not, the switching vector that satisfies midpoint balance at the expense of some circulating current suppression is selected as the optimal solution. The selection of the optimal switching vector comprehensively considers the three control objectives of inner-loop current tracking, midpoint voltage balance, and circulating current suppression.

[0118] Figure 4 The diagram shows a topology of two rectifiers connected in parallel, representing an effective embodiment of the present invention. Figure 5 This is a control block diagram of a parallel rectifier in an embodiment of the present invention. The execution process of this embodiment is as follows: The sampled DC voltage and DC current are processed by an adaptive droop controller with voltage feedforward and a PI regulator to obtain the d-axis reference value of the grid current. To ensure a high power factor for the rectifier, the q-axis reference value of the grid current is set to 0. The calculated grid current reference value is then sent to a model-free predictive controller. The model-free predictive controller selects the optimal switching state based on the inner-loop control objective and the model-free predictive control algorithm, and outputs it to the switching sequence of the T-type three-level rectifier.

[0119] To verify the effectiveness of the control strategy provided by this invention, two units with different power levels were used. Figure 5 Taking a parallel T-type three-level rectifier system as an example, the simulation diagram of the grid current difference update is shown below. Figure 6 As shown. When the DC-side voltage reference value and load power change, the DC voltage and AC-side power of the parallel rectifier are as follows: Figure 7 As shown. When filter parameters mismatch occurs, the effects of model-free predictive control on grid current tracking, circulating current suppression, and midpoint voltage balancing compared to existing technologies are as follows: Figure 8 As shown.

[0120] from Figure 6 (a) It can be seen that the current difference caused by the switching vectors of different groups conforms to Figure 2 Phase relationship diagram of different switching vectors. From Figure 6 (b) It can be seen that after stable operation, the current difference caused by the other three sets of switch vectors does not update, indicating that the optimal solution is selected from the seven switch combinations that make the common-mode voltage zero. From Figure 7(a) It can be seen that when the DC voltage reference value increases by 100V, the difference between the DC voltage and the reference value becomes -100V. Then, after 0.5s, the DC voltage recovers to near the DC voltage reference value, demonstrating the effectiveness of adaptive droop control with voltage feedforward for DC voltage stabilization. From Figure 7 (b) It can be seen that the two rectifiers with rated power of 5kW and 10kW respectively achieved good power distribution, and the power factors of both rectifiers were close to 1, verifying the effectiveness of adaptive droop control with voltage feedforward in power distribution and high power factor control. From Figure 8 (a) It can be seen that, under parameter adaptation conditions, traditional model predictive control exhibits poor current tracking performance and severe waveform distortion. In contrast, the model-free predictive control of this invention achieves the goal of current tracking. From Figure 8 (b) It can be seen that, under the parameter adaptation conditions, the effective value of the circulating current of the traditional model predictive control is 8.835A, while the effective value of the circulating current of the model-free predictive control of the present invention is 137.1mA. The model-free predictive control has a significant advantage in terms of circulating current suppression. The filter parameter mismatch has little effect on the balancing effect of the midpoint voltage. The control effects of the traditional model predictive control and the model-free predictive control are comparable, with a midpoint voltage deviation of 1.4V for both.

[0121] This invention has described the principles and implementation methods of the invention through specific embodiments. Those skilled in the art will recognize that the above embodiments are merely illustrative of the technical concept and features of the invention, intended to enable those skilled in the art to understand the content of the invention and implement it accordingly. They should not be construed as limiting the scope of protection of the invention. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. All equivalent changes or modifications made according to the spirit and essence of the invention should be covered within the scope of protection of the invention.

Claims

1. A model-free predictive control method for parallel T-type three-level rectifiers, characterized in that, The method includes the following steps: S1. Parallel rectifiers have a multi-objective optimization problem, including DC voltage stability, power distribution, high power factor, grid current tracking, circulating current suppression, and midpoint voltage balance. To address this, a dual closed-loop control strategy is adopted to achieve multi-objective optimization control. S2. Establish a discretized mathematical model of the grid current, midpoint voltage, and circulating current of the inner loop control target of the parallel T-type three-level rectifier, and determine the model-free predictive control method for the inner loop control target. S3. Sample the grid current at time k. Switch status The DC-side capacitor voltage was calculated, and the midpoint voltage deviation was obtained. The sampled grid current is transformed from a three-phase coordinate system to a two-phase rotating coordinate system, and the current difference at time k is calculated using the grid current in the two-phase rotating coordinate system and the time delay element. ; S4. Based on the current difference sampled at time k and the switching state, update the current difference data table. According to the model-free predictive control method of current tracking, update the current difference data caused by the switching state of the coaxial line in the phase tables of the 27 switching vectors. S5. Based on the collected data and the updated current difference data table, predict the current difference caused by different switching vectors at time k+1. Substitute into the current tracking cost function Calculate and sort the current error in the two-phase rotating coordinate system caused by different switching vectors. Based on the sorted switching vectors, select the optimal switching vector that simultaneously takes into account current tracking, midpoint voltage balance, and circulating current suppression. The specific steps for updating the S4 current difference are as follows: The mathematical model of the current difference shows that it consists of the natural response of the zero-switching vector and the forced response of the non-zero vector. In the formula, , The natural response caused by the zero vector. , The forced response caused by a non-zero vector. , Switching vector The resulting current difference; , Switching vector The rectifier input voltage caused by the current difference, combined with the phase diagram of the 27 switching vectors, allows us to derive the current difference relationship of a set of switching vectors under the same phase condition. During the current difference update process, the current difference sampled at time k and the switching state can determine the current difference data of the switching vectors that need to be updated. Based on the phase relationship of the switching vectors, a set of switching vectors in the same phase is updated. In this way, the current difference data updated at each time is no longer a single data point but a set, which greatly improves the current difference update frequency. In addition, due to the relationship between the common-mode voltage and the switching state, the optimal switching vectors during steady-state operation are basically limited to the 7 switching vectors that make the common-mode voltage 0. These 7 switching vectors belong to 4 different phase relationship groups, which further improves the current difference update frequency.

2. The model-free predictive control method for parallel T-type three-level rectifiers according to claim 1, characterized in that, The S1 dual closed-loop control specifically includes: The overall control scheme for the parallel rectifier adopts a dual-loop control. The outer loop uses adaptive droop control with voltage feedforward to achieve power distribution, high power factor, and stable DC voltage. The inner loop uses model-free predictive control to achieve midpoint voltage balance, circulating current suppression, and grid current tracking. Specifically, the sampled DC voltage and DC current are used to obtain the grid current d-axis reference value through the adaptive droop controller with voltage feedforward and the PI regulator. To ensure the high power factor of the rectifier, the grid current q-axis reference value is set to 0. The calculated grid current reference value is sent to the model-free predictive controller. The model-free predictive controller selects the optimal switching state based on the inner loop control objective and the model-free predictive control algorithm, and outputs it to the switching sequence of the T-type three-level rectifier. The overall scheme realizes multi-objective optimization control of the parallel T-type three-level rectifier.

3. The model-free predictive control method for parallel T-type three-level rectifiers according to claim 2, characterized in that, The specific steps of the adaptive droop control with voltage feedforward are as follows: The adaptive VI droop control characteristics are as follows: In the formula, , This refers to the DC voltage and its reference value. It is direct current. The rated power of the rectifier. The droop coefficient of the VI characteristic is used. The droop coefficient of the adaptive VI droop control characteristic changes with the DC current, thus realizing adaptive droop control. In addition, due to the inherent flaws in the principle of droop control, the DC voltage will droop to some extent. To improve this droop characteristic, a feedforward is added based on the difference between the DC voltage and the reference value to compensate for the impact of droop control on the DC voltage.

4. The model-free predictive control method for parallel T-type three-level rectifiers according to claim 1, characterized in that, The model-free predictive control method for S2 grid current, midpoint voltage, and circulating current is data-driven and does not rely on any hardware parameters or circuit structure. It achieves multi-objective optimization without using weighting factors and reduces the update stagnation problem of current difference. The specific steps of the inner loop model-free predictive control are as follows: The two rectifiers are identical in model structure except for their transmission power and hardware parameters; therefore, the analysis will be based on the first rectifier. The mathematical model of the parallel T-type three-level rectifier in the two-phase rotating coordinate system is as follows: In the formula, , This refers to the grid current along the dq axis of the first rectifier. For filter inductance, For the parasitic resistance of the filter, , The grid voltage is the dq axis. , The input voltage of the rectifier. The rotational angular velocity of the dq axis; Discretization using Euler's formula yields the discrete model of the parallel T-type three-level rectifier as follows: In the formula, Sampling time, To predict the next moment, The sampling period is [value missing]. The model-free predictive control for current tracking is based on current difference data. The current difference data caused by different switching vectors predicted at time k+1 is represented as follows: In the formula, Let m be the dq-axis grid current, where m = d and q. The switch vector stored in the data table The resulting current difference; For grid current The sampled value at time 10:

00. for Time-predicted switching vector The resulting grid current; given a sufficiently short sampling time, the old data stored in the data table can be used to predict the current difference caused by the switch vector at the next moment; the cost function of current tracking is: In the formula, , This is the reference value of the dq-axis grid current at time k+1; , The switch vector predicted at time k+1 The resulting dq-axis grid current; The mathematical model for the DC side midpoint voltage deviation is: In the formula, The current flowing into the neutral point, Let n be the three-phase power grid current, where n = A, B, C; It is in three-phase switch state. For DC side capacitors, This refers to the DC side midpoint voltage deviation. The discretized midpoint voltage deviation is as follows: From the discretized midpoint voltage deviation formula, it can be seen that, When the neutral point voltage is balanced, model-free predictive control of the neutral point voltage can be achieved based on the relationship between the current flowing into the neutral point and the switching state. Based on the flow path of the zero-sequence circulation, the mathematical model of the circulation is determined as follows: In the formula, It is a zero-sequence circulation. The common-mode voltage of the two rectifiers is where ;R x L is the parasitic resistance of the filter inductor of the two rectifiers. x . represents the filter inductor for two rectifiers. The formula represents the midpoint voltage deviation between the two rectifiers. It can be seen from this formula that the main influencing factors of zero-sequence circulating current are the midpoint voltage and common-mode voltage difference between the two rectifiers. Based on the relationship between common-mode voltage and switching state and the model-free predictive control of midpoint voltage, model-free predictive control of circulating current can be achieved.

5. The model-free predictive control method for parallel T-type three-level rectifiers according to claim 1, characterized in that, The specific steps for selecting the optimal switch vector in S5 are as follows: The current tracking error caused by 27 switching vectors is calculated and sorted according to the cost function of current tracking. The sorted switching vectors are iterated and selected until the optimal switching vector is selected. The specific steps for selecting the optimal switching vector are as follows: First, the position of the switching vector in the sorted switching vectors is determined. The position after sorting represents the magnitude of the current tracking error. The earlier the position, the smaller the current tracking error. Under the condition of satisfying current tracking, it is determined whether the midpoint voltage balance and circulating current suppression are satisfied. If they are satisfied, the optimal solution is selected. If not, the selection condition is relaxed, and the switching vector that satisfies circulating current suppression is selected as the optimal solution. If circulating current suppression is not satisfied, it is determined whether the current midpoint voltage deviation meets the requirements. If it does, the loop continues. If not, the switching vector that satisfies midpoint balance at the expense of some circulating current suppression is selected as the optimal solution. The selection of the optimal switching vector comprehensively considers the three control objectives of inner loop current tracking, midpoint voltage balance, and circulating current suppression.

Images