A Data-Driven Predictive Control Method and System for a Three-Level Inverter

Abstract

The present invention proposes a data-driven predictive control method and system for a three-level inverter, including: establishing a compact dynamic linearization model considering the unknown uncertainties of the system, where the model includes the residual unknown uncertainty terms of the system; calculating the reference voltage vector for the next switching period and determining the sector in which it is located; removing the redundant small vectors with higher common-mode amplitudes in each sector of the three-level inverter space vector; comparing the magnitudes of the upper and lower DC-side capacitor voltages and the sector where the reference vector is located, and adding the virtual small vectors constructed in each sector to the candidate vectors; or, adding the small vectors in the sector where it is located to the candidate vectors; selecting three optimal vectors from the candidate vectors as the next switching action vectors of the three-level inverter, and applying the selected three optimal vectors and their corresponding action times to the control of the switching tubes in the next period of the three-level inverter.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-level inverters in power electronics, and particularly relates to a data-driven predictive control method and system for a three-level inverter. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] In recent years, new energy power generation technologies such as photovoltaic and wind power have developed rapidly. As an interface between new energy and the power grid and load, power converters are crucial for ensuring the safe and efficient operation of new energy power generation systems. Among different topological structures, three-level inverters have the advantages of low harmonic distortion, low voltage stress, and high efficiency, and are widely used in medium and low voltage systems. At the same time, finite set model predictive control has become one of the main control methods for three-level inverters due to its advantages such as simple implementation, fast dynamic response, and multi-objective optimization function, so as to achieve various performance requirements such as high power quality, low common-mode voltage, and neutral point voltage balance. However, the performance of model predictive control depends severely on the accuracy of system parameters. In practical applications, due to measurement errors, random load switching, external disturbances, etc., system parameters are extremely likely to change, and it is very difficult to establish an accurate prediction model. In contrast, data-driven predictive control only uses input / output data to establish a prediction model, getting rid of the problem of the inherent dependence of the prediction model on system parameters, so the system shows strong robustness and adaptability when parameters change.

[0004] The inventors found in their research that when using traditional data-driven predictive control to establish a current prediction model for a three-level inverter, the following problems exist:

[0005] 1. System uncertainty factors such as unknown dynamics and external disturbances are not considered, resulting in poor prediction accuracy and reduced robustness when system parameters change.

[0006] 2. Secondly, since only one vector is used in each sampling period, the volt-second balance cannot be satisfied, the control error is large, and large current harmonics are generated.

[0007] 3. In addition, for the performance requirements of multiple control objectives of three-level inverters, traditional methods do not consider the problem of common-mode voltage suppression, and still need to predict the capacitor voltage with the DC capacitor parameter to achieve neutral point balance control. Summary of the Invention

[0008] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a data-driven predictive control method for a three-level inverter, which only uses the input / output data of the three-level inverter to simultaneously improve the current quality of the three-level inverter, suppress the common-mode voltage, enhance the neutral point balance ability, and improve the robustness of the system when parameters are perturbed.

[0009] To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

[0010] In a first aspect, a data-driven predictive control method for a three-level inverter is disclosed, including:

[0011] Establish a compact-form dynamic linearization prediction model considering the unknown uncertainty of the system, and the model includes a system residual unknown uncertainty term;

[0012] Regard the system unknown uncertainty term as the system extended state, design a generalized extended observer to estimate this state, compensate the calculation result into the established prediction model, calculate the reference voltage vector in the next switching period, and judge the sector where it is located;

[0013] Eliminate the redundant small vectors with higher common-mode amplitude in each sector of the three-level inverter space vector, and construct virtual vectors with enhanced midpoint balancing ability using the large vectors and the two small vectors adjacent to the large vectors in each sector. The constructed virtual vectors are used to replace the eliminated small vectors;

[0014] Compare the magnitudes of the upper and lower capacitor voltages on the DC side at present and the sector where the reference vector is located, and add the constructed virtual small vectors in each sector to the candidate vectors; or, add the small vectors in the sector where it is located to the candidate vectors;

[0015] Select the optimal vector from the candidate vectors as the next switching action vector of the three-level inverter, and apply the selected optimal vector and its corresponding action time to the switching tube control of the three-level inverter in the next cycle.

[0016] As a further technical solution, compare the magnitudes of the upper and lower capacitor voltages V p , V n on the DC side at present. If V p > V n and the reference vector is located in sectors I, III, V, or V p < V n and the reference vector is located in sectors II, IV, VI, add the constructed virtual small vectors in each sector to the candidate vectors; otherwise, add the small vectors in the sector where it is located to the candidate vectors.

[0017] As a further technical solution, select three vectors from the candidate vectors as the next switching action vectors of the three-level inverter:

[0018] First, regard the small vectors or virtual small vectors in the candidate vectors as the first vector to improve the midpoint balancing ability;

[0019] Secondly, a simplified cost function without weight factors is constructed, and the cost function values of the remaining candidate vectors are calculated respectively. Two vectors with the minimum cost function are selected from them as the second and third vectors, and the action time of each vector is calculated according to the volt-second balance principle;

[0020] Finally, the three selected optimal vectors and their corresponding action times are applied to the switching tube control of the three-level inverter in the next cycle.

[0021] In a second aspect, a data-driven predictive control system for a three-level inverter is disclosed, including:

[0022] A prediction model establishment module, configured to: establish a compact-form dynamic linearization prediction model considering the unknown uncertainty of the system, and the model includes the residual unknown uncertainty term of the system;

[0023] A reference voltage vector calculation module, configured to: regard the residual unknown uncertainty term of the system as the system extended state, and transform the established compact-form dynamic linearization model into a state space form; design a generalized extended observer based on the state space form to observe the system extended state; based on the observation of the system extended state by the generalized extended observer, use the calculation result of the system unknown uncertainty term for compensation, calculate the reference voltage vector in the next switching cycle and judge the sector where it is located;

[0024] A candidate vector selection module, configured to: eliminate the redundant small vectors with higher common-mode amplitudes in each sector of the space vector of the three-level inverter, and construct virtual vectors with improved midpoint balance ability by using the large vectors and the two small vectors adjacent to the large vectors in each sector, and the constructed virtual vectors are used to replace the eliminated small vectors;

[0025] Compare the magnitudes of the upper and lower DC-side capacitor voltages and the sector where the reference vector is located, and add the constructed virtual small vectors in each sector to the candidate vectors; or, add the small vectors in the sector where it is located to the candidate vectors;

[0026] An action vector calculation module, configured to: select the optimal vector from the candidate vectors as the next switching action vector of the three-level inverter, and apply the selected optimal vector and its corresponding action time to the switching tube control of the three-level inverter in the next cycle.

[0027] The above one or more technical solutions have the following beneficial effects:

[0028] 1. The technical solution of the present invention only uses historical input and output data to establish a three-level inverter current prediction model considering the unknown uncertainty of the system, avoids introducing any system parameters, and completely eliminates the dependence on the system parameters of the controller;

[0029] 2. The technical solution of the present invention designs a generalized extended observer to identify and compensate the influence of system uncertainties into the prediction model, significantly improving the robustness when the system parameters change;

[0030] 3. The technical solution of the present invention constructs a virtual small vector by a large vector and two adjacent small vectors to replace the redundant small vector with a relatively high common-mode amplitude, effectively suppressing the common-mode voltage and improving the neutral-point voltage balancing ability;

[0031] 4. The technical solution of the present invention selects three candidate vectors to synthesize the reference voltage vector, satisfying the volt-second balance principle, greatly reducing the control error, eliminating the cumbersome design of the weighting coefficients, and simultaneously realizing functions such as improving the current quality, suppressing the common-mode voltage, and enhancing the neutral-point balancing ability. The implementation is simple and highly reliable;

[0032] 5. The technical solution of the present invention realizes the improvement of the current quality, the suppression of the common-mode voltage, and the enhancement of the neutral-point balancing ability based on multi-vector data-driven predictive control in the discrete domain, which is applicable to digital controllers, and this is of great significance to the application of three-level inverters.

[0033] Advantages of additional aspects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The specification drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0035] Figure 1 It is a schematic control block diagram of an embodiment of the present invention;

[0036] Figure 2 It is a space vector diagram of SVPWM of a three-level inverter implemented by the present invention;

[0037] Figure 3 It is a schematic diagram of constructing a virtual small vector in sectors I and II implemented by the present invention;

[0038] Figure 4 It is the waveforms of the neutral-point voltage, current, and common-mode voltage using the traditional data-driven predictive control and the present invention;

[0039] Figure 5 It is the current waveforms using the traditional data-driven predictive control and the present invention when the system reference current and load parameters change;

[0040] Figure 6 It is the neutral-point unbalance control waveform using the present invention and removing the constructed virtual small vector from the candidate vectors. DETAILED DESCRIPTION OF THE INVENTION

[0041] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0042] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0043] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0044] Embodiment 1

[0045] As Figure 1 shown, this embodiment discloses a data-driven predictive control method for a three-level inverter. In one control cycle, it includes:

[0046] Step 1: Set the output reference curve, establish a compact-form dynamic linearization prediction model considering the unknown uncertainties of the system, estimate the value of the pseudo-partial derivative matrix, and construct a calculation reference input expression. The established prediction model does not depend on any system parameters.

[0047] The specific process is as follows: Through coordinate transformation, the historical output three-phase current data of the three-level inverter is transformed into the two-phase stationary coordinate system (αβ), and a compact-form dynamic linearization prediction model considering the unknown uncertainties of the system is established as shown in the following formula:

[0048] y(k + 1) = y(k) + φ(k)Δu(k) + ξ(k);

[0049] In this step, y(k + 1) is the predicted current output at the next moment, Δu(k) is the input change amount between the current moment and the previous moment, ξ(k) is the residual unknown uncertainty term of the system, and φ(k) is the pseudo-partial derivative matrix, which is a time-varying parameter for establishing the dynamic linearization model of the system output and input. Since its exact value is difficult to obtain, it can be calculated by an estimation algorithm as:

[0050]

[0051] where is the estimated value of the pseudo-partial derivative matrix φ(k); μ > 0 is used to limit the change of the pseudo-partial derivative parameter; the addition of the step size factor η ∈ (0, 2] is to make the controller design more flexible. In order to make the pseudo-partial derivative estimation algorithm have stronger tracking ability for time-varying parameters, an algorithm reset mechanism is introduced:

[0052]

[0053] where are the set initial values respectively, where \(i = 1,\cdots,m\), \(j = 1,\cdots,n\); \(\varepsilon\) is a sufficiently small positive number. According to the setting, output the reference curve \(y^*(k + 1)\), and calculate the reference input voltage vector \(u^*(k)\):

[0054] According to the setting, output the reference curve \(y^*(k + 1)\), and calculate the reference input voltage vector \(u^*(k)\):

[0055]

[0056] where \(\lambda>0\) is used to limit the change of the control input, and the step factor \(\rho>0\);

[0057] Step 2: Regard the unknown uncertainty term of the system as the system extended state, design a generalized extended observer to estimate this state, compensate the calculation result into the established prediction model, calculate the reference voltage vector in the next switching period and judge the sector it is in, so as to effectively suppress the influence brought by the unknown uncertainty of the system and improve the robustness of the controller.

[0058] The specific process is as follows: Regard \(\xi(k)\) as the system extended state \(x_2(k)\), define \(\omega(k)=\xi(k + 1)-\xi(k)\), and transform the established compact-form dynamic linearization model into the state-space form, and its expression is:

[0059]

[0060] where \(I\) 2×2 represents the \(2\times2\) identity matrix.

[0061] Design a generalized extended observer to solve the observation result of the system extended state Its expression is:

[0062]

[0063] where are the estimated values of \(x(k)\) and \(y(k)\) respectively, \(L\) is the gain matrix of the generalized extended observer.

[0064] Compensate the calculation result of the system unknown uncertainty term into the prediction model established in Step 1, calculate the reference voltage vector in the next switching period and judge the sector it is in, and its expression is:

[0065]

[0066] In this step, calculate the unknown uncertainty term of the system by designing a generalized extended observer, and obtain the reference input voltage vector and the sector it is in.

[0067] Step 3: Eliminate the redundant small vectors with higher common-mode amplitude from the space vectors, and construct virtual small vectors with improved midpoint balancing ability using the large vectors and the two small vectors adjacent to the large vectors in each sector, and replace the eliminated high common-mode small vectors with them, so as to effectively suppress the common-mode voltage amplitude within ±V DC / 6 (V DC is the DC bus voltage), and at the same time has the ability to actively regulate the midpoint of the DC side in each sector.

[0068] As Figure 2 shown, eliminate the 6 redundant small vectors V N1 , V P2 , V N3 , V P4 , V N5 , V P6 with higher common-mode amplitude in each sector of the three-level SVPWM space vectors, and construct virtual small vectors V NL1 , V PL2 , V NL3 , V PL4 , V NL5 , V PL6 with improved midpoint balancing ability using the large vectors and the two small vectors adjacent to the large vectors in each sector. The expressions are as follows:

[0069]

[0070] Among them, V Lx represents the large vector in each sector, respectively represent the two small vectors adjacent to the corresponding large vector V Lx .

[0071] The constructed virtual vectors are used to replace the eliminated small vectors. Specifically, taking sectors I and II as examples, as Figure 3 shown in (a) and (b) of it, in sector I, construct the virtual vector V L1 using the large vector V N2 , the adjacent small vector V N6 ; in sector II, construct the virtual vector V NL1 using the large vector V L2 , the adjacent small vector V P1 , V P3 . NL1

[0072] In this step, considering the common-mode voltage suppression, eliminate the candidate vectors with high common-mode voltage and construct virtual small vectors to improve the midpoint balancing ability.

[0073] Step 4: Compare the magnitudes of the current upper and lower capacitor voltages V p , V n of the DC side. If Vp > V n and the reference vector is located in sectors I, III, V (such as Figure 2 ), or V p < V n and the reference vector is located in sectors II, IV, VI, add the virtual small vectors constructed in each sector to the candidate vectors; otherwise, add the small vector in the sector where it is located to the candidate vectors.

[0074] Specifically, taking the reference vector located in sector I as an example, when V p > V n the candidate vectors are the four common vectors V0, V L1 , V M6 , V M1 and the constructed virtual small vector V NL1 , otherwise the candidate vectors are V0, V L1 , V M6 , V M1 and the small vector V P1 .

[0075] In this step, by comparing the magnitudes of the upper and lower capacitor voltages on the DC side and according to the sector where the current reference input voltage vector is located, determine the candidate active vector for the next switching period, thereby realizing the neutral point balance control.

[0076] Step 5: Construct a simplified cost function without a weighting factor to reduce the complexity of the cost function, and use the small vector or virtual small vector as the first vector. Select two vectors with the minimum cost function among the remaining four common candidate vectors, calculate the action time of each vector according to the volt-second balance principle, and apply the selected three vectors and the corresponding times to the inverter switching tubes for control to achieve output current tracking.

[0077] The specific steps are as follows: Select three vectors from the candidate vectors as the next switching action vectors of the three-level inverter. First, use the small vector or virtual small vector in the candidate vectors as the first vector to improve the neutral point balance ability. Second, construct a simplified cost function without a weighting factor, and its expression is:

[0078]

[0079] where respectively represent the components on the α, β coordinate axes.

[0080] Calculate the cost function values of the remaining candidate vectors respectively, and select two vectors with the minimum cost function as the second and third vectors, and its expression is:

[0081]

[0082] where g2，x (x = 1, …, 4) represents the calculated values of the cost functions of four common candidate vectors.

[0083] According to the volt-second balance principle, the action times T1(k), T2(k), and T3(k) of each vector are calculated as shown in the following formula:

[0084]

[0085] Among them, T s is the control period, and v xα / v xβ (x = α, β) are the components of the selected three vectors on the α and β coordinate axes.

[0086] Finally, the selected three optimal vectors and their corresponding action times are applied to the switching tube control of the three-level inverter in the next cycle.

[0087] In this step, by constructing a simplified cost function without a weighting factor, the optimal three vectors are calculated from the candidate vectors to realize the calculation of the action time of each vector.

[0088] It should be noted that the technical solution of the present invention includes the establishment of a current prediction model considering system unknown uncertainties, system unknown uncertainty compensation, construction of virtual small vectors with improved midpoint balance ability, construction of a simplified cost function without a weighting factor, selection of the optimal three candidate vectors and calculation of their action times. The present invention adopts data-driven predictive control based on multiple vectors to simultaneously achieve output current control, common-mode voltage control, and midpoint balance control of the three-level inverter.

[0089] The present invention realizes the improvement of current quality, suppression of common-mode voltage, and enhancement of midpoint balance ability without relying on any system parameters, and improves the robustness of the system to parameter changes.

[0090] The present invention establishes a current prediction model of a three-level inverter considering system unknown uncertainties by using historical input and output data, eliminates the dependence of the controller on system parameters, compensates for the influence of system uncertainties, and improves the robustness of the system to parameter changes. At the same time, virtual small vectors with improved midpoint balance ability are constructed with large vectors in each sector and their two adjacent small vectors, and redundant small vectors with higher common-mode amplitude are replaced. By selecting three candidate vectors to synthesize the reference voltage vector, the control error is greatly reduced, and the cumbersome design of the weighting coefficient of the cost function is omitted. At the same time, the improvement of current quality, suppression of common-mode voltage, and enhancement of midpoint balance ability are realized. The method is simple and reliable, and has great significance for the application of three-level inverters.

[0091] The present invention is not limited to the operation of a three-level inverter in grid-connected or islanding modes, is not limited to a specific three-level inverter topology, and is not limited to the form of the DC-side power supply. It is applicable to different scenarios such as low voltage, medium voltage, and high voltage, and has strong scalability and practicability.

[0092] Simulation cases:

[0093] Figure 4 (a) and (b) in [reference] are the waveforms of the midpoint voltage, current, and common-mode voltage when using traditional data-driven predictive control and the method of this patent respectively. From the simulation results, it can be seen that when using traditional data-driven predictive control, the current harmonic is large and the common-mode amplitude is one-third of the DC voltage. The method of the present invention has small current harmonics and the common-mode amplitude is one-sixth of the DC voltage, effectively improving the current quality and suppressing the common-mode voltage.

[0094] Figure 5 (a) and (b) in [reference] are the current waveforms when using traditional data-driven predictive control and the method of this patent respectively when the system reference current and load parameters change. From the simulation results, it can be seen that when using traditional data-driven predictive control, the current oscillates greatly and the response is slow when the system parameters change. The method of the present invention has a smooth current transition process and a fast response when the system parameters change, effectively improving the robustness when the system parameters change.

[0095] Figure 6 (a) and (b) in [reference] are the midpoint imbalance control waveforms when using the method of this patent and when removing the constructed virtual small vector from the candidate vectors respectively. From the simulation results, it can be seen that due to the construction of a virtual small vector with midpoint balance, the present invention can quickly achieve midpoint voltage balance when the DC-side midpoint is unbalanced, effectively improving... The constructed virtual small vector of the present invention can

[0096] From the above simulation results, it can be seen that the multi-vector data-driven predictive control method for a three-level inverter proposed by the present invention only based on the input / output data of the three-level inverter simultaneously realizes the improvement of current quality, the suppression of common-mode voltage, the enhancement of midpoint balance ability, and the improvement of the robustness performance when the system parameters change.

[0097] Embodiment 2

[0098] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the above method are realized.

[0099] Embodiment 3

[0100] The purpose of this embodiment is to provide a computer-readable storage medium.

[0101] A computer-readable storage medium stores a computer program thereon, and when the program is executed by a processor, it performs the steps of the above method.

[0102] Embodiment 4

[0103] The purpose of this embodiment is to provide a three-level inverter data-driven predictive control system, including:

[0104] A prediction model establishment module, configured to: transform the historical three-phase current data output by the three-level inverter into a two-phase stationary coordinate system, and establish a compact-form dynamic linearization prediction model considering the unknown uncertainty of the system, where the model includes a system residual unknown uncertainty term;

[0105] A reference voltage vector calculation module, configured to: regard the system residual unknown uncertainty term as a system extended state, and transform the established compact-form dynamic linearization model into a state-space form; design a generalized extended observer based on the state-space form to observe the system extended state; based on the observation of the system extended state by the generalized extended observer, use the calculation result of the system unknown uncertainty term for compensation, calculate the reference voltage vector for the next switching period and determine the sector where it is located;

[0106] A candidate vector selection module, configured to: eliminate the redundant small vectors with higher common-mode amplitudes in each sector of the three-level inverter space vector, and construct virtual vectors with improved midpoint balancing ability using the large vectors and the two small vectors adjacent to the large vectors in each sector, and the constructed virtual vectors are used to replace the eliminated small vectors;

[0107] Compare the magnitudes of the upper and lower DC-side capacitor voltages and the sector where the reference vector is located, and add the constructed virtual small vectors in each sector to the candidate vectors; or, add the small vectors in the sector where it is located to the candidate vectors;

[0108] An acting vector calculation module, configured to: select three optimal vectors from the candidate vectors as the next switching acting vectors of the three-level inverter, and apply the selected three optimal vectors and their corresponding acting times to the control of the switching tubes in the next cycle of the three-level inverter.

[0109] The steps involved in the devices in the above Embodiments 2, 3, and 4 correspond to those in Method Embodiment 1, and the specific implementation manners can be referred to the relevant description parts of Embodiment 1. The term "computer-readable storage medium" should be understood to include a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0110] Those skilled in the art should understand that the various modules or steps of the present invention described above can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0111] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative efforts on the basis of the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A data-driven predictive control method for a three-level inverter, characterized in that, Including: Establish a compact - form dynamic linearization prediction model considering the unknown uncertainties of the system. The model includes the residual unknown uncertainty term of the system; Take the unknown uncertainty term of the system as the extended state of the system, design a generalized extended observer to estimate this state, compensate the calculation result into the established prediction model, calculate the reference voltage vector in the next switching period and judge the sector it is in; Eliminate the redundant small vectors with higher common - mode amplitude in each sector of the space vector of the three - level inverter, and construct virtual vectors with enhanced mid - point balancing ability using the large vectors and the two small vectors adjacent to the large vectors in each sector. The constructed virtual vectors are used to replace the eliminated small vectors; Compare the upper and lower capacitor voltages V p and V n on the current DC side. If V p > V n and the reference vector is located in sectors I, III, or V, or V p < V n and the reference vector is located in sectors II, IV, or VI, add the virtual small vectors constructed in each sector to the candidate vectors; otherwise, add the small vectors in the sector where it is located to the candidate vectors; Select the optimal vector from the candidate vectors as the next switching action vector of the three - level inverter, and apply the selected optimal vector and its corresponding action time to the control of the switching tubes in the next period of the three - level inverter.

2. The data-driven predictive control method of a three-level inverter according to claim 1, characterized in that When establishing a compact - form dynamic linearization prediction model considering the unknown uncertainties of the system, it includes: Transform the historical output three - phase current data of the three - level inverter to the two - phase stationary coordinate system through coordinate transformation; Based on the transformed data, set the output reference curve, establish a compact - form dynamic linearization prediction model considering the unknown uncertainties of the system, estimate the value of the pseudo - partial derivative matrix, and construct the calculation expression of the reference input. The established prediction model does not depend on any system parameters.

3. A data-driven predictive control method for a three-level inverter as claimed in claim 1, characterized in that, Take the unknown uncertainty term of the system as the extended state of the system. Then, transform the established compact - form dynamic linearization model into the state - space form; design a generalized extended observer based on the state - space form to observe the extended state of the system; based on the observation of the extended state of the system by the generalized extended observer, use the calculation result of the unknown uncertainty term of the system for compensation, calculate the reference voltage vector in the next switching period and judge the sector it is in.

4. A data-driven predictive control method for a three-level inverter as claimed in claim 1, characterized in that, Select three vectors from the candidate vectors as the next switching action vectors of the three - level inverter.

5. A data-driven predictive control method for a three-level inverter as claimed in claim 4, characterized in that, Take the small vector or virtual small vector in the candidate vectors as the first vector to improve the mid - point balancing ability.

6. The data-driven predictive control method for a three-level inverter according to claim 5, characterized in that Construct a simplified cost function without a weighting factor, calculate the cost function values of the remaining candidate vectors respectively, select two vectors with the minimum cost function as the second and third vectors, and calculate the action time of each vector according to the volt - second balance principle; Apply the selected three optimal vectors and their corresponding action times to the control of the switching tubes in the next period of the three - level inverter.

7. A data - driven predictive control system for a three - level inverter, characterized by including: A prediction model establishment module, configured to: establish a compact - form dynamic linearization prediction model considering the unknown uncertainties of the system. The model includes the residual unknown uncertainty term of the system; A reference voltage vector calculation module, configured to: take the unknown uncertainty term of the system as the extended state of the system, design a generalized extended observer to estimate this state, compensate the calculation result into the established prediction model, calculate the reference voltage vector in the next switching period and judge the sector it is in; The candidate vector selection module is configured to: eliminate redundant small vectors with higher common-mode amplitudes in each sector of the space vectors of the three-level inverter, and construct virtual vectors with enhanced midpoint balancing ability by using the large vectors and the two small vectors adjacent to the large vectors in each sector, and the constructed virtual vectors are used to replace the eliminated small vectors; Compare the magnitudes of the upper and lower DC-side capacitor voltages V p and V n . If V p > V n and the reference vector is in sectors I, III, or V, or V p < V n and the reference vector is in sectors II, IV, or VI, add the virtual small vectors constructed in each sector to the candidate vectors; otherwise, add the small vectors of the sector where it is located to the candidate vectors. The active vector calculation module is configured to: select the optimal vector from the candidate vectors as the next switching active vector of the three-level inverter, and apply the selected optimal vector and its corresponding action time to the control of the switching tubes in the next cycle of the three-level inverter.

8. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1-6 above.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it executes the steps of the method according to any one of claims 1-6 above.

Images