Model-free, high-robustness predictive control method and system for a 3-level SNPC inverter

Abstract

The present invention relates to the field of three-level inverters, and more specifically discloses a model-free, high-robustness predictive control method and system for a three-level SNPC inverter, which includes: performing real-time sampling and coordinate transformation on the inverter to obtain current and voltage components in a two-phase stationary coordinate system; establishing a hyperlocal model, unifying and equating system parameter deviations and external disturbances into concentrated disturbances, designing a full-order sliding mode observer and performing real-time estimation; calculating a reference voltage vector based on disturbance estimates and a reference current obtained by extrapolation; selecting the optimal dual vector combination by spatial vector optimization, determining the action time, selecting redundant small vectors according to the midpoint voltage deviation, and generating a PWM drive signal.

Description

[Technical Field]

[0001] The present invention relates to the technology of three-level inverters, and more specifically to a model-free, high-robustness predictive control method and system for three-level SNPC inverters. [Background technology]

[0002] Three-phase multilevel inverters are widely used in applications such as motor drives, grid-connected power generation, and battery energy storage systems due to their advantages, including high output voltage quality and low electromagnetic interference. Three-level SNPC converters, based on the conventional three-phase three-level topology, introduce a segmented structure of common and independent modules, maintaining three-level output capability while significantly reducing the number of active switching devices, thereby lowering system costs and losses.

[0003] Previous research has shown that finite control set model predictive control (FCS-MPC) offers advantages such as fast response and flexible constraint handling in power electronics systems. However, its control variable prediction process generally depends on the target parameters, such as resistance and inductance. If there is a discrepancy, i.e., a mismatch between the parameters used in the controller and the actual target parameters, prediction errors occur, leading to a degradation of control performance. In 2021, the IEEE TIE paper "Robust Model Predictive Control for a Three-Phase PMSM Motor With Improved Control Precision" pointed out that FCS-MPC is sensitive to parameters, and parameter mismatches lead to prediction errors in the control target, resulting in poor controllability. Experimental comparisons showed that when inductance mismatches exist, for example L'=2L, the current ripple of conventional MPCs increases significantly compared to operating conditions without mismatches. As can be seen, in scenarios such as three-level SNPC inverters, if predictive control still explicitly depends on nominal filtering parameters and equivalent impedance parameters, the discreteness of the elements, degradation due to temperature rise, and parameter drift due to changes in operating conditions inevitably lead to prediction deviations, which can be expressed as increased current tracking errors, steady-state ripple, and increased harmonic content, potentially further affecting the midpoint potential balance control effect.

[0004] Therefore, in order to improve the robustness and electrical energy quality of the system under parameter mismatch and changing operating conditions, it is necessary to propose a robustness predictive control method that can be applied to a 3-level SNPC converter and that can explicitly take into account the effects of parameter mismatch. [Overview of the project] [Problems that the invention aims to solve]

[0005] To address the problems of conventional model-based predictive control, which are highly dependent on parameter accuracy and lack robustness, this invention provides a model-free, high-robustness predictive control method and system for a 3-level SNPC inverter. By constructing an ultra-local model, the system uncertainty is concentrated in the disturbance term, estimated and compensated for in real time using an observer, and midpoint potential balance is ensured while simultaneously improving the electrical energy quality of the output current and the system robustness. [Means for solving the problem]

[0006] To solve the above technical problems, the present invention specifically provides the following technical solutions.

[0007] A model-free, high-robustness predictive control method for a 3-level SNPC inverter, comprising the following steps:

[0008] S1. Sample the three-phase output current of the inverter, the power system side voltage, and the DC side capacitor voltage, and convert them to obtain the current and voltage in a two-phase stationary coordinate system. S2. Construct a hyperlocal model of the inverter, unify the uncertainty of the system parameters into a concentrated disturbance and make it equivalent, estimate the concentrated disturbance in real time using a full-order observer, and obtain the disturbance estimate. S3. According to the current command, the reference current at a future time is predicted using extrapolation, and the current sampling current and the disturbance estimate are linked to calculate the reference voltage vector for the next sampling period based on the ultralocal model. S4. A set of candidate voltage vectors is selected based on the position of the reference voltage vector in the spatial vector diagram, the optimal combination of dual vectors is evaluated and selected using a cost function, its duration is determined, redundant small vectors are selected according to the DC-side midpoint voltage deviation, and a PWM drive signal is generated to realize modulation control.

[0009] In one preferred version of the present invention, S1 specifically means, Within each sampling period, real-time sampling is performed on the 3-level SNPC inverter, the three-phase output current is acquired via a current sensor, and the three-phase power system voltage and the DC upper and lower capacitor voltages are acquired via a voltage sensor. This includes performing Clark transforms on the three-phase output current and the three-phase power system voltage, respectively, to obtain the current component and the power system voltage component in a two-phase stationary coordinate system as input quantities for a subsequent hyperlocal model and a full-order observer.

[0010] As one preferred approach of the present invention, in S2, constructing a hyperlocal model of the inverter and unifying the uncertainty of the system parameters into a concentrated disturbance for equivalence is, specifically, In the aforementioned two-phase stationary coordinate system, filtering inductance, resistance, power system impedance, midpoint voltage deviation, and conduction voltage drop of the power device are introduced as gain coefficients, and the uncertainty of the system is concentrated as a disturbance. By unifying and equivalentizing the data to JPEG0007883334000002.jpg56, and achieving simplification of the model structure and parameter decoupling, the ultralocal model is achieved. This includes constructing JPEG0007883334000003.jpg730, Here, JPEG0007883334000004.jpg727 is the output voltage of a 3-level SNPC inverter in a two-phase stationary coordinate system. JPEG0007883334000005.jpg723 represents the output current in a two-phase stationary coordinate system, where T is the matrix transpose and α is a design parameter related to the filtering inductance, used to reduce the control algorithm's dependence on actual filtering inductance, resistance, and other parameters.

[0011] As one preferred approach of the present invention, in S2, a full-order observer is used to estimate the concentrated disturbance in real time and obtain the disturbance estimate, specifically, The method involves constructing a full-order observer based on a hyperlocal model, wherein the full-order observer uses a sliding-mode observer structure, and the input to the full-order observer is the sampling current. JPEG0007883334000006.jpg56 and output voltage JPEG0007883334000007.jpg47, and the state variables of the full-order observer include current observations and concentrated disturbances, By using a full-order observer to estimate concentrated disturbances in real time, and completing iterative updates within each sampling period using the current sample value and the observed values ​​from the previous period through discretization processing, real-time estimation with low memory demand is achieved, thereby generating disturbance estimates. This includes obtaining JPEG0007883334000008.jpg67.

[0012] In one preferred approach of the present invention, in S3, predicting the reference current at a future time using an extrapolation method according to the current command is, specifically, Current commands according to power grid connection commands or the output of the current external loop regulator. To obtain JPEG0007883334000009.jpg77, Using second-order Lagrangian extrapolation, the reference current for the next two sampling periods is determined based on the current commands for the current and the first two sampling periods. This includes predicting JPEG0007883334000010.jpg517.

[0013] In one preferred approach of the present invention, in S3, the current sampling current and the disturbance estimate are linked to calculate the reference voltage vector for the next sampling period based on the ultralocal model, specifically, Current sampling current JPEG0007883334000011.jpg612 and disturbance estimates Link JPEG0007883334000012.jpg613 to predict current Obtain JPEG0007883334000013.jpg618, and With the ultra-local model, predict the current for the next two sampling periods JPEG0007883334000014.jpg517 will track the reference current within the next two beats JPEG0007883334000015.jpg517, and calculate the reference voltage vector for the next sampling period JPEG0007883334000016.jpg67, where the reference voltage vector JPEG0007883334000017.jpg67 does not explicitly depend on the nominal parameters of the filtering inductance, resistance, and power system side impedance.

[0014] In a preferred embodiment of the present invention, in S4, based on the position of the reference voltage vector in the space vector diagram, select a set of candidate voltage vectors, evaluate and select the optimal combination of dual vectors by a cost function, and obtain its operating time. Specifically, Based on the position of the reference voltage vector in the 3-level SNPC inverter space voltage vector diagram, determine the sector and sub-triangle region where the reference voltage vector is located, and construct the large vector, small vector, and zero vector related to the sub-triangle region as a set of candidate voltage vectors, and construct a plurality of candidate line segments composed of two voltage vectors with the reference voltage vector as the target. Select a preset cost function, calculate the cost value from the reference voltage vector to the midpoint of each candidate line segment with the error between the reference voltage vector and the equivalent output voltage vector as an index, select the line segment with the smallest cost function value, and use the voltage vectors at both ends as the two operating voltage vectors for this sampling period. Based on the geometric relationship of the space vector diagram or the optimization condition for approximating the equivalent output voltage vector to the reference voltage vector, obtain the operating times T1 and T2 of the two operating voltage vectors, and include realizing two-stage modulation of the dual vectors.

[0015] As a preferred embodiment of the present invention, in S4, selecting redundant small vectors according to the DC-side midpoint voltage deviation and generating PWM drive signals to achieve modulation control specifically includes: Calculating the midpoint voltage deviation based on the DC-side upper capacitor voltage VP and the lower capacitor voltage VN. When there are redundant small vectors, on the premise of meeting the current control target, different redundant small vectors are selected according to the sign of the midpoint voltage deviation and participate in dual vector synthesis to achieve the balance of the DC-side midpoint potential; According to the selected two active voltage vectors and their action times, generating PWM drive signals for the 3-level SNPC inverter common module and the three-phase bridge arm, completing the model-free predictive control of one sampling period, and repeating the above steps in subsequent sampling periods.

[0016] The <0000****>[7]] model-free high robustness predictive control system of a 3-level SNPC inverter is used to realize the model-free high robustness predictive control method of a 3-level SNPC inverter. A sampling module that acquires the three-phase output current, the three-phase power system-side voltage, and the DC-side upper and lower capacitor voltages in real time and completes the Clark transformation to output the current components and voltage components in the two-phase stationary coordinate system; An observation module that receives the current components, voltage components, and inverter output voltage, estimates the extended state in real time by a full-order sliding mode observer constructed based on the super-local model, and outputs the current observation value and the concentrated disturbance estimation value; A reference voltage calculation module that receives the current command, current observation value, and concentrated disturbance estimation value, predicts the future reference current using an extrapolation algorithm, and calculates the reference voltage vector; A modulation module that selects a candidate voltage vector set based on the reference voltage vector position, constructs a candidate line segment, selects the optimal active voltage vector based on the cost function, and obtains its action time; It includes a control module that converts the selected operating voltage vector and its operating time into a PWM drive signal for a 3-level SNPC inverter common module and a three-phase bridge arm, and selects redundant small vectors according to the midpoint voltage deviation to achieve midpoint potential balance control.

[0017] A computer-readable storage medium storing a computer program, wherein when the computer program is invoked by a processor, it executes a model-free, high-robustness predictive control method and system for a 3-level SNPC inverter, the processor being a digital signal processor, a microcontroller, or a field-programmable gate array, used to achieve real-time predictive control of the 3-level SNPC inverter. [Effects of the Invention]

[0018] The present invention has the following beneficial effects compared to the conventional technology.

[0019] 1. By constructing an ultralocal model and a full-order sliding mode observer, the present invention unifies and equips nonlinear factors such as filtering inductance deviation, resistance deviation, uncertainty in power system impedance, midpoint voltage fluctuations, and conduction voltage drop of power devices into concentrated disturbances and estimates them in real time. Compared to conventional predictive control methods that rely on accurate parameter models, the present invention fundamentally eliminates the impact of parameter mismatches on current prediction accuracy, and maintains low current tracking errors and low total harmonic distortion even under conditions such as element degradation, temperature rise changes, and rapid changes in operating conditions, thereby significantly improving the robustness and long-term operational stability of the system.

[0020] 2. The full-order sliding mode observer used in this invention has a simplified structure and can complete the iterative updating of disturbance estimation using only the data from the current and previous sampling periods, without storing long-term historical data, thereby significantly reducing memory usage and computational complexity. Due to this characteristic, the algorithm is applicable particularly to 3-level SNPC inverter applications with high switching frequencies and high power levels, ensuring real-time control performance and dynamic response speed. [Brief explanation of the drawing]

[0021] To more clearly illustrate embodiments of the present invention or technical concepts in the prior art, the following briefly introduces the drawings that may be used in the embodiments or prior art descriptions. Obviously, the drawings in the following description are illustrative only, and those skilled in the art can obtain other implementation drawings based on the provided drawings without expending any creative effort. [Figure 1] This is a structural diagram of a 3-level SNPC inverter system. [Figure 2] This is a basic spatial vector diagram of a 3-level SNPC inverter system. [Figure 3] This is a block diagram of the predictive control of the fully-order sliding mode observer (SMO) of the present invention. [Figure 4] This is an overall control block diagram of the model-free predictive control method for a 3-level SNPC inverter system according to the present invention. [Figure 5] This is a comparison diagram of the grid-connected current waveform and THD of a normal dual-vector MPC and the method of the present invention during inductance parameter matching work, where (a) is a dual-vector predictive current control method and (b) is a model-free dual-vector predictive current control method based on SMO. [Figure 6]This is a comparison diagram of the grid-connected current waveform and THD of a conventional dual-vector MPC and the method of the present invention under inductance parameter mismatch conditions, where (a) is a conventional dual-vector predictive current control method and (b) is a model-free dual-vector predictive current control method based on SMO. [Figure 7] The present invention relates to the three-phase current waveform output when the load of a three-level SNPC inverter changes rapidly. [Figure 8] This shows the fluctuations in the midpoint voltage during the operation period of the 3-level SNPC inverter using the method of the present invention. [Modes for carrying out the invention]

[0022] The following clearly and completely describes the technical concepts in the embodiments of the present invention, linking them to the drawings of the embodiments. Clearly, the embodiments described are only a subset of the embodiments of the present invention, not all embodiments. All other embodiments obtained based on the embodiments of the present invention without the creative effort of a person skilled in the art are all within the scope of the protection of the present invention.

[0023] The following explains the concepts related to this application, first by linking them to the drawings. It should be noted that the following explanations of each concept are merely for the purpose of making the content of this application easier to understand, and do not represent limitations on the scope of protection of this application. Furthermore, the embodiments and features in this application can be combined with each other, as long as they do not conflict. The following will describe this application in detail, referring to the drawings and linking them to the embodiments.

[0024] Example 1 As shown in Figures 3-8, the present invention provides a model-free, high-robustness predictive control method for a 3-level SNPC inverter, which is based on a mathematical model of the 3-level SNPC inverter and a hyperlocal model. By introducing JPEG0007883334000018.jpg732, filtering inductance / resistance deviation, power system voltage disturbances, dead zones, and device non-idealisms are concentrated disturbances. The files will be unified and absorbed into JPEG0007883334000019.jpg56. The control law was estimated online at the time of implementation using the design parameter α and the observer. By using only JPEG0007883334000020.jpg66, explicit reliance on nominal L, R, and other electrical parameters is avoided, and if parameter drift occurs, its effects are... This method is compensated by real-time updates of JPEG0007883334000021.jpg56, and finally, dual-vector two-stage modulation is combined to achieve current predictive control and midpoint potential balance. This method is applied to the 3-level SNPC grid-connected inverter system shown in Figure 1.

[0025] Specifically, this includes the following steps:

[0026] S1. The three-phase output current, power system voltage, and DC capacitor voltage of the inverter are sampled and converted to obtain the current and voltage in a two-phase stationary coordinate system, specifically including the following:

[0027] Each sampling period Within JPEG0007883334000022.jpg55, the three-phase output current is transmitted via a Hall current sensor. Sample JPEG0007883334000023.jpg533 and the voltage from the three-phase power system side via the voltage sensor. JPEG0007883334000024.jpg543 and DC side upper capacitor voltage JPEG0007883334000025.jpg510 and lower capacitor voltage Sample JPEG0007883334000026.jpg511, where k corresponds to the k-th sampling time or sampling point.

[0028] The three-phase output current and the three-phase power system voltage are each converted to obtain the current components in a two-phase stationary coordinate system (αβ coordinate system). JPEG0007883334000027.jpg520 and power system voltage component The image JPEG0007883334000028.jpg625 is obtained, and this transformation result is used as the input quantity for the subsequent ultralocal model establishment and full-order observer.

[0029] S2. Construct a hyperlocal model of the inverter, unify and equip the uncertainty of the system parameters with a concentrated disturbance, estimate the concentrated disturbance in real time using a full-order observer, and obtain disturbance estimates, specifically including the following:

[0030] In S21.3 level SNPC grid-connected inverters, parameters such as filtering inductance, resistance, and power system impedance exhibit certain discrepancies in actual operation. Taking a two-phase stationary coordinate system as an example, the nominal model may be expressed as follows: JPEG0007883334000029.jpg1061 Here, JPEG0007883334000030.jpg56 is the grid connection current vector, JPEG0007883334000031.jpg58 is the inverter output voltage vector. JPEG0007883334000032.jpg57 is the power system side voltage vector, JPEG0007883334000033.jpg510 represents the equivalent voltage error term due to the dead zone, voltage sampling deviation, etc., where L and R are the nominal filtering inductance and resistance, respectively.

[0031] Considering the parameter mismatch and the effects of the device's conduction voltage drop, the actual model may also be written as follows: JPEG0007883334000034.jpg10128 Here, ΔL and ΔR are parameter deviations. JPEG0007883334000035.jpg57 is the conduction resistance of a power device. JPEG0007883334000036.jpg512 shows the equivalent voltage error due to the dead zone and sampling error.

[0032] To reduce the dependence on nominal parameters, we introduce a first-order hyperlocal model gain α>0 and reconstruct the above equation into the following hyperlocal model: JPEG0007883334000037.jpg1035 Here, JPEG0007883334000038.jpg727 is the output voltage of a 3-level SNPC inverter in a two-phase stationary coordinate system. JPEG0007883334000039.jpg725 represents the output current in a two-phase stationary coordinate system, where T is the matrix transpose, and α is a design parameter related to the filtering inductance, used to reduce the control algorithm's dependence on actual filtering inductance, resistance, and other parameters, with α selected to an approximate level of 1 / L. The amplitude of JPEG0007883334000040.jpg56 is reduced, making it easier to converge observations, and in engineering terms, it may be determined according to the nominal value of the filtering inductance, allowing for settlement within a certain range. JPEG0007883334000041.jpg727 is a concentrated disturbance term, used to absorb parameter discrepancies and external disturbances. JPEG0007883334000042.jpg22128 The above modeling method allows all parameter errors and external disturbances of filtering inductance, resistance, and power system side impedance to be concentrated as disturbance terms. This is absorbed into JPEG0007883334000043.jpg512, laying the foundation for subsequent ultralocal model predictive control.

[0033] S22. Based on the above ultralocal model, design a full-order sliding mode observer (SMO) for concentrated disturbances Estimate JPEG0007883334000044.jpg56 in real time, First, construct the extended state vector, and the extended state vector JPEG0007883334000045.jpg1120 and its observed values Let's call it JPEG0007883334000046.jpg1219, and here, JPEG0007883334000047.jpg67 and JPEG0007883334000048.jpg66 shows the current observation value and the concentrated disturbance observation value, respectively. The observation errors are, The filename is JPEG0007883334000049.jpg665. The continuous-time full-order sliding mode observer equation is as follows: JPEG0007883334000050.jpg3166 Here, k1 and k2 are the sliding mode observation gains, ω is the system connection angular frequency, and j is the imaginary unit for representing the orthogonal rotation operator on the αβ plane.

[0034] Substituting these values, we obtain the error dynamics. By appropriately selecting k1 and k2 in JPEG0007883334000051.jpg2257, the observation error converges to zero within a finite time. JPEG0007883334000052.jpg56 and We can guarantee that accurate estimation of JPEG0007883334000053.jpg56 can be achieved by first selecting k2 / k1 as the observer equivalent bandwidth parameter (trade-off between dynamics and noise) during the setting process, and then selecting k1 to satisfy the upper limit of disturbance and the noise immunity requirement. The filename is JPEG0007883334000054.jpg535.

[0035] The above observer is sampled with a sampling period T s Discretize according to the formula, and obtain discrete equations using the forward Euler method. JPEG0007883334000055.jpg17128 Current current measurement value at each sampling period Only JPEG0007883334000056.jpg626 is used, voltage By controlling the observed values ​​of JPEG0007883334000057.jpg628 and the previous period, the current observed value JPEG0007883334000058.jpg537 and disturbance estimates By updating JPEG0007883334000059.jpg636, real-time estimation of concentrated disturbances can be achieved, and the observer structure is as shown in Figure 3.

[0036] S3. Following the current command, the reference current at future times is predicted using extrapolation, and the current sampling current and disturbance estimates are linked to calculate the reference voltage vector for the next sampling period based on a hyperlocal model, specifically including the following:

[0037] S31. Based on the grid-connected operation target or the output of the higher-level power controller, the current command in the two-phase stationary coordinate system given to the external loop is performed. The image is named JPEG0007883334000060.jpg611, and in order to improve the accuracy of model-free prediction, a reference current for two future sampling periods is obtained using second-order Lagrangian extrapolation. JPEG0007883334000061.jpg13128 Here, JPEG0007883334000062.jpg641 is a current command for the first two sampling periods.

[0038] S32. Based on the ultralocal model, concentrated disturbances are observed Under the condition of replacing with JPEG0007883334000063.jpg614, the current prediction equation is obtained. JPEG0007883334000064.jpg781 To achieve steady-state error-free tracking of current, predict current at time k+2. JPEG0007883334000065.jpg519 reference current within the next 2 beats When tracking JPEG0007883334000066.jpg618, the estimated reference voltage vector is as follows: JPEG0007883334000067.jpg1183 The disturbance estimation term has already absorbed parameter errors such as filtering inductance, resistance, and power system impedance, and thus obtained JPEG0007883334000068.jpg612 no longer explicitly depends on nominal parameters such as L and R, thereby maintaining good current tracking performance even with parameter discrepancies and changes in working conditions.

[0039] S4. Based on the position of the reference voltage vector in the spatial vector diagram, a set of candidate voltage vectors is selected, the optimal dual vector combination is evaluated and selected using a cost function, its duration is determined, redundant small vectors are selected according to the DC-side midpoint voltage deviation, and a PWM drive signal is generated to realize modulation control, specifically including the following:

[0040] S41. Reference voltage vector After determining JPEG0007883334000069.jpg612, based on its position in the 3-level SNPC spatial voltage vector diagram shown in Figure 2, i.e., the sector (sectors 1-6) and sub-triangular region (each sector contains 4 sub-regions), the large vector, small vector, and zero vector within the corresponding region are selected to form a candidate voltage vector set {u i Construct}.

[0041] Any single candidate vector u i For this, the cost function is the Euclidean distance between the reference voltage and this vector. Defined as JPEG0007883334000070.jpg637, Two vectors u in the candidate set p u qIf we construct multiple line segments consisting of and assume that their action time is inversely proportional to the cost function, then the action time distribution of the two vectors is JPEG0007883334000071.jpg741 is obtained, and here, T s This is the sampling period, If it is necessary to superimpose zero vectors, the zero vector action time It can also be named JPEG0007883334000072.jpg530.

[0042] For each pair of candidate dual vectors, calculate their combined voltage. JPEG0007883334000073.jpg1152JPEG0007883334000074.jpg630 is the smallest criterion for optimal dual vector JPEG0007883334000075.jpg613 and its duration Select JPEG0007883334000076.jpg518.

[0043] S42. To achieve midpoint potential balance, obtained by sampling Using JPEG0007883334000077.jpg526 to determine the midpoint voltage deviation Calculate JPEG0007883334000078.jpg531.

[0044] When redundancy exists in small vectors, Select redundant small vectors based on different codes of JPEG0007883334000079.jpg511. I participated in the synthesis of JPEG0007883334000080.jpg511. In the case of JPEG0007883334000081.jpg518, a small vector that is advantageous for improving the upper capacitor voltage is preferentially selected. In the case of JPEG0007883334000082.jpg518, automatic balancing of the midpoint potential is achieved by preferentially selecting small vectors that are advantageous for improving the lower capacitor voltage.

[0045] The finally selected dual vector and its action time are mapped to the specific switching states of the three-level SNPC common module and the three-phase bridge arm, and the gate drive signals of each power device are generated by PWM or timing comparison method, completing the predictive control and neutral point potential balance of the three-level SNPC inverter.

[0046] To sum up, this embodiment combines the hyperlocal model, full-order observer and dual-vector hyperlocal model predictive control. When there are filter parameter deviations and uncertainties in the power system side impedance, it can significantly reduce the harmonics and ripples of the system connection current, and improve the robustness against system parameter mismatches and external disturbances.

[0047] To further verify the technical effects of the method described in the present invention, a simulation software is used to construct a three-level SNPC inverter closed-loop control system simulation platform, and the conventional FCS-MPC dual-vector predictive current control method in the prior art is selected as the comparison object, and a comparative experiment is carried out under the same conditions as the dual-vector model-free predictive current control method based on the full-order sliding mode observer (SMO) proposed in the present invention. The two methods both use the same sampling period T s , switching frequency f s , DC bus voltage U dc , current reference i * , filtering structure and its nominal parameters. Except for the "parameter mismatch" setting required for comparison, other simulation parameters are the same, and the simulation parameters are as shown in Table 1.

[0048] [Table 1] Simulation parameters JPEG0007883334000083.jpg65145

[0049] The present invention sets out the following two typical working conditions.

[0050] (1) Parameter adaptation work status: The controller parameters are matched to the parameters being controlled and are used to compare the steady-state electrical energy quality of the two methods under normal conditions. (2) Inductance parameter mismatch working conditions: The inductance parameter used in the controller is 1.5 times the actual inductance and is used to compare the robustness of the two methods under parameter mismatch conditions.

[0051] Under each operating condition, once the system is running in a steady state, three-phase grid interconnection current data for two consecutive fundamental wave periods is extracted and FFT harmonic analysis is performed, and THD is calculated based on a unified FFT setting. A comparison of the key indicators of the two methods under different operating conditions is summarized in Table 2.

[0052] [Table 2] Comparison of key indicators JPEG0007883334000084.jpg51150

[0053] Here, Figures 5(a) and (b) show the simulation results of the conventional dual-vector predictive current control method and the model-free predictive current control method based on the SMO of the present invention, respectively, under parameter adaptation conditions. As can be seen from Figure 5, both methods were able to achieve stable tracking, but the present invention showed smaller current ripple, lower harmonic content, and a further reduction in THD from approximately 1.50% to approximately 1.24%.

[0054] Furthermore, Figures 6(a) and (b) show the comparison results of the two methods under the condition of inductance parameter mismatch (controller inductance parameter is 1.5 times the actual inductance), respectively. As shown in Figure 6, the conventional method's prediction calculation explicitly depends on the inductance parameter, so the mismatch leads to prediction deviation, resulting in increased current distortion and a THD increase of approximately 2.06%. However, the present invention uses a hyperlocal model to equate uncertainties such as parameter deviations to concentrated disturbances and compensates for them with SMO online estimation, thus maintaining a relatively good current waveform even under the same mismatch condition, with a THD of approximately 1.26%, demonstrating that the present invention has stronger parameter robustness. In addition, as shown in Figure 7, when the load of the 3-level SNPC inverter changes rapidly, the output current of the present invention shows a rapid response, a smooth transition process, and good dynamic performance. As shown in Figure 8, the fluctuations of the DC-side upper and lower capacitor voltages were controlled within ±0.5V during the operation period of the 3-level SNPC inverter, indicating a relatively good midpoint potential balance control effect.

[0055] Based on the results described above, the model-free, high-robustness predictive control method for a 3-level SNPC inverter according to the present invention effectively reduces harmonics in the output current of the SNPC inverter, improves the robustness of the system in the event of parameter mismatches and changes in working conditions, and achieves both rapid dynamic response and good DC-side midpoint voltage balance control effects.

[0056] Example 2 As shown in Figure 4, a model-free, high-robustness predictive control system for a 3-level SNPC inverter is used to realize a model-free, high-robustness predictive control method for a 3-level SNPC inverter. Real-time sampling is performed on a 3-level SNPC inverter system to obtain the three-phase output current, three-phase power system voltage, and DC upper and lower capacitor voltages. After performing a Clark transform on the three-phase output current and three-phase power system voltage, the current in the two-phase stationary coordinate system is obtained. JPEG0007883334000085.jpg512 and power system voltage A sampling module for forming JPEG0007883334000086.jpg515, Current in a two-phase stationary coordinate system JPEG0007883334000087.jpg613, Power system voltage JPEG0007883334000088.jpg513 and inverter output voltage A fully-order sliding mode observer, constructed with JPEG0007883334000089.jpg411 as input, estimates the extended state in real time within a hyperlocal model framework and provides concentrated disturbance estimates. JPEG0007883334000090.jpg613 and current observation values An observation module for outputting JPEG0007883334000091.jpg714, The full-order sliding mode observer used in the observation module corrects concentrated disturbances using current observation errors, thereby improving the observation accuracy of parameter discrepancies and external disturbances. Based on the ultralocal model, current instruction By combining JPEG0007883334000092.jpg712 with other data, the reference current at future times is extrapolated and predicted, and the current sampling current and concentrated disturbance estimates are used to create the reference voltage vector for the next sampling period. A reference voltage calculation module for calculating JPEG0007883334000093.jpg816, wherein the reference voltage vector is used to compensate for parameter mismatches and external disturbances, and the calculation process is a reference voltage calculation module that does not explicitly depend on the nominal parameters of filtering inductance, resistance, and power system side impedance. A modulation module for selecting a set of candidate voltage vectors based on the position of a reference voltage vector, constructing candidate line segments, selecting the optimal acting voltage vector based on a cost function, and determining its acting time, The system includes a control module for selecting redundant small vectors according to the midpoint voltage deviation to achieve midpoint potential balance control, which converts the selected operating voltage vector and its operating time into PWM drive signals for the 3-level SNPC inverter common module and the three-phase bridge arm, and controls the operation of the 3-level SNPC inverter system.

[0057] Example 3 A system based on the aforementioned control method, The system includes a main circuit and a control subsystem for a 3-level SNPC inverter. The main circuit of the inverter includes a DC capacitor, a DC power supply, a level generation unit, and a three-phase reverse-shifting bridge arm and filter connected to a power grid or load. The control subsystem uses the aforementioned ultra-local predictive control method to drive the level generation unit and generate the corresponding switching control signals to control the main circuit of the inverter.

[0058] Example 4 A computer-readable storage medium storing a computer program, wherein when the computer program is invoked by a processor, it executes a model-free, high-robustness predictive control method and system for a 3-level SNPC inverter, the processor being a digital signal processor, a microcontroller, or a field-programmable gate array, used to achieve real-time predictive control of the 3-level SNPC inverter.

[0059] As can be seen from the above description, the above embodiment of the present invention achieves the following technical effects.

[0060] This invention, by introducing ultralocal models and full-order sliding mode observers (SMOs), centralizes uncertainties in inverter systems (e.g., filtering inductance, resistance, power system impedance, and switching device nonlinearity) and reduces disturbances. The system effectively models the data as JPEG0007883334000094.jpg613 and estimates them in real time. Compared to conventional control methods based on accurate parameter models, the present invention can significantly improve the robustness and adaptability of the system by still ensuring relatively low current tracking errors and relatively small current total harmonic distortion (THD) even with parameter mismatches, element degradation, and external disturbances.

[0061] The fully-order sliding mode observer in this invention reduces reliance on long-term historical data through a simplified observer structure. Compared to conventional methods, the observer performs disturbance estimation relying only on data from the current and previous sampling periods, significantly reducing memory requirements and decreasing the computational complexity of the system. It is particularly applicable to application scenarios requiring real-time control and high dynamic performance, such as high-power inverters and power grid interconnection systems.

[0062] The examples and / or embodiments described above are used solely to illustrate preferred examples and / or embodiments of realizing the art of the present invention and do not limit the embodiments of the art of the present invention in any way. Those skilled in the art can make some changes or modifications to other equivalent examples without departing from the technical means disclosed in the summary of the present invention, and these should still be considered substantially the same art or examples as the present invention.

[0063] This specification describes the principles and embodiments of the present application using specific examples, but the above description of embodiments is used solely to assist in understanding the methods and core ideas of the present application. As stated above, these are merely preferred embodiments of the present application, and it should be noted that, due to the limitations of literal expression, there are objectively infinite specific structures, and a person skilled in the art can make multiple improvements, modifications, or changes without departing from the principles of the present application, and can combine the above technical features in an appropriate manner, and any of these improvements, modifications, changes, or combinations, or the direct application of the concept and technical ideas of the invention to other cases without improvements, should all be considered within the scope of protection of the present application.

Claims

1. A model-free, high-robustness predictive control method for a 3-level SNPC inverter, The three-phase output current of the inverter, the power system voltage, and the DC side capacitor voltage are sampled and converted to obtain the current and voltage in a two-phase stationary coordinate system. The process involves constructing a hyperlocal model of the inverter, unifying the uncertainty of the system parameters into a concentrated disturbance for equivalence, estimating the concentrated disturbance in real time using a full-order observer, and obtaining a disturbance estimate. Following the current command, the reference current at a future time is predicted using extrapolation, and the current sampling current and the disturbance estimate are linked to calculate the reference voltage vector for the next sampling period based on the ultralocal model. This includes selecting a set of candidate voltage vectors based on the position of the reference voltage vector in the spatial vector diagram, evaluating and selecting the optimal combination of dual vectors using a cost function, determining its duration, selecting redundant small vectors according to the DC-side midpoint voltage deviation, and generating a PWM drive signal to realize modulation control. Predicting a reference current at a future time using extrapolation, according to a current command, specifically means: Current commands according to power grid connection commands or the output of the current external loop regulator. To obtain, Using second-order Lagrangian extrapolation, the reference current for the next two sampling periods is determined based on the current command for the current and the first two sampling periods. This includes predicting, Specifically, by linking the current sampling current with the disturbance estimate and calculating the reference voltage vector for the next sampling period based on the ultralocal model, Current sampling current and disturbance estimates Connecting and predicting current To obtain, In the aforementioned ultralocal model, predict the currents for two future sampling periods. The reference current within the next two beats Track the reference voltage vector for the next sampling period. This involves calculating the reference voltage vector A model-free, highly robust predictive control method for a three-level SNPC inverter, characterized in that it does not explicitly depend on nominal parameters of filtering inductance, resistance, and power system side impedance.

2. Specifically, sampling and converting the three-phase output current, power system voltage, and DC capacitor voltage of the inverter to obtain the current and voltage in a two-phase stationary coordinate system means: Within each sampling period, real-time sampling is performed on the 3-level SNPC inverter, the three-phase output current is acquired via a current sensor, and the three-phase power system voltage and the DC side upper and lower capacitor voltages are acquired via a voltage sensor. The model-free, high-robustness predictive control method for a three-level SNPC inverter according to claim 1, characterized by comprising performing a Clark transform on the three-phase output current and the three-phase power system voltage, respectively, to obtain the current component and the power system voltage component in a two-phase stationary coordinate system.

3. Constructing a hyperlocal model of the inverter and unifying the uncertainty of the system parameters into a concentrated disturbance for equivalence means, specifically, In the aforementioned two-phase stationary coordinate system, the effects of the real-time sampling parameter deviation and non-ideal factors are concentrated as disturbances. By unifying and equating them, the ultralocal model This includes constructing, Here, This is the output voltage of a 3-level SNPC inverter in a two-phase stationary coordinate system. The model-free, high-robustness predictive control method for a three-level SNPC inverter according to claim 2, characterized in that is the output current in a two-phase stationary coordinate system, T is the transpose of the matrix, and α is a design parameter related to the filtering inductance.

4. Specifically, using a full-order observer to estimate the aforementioned concentrated disturbance in real time and obtain disturbance estimates means, The method involves constructing a full-order observer based on a hyperlocal model, wherein the full-order observer uses a sliding-mode observer structure, and the input to the full-order observer is the sampling current. and output voltage Therefore, the state variables of the full-order observer include current observations and concentrated disturbances, By using a full-order observer to estimate concentrated disturbances in real time, and by discretizing them, iterative updates are completed within each sampling period using the current sample value and the observed values ​​from the previous period, thereby generating disturbance estimates. A model-free, high-robustness predictive control method for a three-level SNPC inverter according to claim 3, characterized by comprising obtaining the above.

5. Selecting a set of candidate voltage vectors based on the position of the aforementioned reference voltage vector in the spatial vector diagram, evaluating and selecting the optimal dual-vector combination using a cost function, and determining its duration, specifically involves: Based on the position of the reference voltage vector in the 3-level SNPC inverter spatial voltage vector diagram, the sector and sub-triangular region where the reference voltage vector is located are determined, the large vector, small vector and zero vector associated with the sub-triangular region are formed into a candidate voltage vector set, and multiple candidate line segments consisting of two voltage vectors are constructed with the reference voltage vector as the target. A pre-set cost function is selected, and the cost value from the reference voltage vector to the midpoint of each candidate line segment is calculated using the error between the reference voltage vector and the equivalent output voltage vector as an indicator. The line segment with the smallest cost function value is selected, and the voltage vectors at both ends of that segment are used as the two working voltage vectors for this sampling period. Based on the geometric relationship of the spatial vector diagram or the optimization conditions that approximate the equivalent output voltage vector to the reference voltage vector, the action time T of the two action voltage vectors １ , T ２ A model-free, high-robustness predictive control method for a three-level SNPC inverter according to claim 4, characterized in that it includes determining the following.

6. Specifically, selecting redundant small vectors according to the DC-side midpoint voltage deviation and generating a PWM drive signal to achieve modulation control means: Based on the DC-side upper capacitor voltage VP and lower capacitor voltage VN, the midpoint voltage deviation is calculated. If redundant small vectors exist, assuming the current control objective is met, different redundant small vectors are selected according to the sign of the midpoint voltage deviation and participate in dual-vector synthesis. The model-free high-robustness predictive control method for a three-level SNPC inverter according to claim 5, comprising generating PWM drive signals for a three-level SNPC inverter common module and a three-phase bridge arm according to two selected operating voltage vectors and their operating times, and completing model-free predictive control for one sampling period.

7. A model-free, high-robustness predictive control system for a 3-level SNPC inverter, used to realize the model-free, high-robustness predictive control method for a 3-level SNPC inverter described in any one of claims 1 to 6. A sampling module for acquiring three-phase output current, three-phase power system voltage, and DC upper and lower capacitor voltages in real time, completing the Clark transformation, and outputting the current and voltage components in a two-phase stationary coordinate system. An observation module receives the current component, voltage component, and inverter output voltage, and a full-order sliding mode observer built on a hyperlocal model estimates the extended state in real time and outputs current observation values ​​and concentrated disturbance estimate values. A reference voltage calculation module receives current commands, current observations, and concentrated disturbance estimates, uses an extrapolation algorithm to predict future reference currents, and calculates a reference voltage vector. A modulation module for selecting a set of candidate voltage vectors based on the position of a reference voltage vector, constructing candidate line segments, selecting the optimal acting voltage vector based on a cost function, and determining its acting time, A model-free, high-robustness predictive control system for a three-level SNPC inverter, characterized by including a control module for converting selected operating voltage vectors and their operating times into PWM drive signals for a three-level SNPC inverter common module and a three-phase bridge arm, and selecting redundant small vectors according to the midpoint voltage deviation to achieve midpoint potential balance control.

8. A computer-readable storage medium in which a computer program is stored, wherein when the computer program is invoked by a processor, a model-free, high-robustness predictive control method for a three-level SNPC inverter according to any one of claims 1 to 6 is executed, wherein the processor is a digital signal processor, a microcontroller, or a field-programmable gate array, and is used to realize real-time predictive control of a three-level SNPC inverter.

Images