Weight factor design method and device based on data driving and model prediction control method and device

Abstract

The invention discloses a weight factor design method based on data driving, and a model prediction control method and device. The model prediction control method comprises the following steps: acquiring a state variable of a T-type three-level energy storage converter at a k moment; establishing a discrete mathematical model of the T-type three-level energy storage converter in a two-phase static alpha-beta coordinate system, and obtaining a reference value of the power grid current at the k + 1 moment according to the Lagrange extrapolation theorem; optimizing a weight factor of a neutral-point voltage balance item on line; determining candidate voltage vectors based on a voltage vector conversion rule; calculating the cost function value of each candidate voltage vector through a cost function, and selecting the voltage vector corresponding to the minimum cost function value as the optimal voltage vector of the kth control period; and applying the gate driving signal corresponding to the optimal voltage vector and the action time of the optimal voltage vector to a power electronic semiconductor device of the T-type three-level energy storage converter. According to the method, the weight factors can be optimized online, and the design process of the weight factors is greatly simplified.

Description

Technical Field

[0001] The present invention belongs to the field of three-level inverter control, and in particular relates to a data-driven weight factor design method, a model predictive control method and a device. Background Art

[0002] With the increasing shortage of traditional fossil energy and the increasingly serious environmental pollution problem worldwide, human beings have to change their traditional thinking and turn their attention to the research of new energy with broader prospects. Wind power and solar photovoltaic power generation have attracted widespread attention among renewable energy sources because of their environmental friendliness. Grid-connected inverters, as the core power conversion device in new energy power generation systems, play an indispensable role. Grid-connected inverters are the bridge for energy transmission between renewable energy power generation systems and traditional power systems. They can convert the electric energy in renewable energy power generation systems into alternating current and feed it into traditional power systems. In order to achieve high-efficiency power generation, high reliability and low current harmonics in new energy power generation systems, the performance improvement of grid-connected inverters is crucial for the stability of new energy power generation systems and new power systems.

[0003] Model predictive control is a promising converter control strategy with the following advantages: easy implementation, fast dynamic response, no proportional integral regulator, simple parameter adjustment, and the ability to handle multiple constraints in a single control loop. Model predictive control can achieve multi-objective control in one control loop and adjust the priority of each control objective by adjusting the weight factor in the cost function.

[0004] However, the adjustment process of the weight coefficient is rather complicated and lacks reliable theoretical guidance. Summary of the invention

[0005] The purpose of the present invention is to provide a data-driven weight factor design method, a model predictive control method and a device. It is used to solve the problem that the weight factor adjustment in the traditional model predictive control method is cumbersome and lacks theoretical basis. The data-driven weight factor design method proposed in the present invention can avoid the cumbersome design process of the weight factor. The weight factor can be optimized online using only the input and output data of the converter.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, the present invention provides a data-driven weight factor design method, the method comprising:

[0008] The average value of the absolute value of the midpoint voltage is calculated online; the average value of the absolute value of the midpoint voltage is:

[0009]

[0010] In the formula, u av is the average value of the absolute value of the midpoint voltage, n = T w / T s , is the total number of statistics, T w is the statistical time, usually the control period T s An integer multiple of ;

[0011] The average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term are expressed in a nonlinear form; the nonlinear form of the average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term is expressed as:

[0012] u av (k+1)=f(u av (k),u av (k-1),…,u av (kn i ),k u (k),k u (k-1),…,k u (kn i ))

[0013] In the formula, u av (k+1) represents the average absolute value of the midpoint voltage of the three-level inverter at time k+1, u av (kn i ), k u (kn i ) represent the three-level inverter in kn i The average value of the absolute value of the midpoint voltage at the moment and the weight factor of the midpoint voltage balance term, f() represents the nonlinear equation of the system, n i is a positive integer related to the system;

[0014] Define the error between the average value of the absolute value of the midpoint voltage of the variable and the weight factor of the midpoint voltage balance term, and make it have a time-varying partial derivative to meet the requirements of the nonlinear system;

[0015] The error between the average value of the absolute value of the midpoint voltage of the definition variable and the weight factor of the midpoint voltage balance term is:

[0016]

[0017] In the formula, Δu av (k+1) is the error of the average value of the absolute value of the midpoint voltage at time k+1, Δk u (k+1) is the error of the weight factor of the midpoint voltage balance term at time k+1;

[0018] For a nonlinear system, there exists a time-varying partial derivative φ such that the system satisfies the following equation:

[0019] Δu av (k+1)=φ(k)Δk u (k)

[0020] In the formula, φ(k) is the partial derivative value at time k, Δk u (k) is the error of the weight factor of the midpoint voltage balance term at time k;

[0021] According to the model-free adaptive control theory, the cost function of the pseudo partial derivative is set, and the partial derivative is derived to obtain the optimal value of the partial derivative;

[0022] According to the model-free adaptive control theory, the cost function of the weight factor of the midpoint voltage balance term is set, and the weight factor of the midpoint voltage balance term is derived after substituting the optimal value of the partial derivative to obtain the optimal value of the weight factor of the midpoint voltage balance term.

[0023] In one embodiment, the cost function of the pseudo partial derivative is set as:

[0024] J(φ(k))=|u av (k)-u av (k-1)-φ(k)Δk u (k-1) 2 +μφ(k)-φ(k-1) 2

[0025] Where μ>0 is the weight factor of the pseudo partial derivative;

[0026] The derivative of the partial derivative is obtained as:

[0027]

[0028] Let the derivative of the partial derivative be equal to zero, and find the optimal value of the partial derivative:

[0029]

[0030] Where φ(k) is the partial derivative value at time k, φ(k-1) is the partial derivative value at time k-1, and Δu av (k) is the error of the average value of the absolute value of the midpoint voltage at time k, Δk u (k-1) is the error of the weight factor of the midpoint voltage balance term at time k-1, and η is the adjustment factor of the partial derivative.

[0031] In one implementation, the cost function for setting the weight factor of the midpoint voltage balance term is:

[0032]

[0033] Where, λ>0 is the weight factor of the midpoint voltage balance term;

[0034] The derivative of the weight factor of the midpoint voltage balance term is:

[0035]

[0036] Let the derivative of the weight factor of the midpoint voltage balance term be equal to zero, and find the optimal value of the weight factor of the midpoint voltage balance term:

[0037]

[0038] Where ρ is the adjustment factor of the weight factor of the midpoint voltage balance term, k u (k) is the weight factor of the midpoint voltage balance term at time k, k u (k-1) is the weight factor of the midpoint voltage balance term at time k-1, φ(k) is the partial derivative value at time k, u av (k) is the average value of the absolute value of the midpoint voltage at time k, It is the reference value of the average value of the absolute value of the midpoint voltage at time k+1.

[0039] In a second aspect, the present invention provides a T-type three-level energy storage converter model predictive control method, the method comprising:

[0040] Collecting state variables of the T-type three-level energy storage converter at time k; the state variables include grid current, grid-side voltage and capacitor voltage;

[0041] A discrete mathematical model of a T-type three-level energy storage converter in a two-phase stationary α-β coordinate system is established, and the reference value of the grid current at time k+1 is obtained according to the Lagrange extrapolation theorem.

[0042] The weight factor of the midpoint voltage balance term is optimized online by using the above data-driven weight factor design method;

[0043] Based on the voltage vector conversion rule, a candidate voltage vector is determined;

[0044] A cost function is constructed according to the predicted values ​​of the grid current and midpoint voltage and the reference value of the grid current at time k+1, and the cost function value of each candidate voltage vector is calculated by the cost function, and the voltage vector corresponding to the minimum cost function value is selected as the optimal voltage vector of the kth control cycle;

[0045] The optimal voltage vector and the gate driving signal corresponding to the action time thereof are applied to the power electronic semiconductor device of the T-type three-level energy storage converter.

[0046] In one implementation, the discrete mathematical model of the T-type three-level energy storage converter in a two-phase static α-β coordinate system is:

[0047]

[0048] In the formula, L and R are the inductance value and equivalent resistance of the filter inductor respectively, T s is the sampling period; i α (k), i β (k) is the grid current sampling value in the α-β coordinate system at time k; i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1; e α (k), e β (k) is the grid voltage sampling value in the α-β coordinate system at time k; u α (k),u β (k) is the AC side output voltage of the converter in the α-β coordinate system at time k, u o (k),u o (k+1) are the DC side midpoint voltage values ​​at time k and k+1 respectively; i o (k) is the midpoint current value of the DC side at time k; C1 and C2 are the capacitance values ​​of the upper and lower bus capacitors of the DC side respectively; T s is the control period of the control system.

[0049] In one implementation, the reference value of the grid current at time k+1 calculated according to the Lagrange extrapolation theorem is:

[0050]

[0051] In the formula, are the reference values ​​of the α-axis grid current at the time k+1, k, k-1, and k-2 respectively, are the reference values ​​of the β-axis grid current at moments k+1, k, k-1, and k-2 respectively.

[0052] In one implementation, the voltage vector conversion rule is:

[0053] Only the voltage vector adjacent to the optimal voltage vector at the previous moment is selected, and the change of the phase voltage and line voltage of the three-level energy storage converter cannot exceed u dc The voltage vector required by the voltage change of / 2 is taken as the candidate voltage vector, and the zero voltage vector only considers the "000" state.

[0054] In one implementation, the candidate voltage vector is determined based on the voltage vector conversion rule as:

[0055] The optimal voltage vector for the k-1 cycle is: V1, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V5, V6, V7, V8, V9, V 10 、V11 、V 12 、V 13 、V 14 、V 15 ;

[0056] The optimal voltage vector for the k-1 cycle is: V2, and the corresponding candidate voltage vector for the k cycle is: None;

[0057] The optimal voltage vector for the k-1 cycle is: V3, and the corresponding candidate voltage vector for the k cycle is: None;

[0058] The optimal voltage vector for the k-1 cycle is: V4, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V6, V7, V 14 、V 15 、V 16 、V 21 、V 22 ;

[0059] The optimal voltage vector for the k-1 cycle is: V5, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V5, V6, V 14 、V 16 、V 21 、V 22 ;

[0060] The optimal voltage vector for the k-1 cycle is: V6, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V5, V6, V8, V9, V 16 、V 17 、V 23 ;

[0061] The optimal voltage vector for the k-1 cycle is: V7, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V7, V8, V 16 、V 17 、V 23 ;

[0062] The optimal voltage vector for the k-1 cycle is: V8, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V6, V7, V8, V 10 、V 11 、V 17 、V 18 、V 24 ;

[0063] The optimal voltage vector for the k-1 cycle is: V9, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V6, V9, V 10 、V 17 、V 18 、V 24 ;

[0064] The optimal voltage vector for k-1 period is: V 10, the corresponding k-cycle candidate voltage vectors are: V1, V8, V9, V 10 、V 12 、V 13 、V 18 、V 19 、V 25 ;

[0065] The optimal voltage vector for k-1 period is: V 11 , the corresponding k-cycle candidate voltage vectors are: V1, V8, V 11 、V 12 、V 18 、V 19 、V 25 ;

[0066] The optimal voltage vector for k-1 period is: V 12 , the corresponding k-cycle candidate voltage vectors are: V1, V 10 、V 11 、V 12 、V 14 、V 15 、V 19 、V 20 、V 26 ;

[0067] The optimal voltage vector for k-1 period is: V 13 , the corresponding k-cycle candidate voltage vectors are: V1, V 10 、V 13 、V 14 、V 19 、V 20 、V 26 ;

[0068] The optimal voltage vector for k-1 period is: V 14 , the corresponding k-cycle candidate voltage vectors are: V1, V4, V5, V 12 、V 13 、V 14 、V 20 、V 21 、V 27 ;

[0069] The optimal voltage vector for k-1 period is: V 15 , the corresponding k-cycle candidate voltage vectors are: V1, V4, V 12 、V 15 、V 20 、V 21 、V 27 ;

[0070] The optimal voltage vector for k-1 period is: V 16, the corresponding k-cycle candidate voltage vectors are: V4, V5, V6, V7, V 16 、V 22 、V 23 ;

[0071] The optimal voltage vector for k-1 period is: V 17 , the corresponding k-cycle candidate voltage vectors are: V6, V7, V8, V9, V 17 、V 23 、V 24 ;

[0072] The optimal voltage vector for k-1 period is: V 18 , the corresponding k-cycle candidate voltage vectors are: V8, V9, V 10 、V 11 、V 18 、V 24 、V 25 ;

[0073] The optimal voltage vector for k-1 period is: V 19 , the corresponding k-cycle candidate voltage vector is: V 10 、V 11 、V 12 、V 13 、V 19 、V 25 、V 26 ;

[0074] The optimal voltage vector for k-1 period is: V 20 , the corresponding k-cycle candidate voltage vector is: V 12 、V 13 、V 14 、V 15 、V 20 、V 26 、V 27 ;

[0075] The optimal voltage vector for k-1 period is: V 21 , the corresponding k-cycle candidate voltage vectors are: V4, V5, V 14 、V 15 、V 21 、V 22 、V 27 ;

[0076] The optimal voltage vector for k-1 period is: V 22 , the corresponding k-cycle candidate voltage vectors are: V4, V5, V 16 、V 21 、V 22 ;

[0077] The optimal voltage vector for k-1 period is: V23 , the corresponding k-cycle candidate voltage vectors are: V6, V7, V 16 、V 17 、V 23 ;

[0078] The optimal voltage vector for k-1 period is: V 24 , the corresponding k-cycle candidate voltage vectors are: V8, V9, V 17 、V 18 、V 24 ;

[0079] The optimal voltage vector for k-1 period is: V 25 , the corresponding k-cycle candidate voltage vector is: V 10 、V 11 、V 18 、V 19 、V 25 ;

[0080] The optimal voltage vector for k-1 period is: V 26 , the corresponding k-cycle candidate voltage vector is: V 12 、V 13 、V 19 、V 20 、V 26 ;

[0081] The optimal voltage vector for k-1 period is: V 27 , the corresponding k-cycle candidate voltage vector is: V 13 、V 14 、V 20 、V 21 、V 27 .

[0082] In one embodiment, the cost function is:

[0083]

[0084] In the formula, is the grid current reference value in the α-β coordinate system at time k+1, i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1, k u (k) is the optimal value of the weight factor at time k, u o (k+1) is the predicted value of the midpoint voltage at time k+1.

[0085] In a third aspect, the present invention provides a T-type three-level energy storage converter model predictive control device, the device comprising:

[0086] A collection module, used to collect state variables of the T-type three-level energy storage converter at time k; the state variables include grid current, grid-side voltage and capacitor voltage;

[0087] The discrete mathematical model module is used to establish a discrete mathematical model of the T-type three-level energy storage converter in a two-phase static α-β coordinate system, and obtain the reference value of the grid current at time k+1 according to the Lagrange extrapolation theorem;

[0088] The weight factor online optimization module uses the above-mentioned data-driven weight factor design method to online optimize the weight factor of the midpoint voltage balance item;

[0089] A candidate voltage vector selection module, used for determining a candidate voltage vector based on a voltage vector conversion rule;

[0090] The optimal voltage vector selection module is used to construct a cost function according to the predicted values ​​of the grid current and the midpoint voltage and the reference value of the grid current at time k+1, calculate the cost function value of each candidate voltage vector through the cost function, and select the voltage vector corresponding to the minimum cost function value as the optimal voltage vector of the kth control cycle;

[0091] The action module is used to apply the optimal voltage vector and the gate drive signal corresponding to the action time thereof to the power electronic semiconductor device of the T-type three-level energy storage converter.

[0092] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0093] The present invention can reduce current ripple and voltage fluctuation under the premise of achieving capacitor voltage balance and stable operation of the three-level converter. The data-driven weight factor design method proposed in the present invention can optimize the weight factor online, solving the problem of cumbersome weight factor adjustment and low flexibility in the traditional model predictive control method. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] The accompanying drawings are part of the present invention and are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but do not constitute an improper limitation of the present invention. Obviously, the accompanying drawings described below are only some embodiments. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0095] Figure 1 A topological diagram of a three-level converter provided in an embodiment of the present invention;

[0096] Figure 2 A voltage vector distribution diagram of a three-level converter provided in an embodiment of the present invention;

[0097] Figure 3 A flow chart of a method for designing weight factors in a data-driven predictive control according to an embodiment of the present invention;

[0098] Figure 4 A voltage vector switching rule diagram of a three-level converter provided in an embodiment of the present invention;

[0099] Figure 5 It is a comparison diagram of steady-state current simulation results of the finite set model predictive control and the data-driven model predictive control provided in the embodiment of the present invention; wherein Figure 5 (a) is the steady-state simulation waveform of the finite set model predictive control. Figure 5 (b) is the steady-state simulation waveform based on data-driven model predictive control;

[0100] Figure 6 The figure is a comparison diagram of the steady-state capacitor voltage simulation results of the finite set model predictive control and the data-driven model predictive control provided in the embodiment of the present invention, wherein Figure 6 (a) is the steady-state simulation waveform of the finite set model predictive control. Figure 6 (b) is the steady-state simulation waveform based on data-driven model predictive control.

[0101] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0102] In order to deepen the recognition and understanding of the present invention, the technical solution of the present invention is further introduced below in conjunction with the accompanying drawings and specific implementation methods. The following embodiments are explanations of the present invention and the present invention is not limited to the following embodiments.

[0103] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0104] An embodiment of the present invention provides a data-driven weight factor design method, the method comprising:

[0105] The average value of the absolute value of the midpoint voltage is calculated online; the average value of the absolute value of the midpoint voltage is:

[0106]

[0107] In the formula, u av is the average value of the absolute value of the midpoint voltage, n = T w / T s, is the total number of statistics, T w is the statistical time, usually the control period T s An integer multiple of ;

[0108] The average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term are expressed in a nonlinear form; the nonlinear form of the average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term is expressed as:

[0109] u av (k+1)=f(u av (k),u av (k-1),…,u av (kn i ),k u (k),k u (k-1),…,k u (kn i ))

[0110] In the formula, u av (k+1) represents the average absolute value of the midpoint voltage of the three-level inverter at time k+1, u av (kn i ), k u (kn i ) represent the three-level inverter in kn i The average value of the absolute value of the midpoint voltage at the moment and the weight factor of the midpoint voltage balance term, f() represents the nonlinear equation of the system, n i is a positive integer related to the system;

[0111] Define the error between the average value of the absolute value of the midpoint voltage of the variable and the weight factor of the midpoint voltage balance term, and make it have a time-varying partial derivative to meet the requirements of the nonlinear system;

[0112] The error between the average value of the absolute value of the midpoint voltage of the definition variable and the weight factor of the midpoint voltage balance term is:

[0113]

[0114] In the formula, Δu av (k+1) is the error of the average value of the absolute value of the midpoint voltage at time k+1, Δk u (k+1) is the error of the weight factor of the midpoint voltage balance term at time k+1;

[0115] For a nonlinear system, there exists a time-varying partial derivative φ such that the system satisfies the following equation:

[0116] Δu av (k+1)=φ(k)Δk u (k)

[0117] In the formula, φ(k) is the partial derivative value at time k, Δk u (k) is the error of the weight factor of the midpoint voltage balance term at time k;

[0118] According to the model-free adaptive control theory, the cost function of the pseudo partial derivative is set, and the partial derivative is derived to obtain the optimal value of the partial derivative;

[0119] According to the model-free adaptive control theory, the cost function of the weight factor of the midpoint voltage balance term is set, and the weight factor of the midpoint voltage balance term is derived after substituting the optimal value of the partial derivative to obtain the optimal value of the weight factor of the midpoint voltage balance term.

[0120] Furthermore, the cost function of the pseudo partial derivative is set as:

[0121] J(φ(k))=|u av (k)-u av (k-1)-φ(k)Δk u (k-1) 2 +μφ(k)-φ(k-1) 2

[0122] Where μ>0 is the weight factor of the pseudo partial derivative;

[0123] The derivative of the partial derivative is obtained as:

[0124]

[0125] Let the derivative of the partial derivative be equal to zero, and find the optimal value of the partial derivative:

[0126]

[0127] Where φ(k) is the partial derivative value at time k, φ(k-1) is the partial derivative value at time k-1, and Δu av (k) is the error of the average value of the absolute value of the midpoint voltage at time k, Δk u (k-1) is the error of the weight factor of the midpoint voltage balance term at time k-1, and η is the adjustment factor of the partial derivative.

[0128] Furthermore, the cost function of the weight factor of the midpoint voltage balance term is set as:

[0129]

[0130] Where, λ>0 is the weight factor of the midpoint voltage balance term;

[0131] The derivative of the weight factor of the midpoint voltage balance term is:

[0132]

[0133] Let the derivative of the weight factor of the midpoint voltage balance term be equal to zero, and find the optimal value of the weight factor of the midpoint voltage balance term:

[0134]

[0135] Where ρ is the adjustment factor of the weight factor of the midpoint voltage balance term, k u (k) is the weight factor of the midpoint voltage balance term at time k, k u (k-1) is the weight factor of the midpoint voltage balance term at time k-1, φ(k) is the partial derivative value at time k, u av (k) is the average value of the absolute value of the midpoint voltage at time k, It is the reference value of the average value of the absolute value of the midpoint voltage at time k+1.

[0136] The data-driven weight factor design method proposed in the present invention can avoid the cumbersome design process of the weight factor.

[0137] Figure 1 The topology diagram of the T-type three-level energy storage converter is given. Each phase bridge arm consists of four power electronic semiconductor devices, and the DC side is composed of capacitors C1 and C2 in series, and the parameters of C1 and C2 are consistent. The midpoint of the DC side is defined as o, and the potential is zero. The grid-side filtering link consists of a single L-type filter, and the resistor R is the equivalent resistance of the filter. x ,i x are the grid voltage and current respectively, (x=a,b,c). c1 and u c2 are the voltages of capacitors C1 and C2, u dc is a DC voltage. Each bridge arm of the T-type three-level energy storage converter has three effective switching states: S x ∈{1, 0, -1}. Assuming that the capacitor voltage is balanced, that is, the voltage of each capacitor is equal to half of the DC voltage, the output voltage generated by these three switching states is u xo ∈{u dc / 2,0,-u dc / 2}. Table 1 summarizes the relationship between the output voltage, switch state and gate signal generated by the converter.

[0138] Table 1 Relationship between output voltage, switch state and gate signal, x = (a, b, c)

[0139]

[0140] In order to establish the mathematical model of the T-type three-level energy storage converter, it is assumed that the power electronic semiconductor devices are ideal switching devices, and the nonlinear effects such as the on-state voltage drop and switching delay of the power electronic semiconductor devices are ignored.

[0141] According to Kirchhoff's voltage law, in the three-phase stationary coordinate system, the dynamic response of the grid current is:

[0142]

[0143] The switching states of the T-type three-level energy storage converter are 27 (3 3 ) species, such as Figure 2 As shown in the α-β coordinate system, these switching states generate a total of 19 non-redundant voltage vectors and 8 redundant voltage vectors. According to the amplitude of the output voltage, the voltage vector is divided into zero vector (V1-V3), small vector (V4-V 15 ), mid vector (V 16 -V 21 ) and large vector (V 22 -V 27 ). There is only one zero vector, and there are two redundant vectors; there are six small vectors, and each has one redundant vector; there are six medium vectors and six large vectors, and there is no redundant vector. The voltage vectors are shown in Table 2.

[0144] Table 2 Voltage vector table

[0145]

[0146] like Figure 3 As shown, the disclosed embodiment provides a T-type three-level energy storage converter model predictive control method, the method specifically comprising the following steps:

[0147] Step S100: collecting state variables of a T-type three-level energy storage converter at time k.

[0148] Specifically, the grid current i at time k is collected by voltage and current sensors. a (k), i b (k), i c (k), grid voltage e a (k), e b (k), e c (k), and capacitor voltage u c1 (k),u c2 (k).

[0149] Step S200: establishing a discrete mathematical model of a T-type three-level energy storage converter in a two-phase stationary α-β coordinate system, and obtaining a reference value of the grid current at time k+1 according to the Lagrange extrapolation theorem.

[0150] The three-phase stationary coordinate system can be converted into a two-phase stationary coordinate system through Clarke transformation. Clarke transformation is:

[0151]

[0152] The mathematical model of the three-level energy storage converter in the two-phase stationary coordinate system is:

[0153]

[0154] Through the forward Euler method, the discrete mathematical model of the T-type three-level energy storage converter in the two-phase stationary coordinate system is:

[0155]

[0156] In the formula, L and R are the inductance value and equivalent resistance of the filter inductor respectively, T s is the sampling period; i α (k), i β (k) is the grid current sampling value in the α-β coordinate system at time k; i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1; e α (k), e β (k) is the grid voltage sampling value in the α-β coordinate system at time k; u α (k),u β (k) is the output voltage of the AC side of the converter in the α-β coordinate system at time k.

[0157]

[0158] The DC side midpoint voltage is:

[0159] u o =u c2 -u c1

[0160] Assuming that the capacitor voltages are equal and the grid currents are balanced, the dynamic process of the DC side midpoint voltage can be modeled as:

[0161]

[0162] i o is the DC side midpoint current, which is given by the following formula:

[0163] i o =|S abc |i abc

[0164] In the formula, |S abc |=[1-|S a|,1-|S b |,1-|S c |],i abc =[i a ,i b ,i c ].

[0165] The predicted value of the midpoint voltage is:

[0166]

[0167] In the formula, u o (k),u o (k+1) are the DC side midpoint voltage values ​​at time k and k+1 respectively; i o (k) is the midpoint current value of the DC side at time k; C1 and C2 are the capacitance values ​​of the upper and lower bus capacitors of the DC side respectively; T s is the control period of the control system.

[0168] According to Lagrange's extrapolation theorem, in the α-β coordinate system, the reference value of the grid current at time k+1 is:

[0169]

[0170] In the formula, in the formula, are the reference values ​​of the α-axis grid current at the time k+1, k, k-1, and k-2 respectively, are the reference values ​​of the β-axis grid current at moments k+1, k, k-1, and k-2 respectively.

[0171] Step S300: using a data-driven weight factor design method to online optimize the weight factor of the midpoint voltage balance item.

[0172] The data-driven weight factor design method is described in the same or similar parts as above and will not be repeated here.

[0173] The average value of the absolute value of the midpoint voltage is calculated online, and the weight factor k of the midpoint voltage balance term in the cost function is adjusted online based on the data-driven principle. u , good control effect can be obtained.

[0174] Through the above calculation process, the weight factor k of the midpoint voltage balance term can be optimized online. u . k u It will participate in the subsequent cost function calculation process to help select the optimal voltage vector.

[0175] Step S400: selecting a candidate voltage vector based on a voltage vector switching rule.

[0176] In the application embodiment, the voltage vector switching rules are formulated by comprehensively considering the characteristics of the three-level energy storage converter itself, reducing the damage to power electronic semiconductor devices and the impact on the power grid, and reducing the complexity and calculation amount of the algorithm.

[0177] The change of phase voltage and line voltage of three-level energy storage converter cannot exceed u dc / 2, because the former may damage the switching devices of the converter, and the latter will have a greater impact on the power grid. Therefore, for the three-level energy storage converter, not all voltage vectors can be directly converted. In addition, in order to reduce the amount of calculation of the algorithm, the candidate voltage vectors of each cycle should be optimized under the premise of meeting the above requirements.

[0178] The voltage vector switching rule in the present invention is as follows: only the voltage vectors that are adjacent to the optimal voltage vector at the previous moment and meet the voltage change requirements in the above analysis are selected as candidate voltage vectors. Figure 4 As shown, in order to reduce the common mode voltage and switching frequency, the zero voltage vector only considers the "000" state.

[0179] Specifically, the candidate voltage vectors corresponding to each voltage vector are shown in Table 3.

[0180] Table 3 Candidate voltage vector switching rules

[0181]

[0182]

[0183] Step S500: construct a cost function based on the predicted values ​​of the grid current and midpoint voltage and the reference value of the grid current at time k+1, calculate the cost function value of each candidate voltage vector through the cost function, and select the voltage vector corresponding to the minimum cost function value as the optimal voltage vector for the kth control cycle.

[0184] Furthermore, the cost function is:

[0185]

[0186] In the formula, is the grid current reference value in the α-β coordinate system at time k+1, i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1, k u (k) is the optimal value of the weight factor at time k, u o (k+1) is the predicted value of the midpoint voltage at time k+1.

[0187] where k uIt is calculated in step S300 and is used to adjust the control priority between the grid current and the midpoint voltage.

[0188] Step S600: applying the optimal voltage vector and the gate driving signal corresponding to the action time thereof to the power electronic semiconductor device of the T-type three-level energy storage converter.

[0189] After a series of calculations in steps S100-S600, the control cycle value, comparison value and other information are transmitted to the FPGA through the communication module of the DSP, and the FPGA is used to implement the pulse distribution of the power device, the dead zone delay and other functions.

[0190] The following is an embodiment of the T-type three-level energy storage converter model predictive control device of the present invention, which can be used to execute the embodiment of the T-type three-level energy storage converter model predictive control method of the present invention. For details not disclosed in the embodiment of the T-type three-level energy storage converter model predictive control device of the present invention, please refer to the embodiment of the T-type three-level energy storage converter model predictive control method of the present invention.

[0191] In one embodiment, a T-type three-level energy storage converter model predictive control device is proposed, the device comprising:

[0192] A collection module, used to collect state variables of the T-type three-level energy storage converter at time k; the state variables include grid current, grid-side voltage and capacitor voltage;

[0193] The discrete mathematical model module is used to establish a discrete mathematical model of the T-type three-level energy storage converter in a two-phase static α-β coordinate system, and obtain the reference value of the grid current at time k+1 according to the Lagrange extrapolation theorem;

[0194] The weight factor online optimization module uses the above-mentioned data-driven weight factor design method to online optimize the weight factor of the midpoint voltage balance item;

[0195] A candidate voltage vector selection module, used for determining a candidate voltage vector based on a voltage vector conversion rule;

[0196] The optimal voltage vector selection module is used to construct a cost function according to the predicted values ​​of the grid current and the midpoint voltage and the reference value of the grid current at time k+1, calculate the cost function value of each candidate voltage vector through the cost function, and select the voltage vector corresponding to the minimum cost function value as the optimal voltage vector of the kth control cycle;

[0197] The action module is used to apply the optimal voltage vector and the gate drive signal corresponding to the action time thereof to the power electronic semiconductor device of the T-type three-level energy storage converter.

[0198] The various functional modules in the embodiments of the present invention may be integrated into one processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0199] As a specific example of the present invention, a three-level energy storage converter simulation model was built in MATLAB / Simulink to test and verify the present invention. The AC side of the three-level energy storage converter is connected to a 220V AC power grid, and the DC side is powered by a battery with a voltage of 600V. The DC side capacitor is 2200μF, and the filter inductor is 4mH. The sampling and control period is 100μs. In order to highlight the advantages of the present invention in improving current quality, the traditional finite control set model predictive control and the data-driven model predictive control method proposed in the present invention are compared.

[0200] Figure 5 The following is a comparison of the steady-state simulation results of the three-phase grid current using the traditional finite control set model predictive control and the data-driven model predictive control when the grid reference current is 50A. Figure 5 (a) is the steady-state simulation waveform of the finite set model predictive control. Figure 5 (b) is the steady-state simulation waveform of the data-driven model predictive control. It can be concluded from the simulation waveform that both control strategies can achieve stable control of the three-level converter. The current THD of the finite set model predictive control and the data-driven model predictive control are 3.14% and 2.87% respectively. It can be seen that the data-driven model predictive control has better current quality.

[0201] Figure 6 This is a comparison chart of the steady-state capacitor voltage simulation results of the traditional finite control set model predictive control and the data-driven model predictive control. Figure 6 (a) is the steady-state simulation waveform of the finite set model predictive control. Figure 6 (b) is the steady-state simulation waveform of the data-driven model predictive control. It can be concluded from the simulation waveform that both control strategies can achieve balanced control of the capacitor voltage of the three-level converter. The maximum difference in capacitor voltage of the traditional finite control set model predictive control is 6V, and the maximum difference in capacitor voltage of the data-driven model predictive control is 3V. The data-driven model predictive control has a better capacitor voltage control effect.

[0202] The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this patent can make some changes or modify the technical contents suggested above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the invention without departing from the content of the technical solution of the present invention still fall within the scope of the solution of the present invention.

Claims

1. A data-driven weight factor design method, characterized in that: The method comprises: The average value of the absolute value of the midpoint voltage is calculated online; the average value of the absolute value of the midpoint voltage is: In the formula, u av is the average value of the absolute value of the midpoint voltage, n = T w / T s , is the total number of statistics, T w is the statistical time, usually the control period T s An integer multiple of ; The average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term are expressed in a nonlinear form; the nonlinear form of the average value of the midpoint voltage absolute value and the weight factor of the midpoint voltage balance term is expressed as: u av (k+1)=f(u av (k),u av (k-1),…,u av (k-n i ),k u (k),k u (k-1),…,k u (k-n i )) In the formula, u av (k+1) represents the average absolute value of the midpoint voltage of the three-level inverter at time k+1, u av (kn i ), k u (kn i ) represent the three-level inverter in kn i The average value of the absolute value of the midpoint voltage at the moment and the weight factor of the midpoint voltage balance term, f() represents the nonlinear equation of the system, n i is a positive integer related to the system; Define the error between the average value of the absolute value of the midpoint voltage of the variable and the weight factor of the midpoint voltage balance term, and make it have a time-varying partial derivative to meet the requirements of the nonlinear system; The error between the average value of the absolute value of the midpoint voltage of the definition variable and the weight factor of the midpoint voltage balance term is: In the formula, Δu av (k+1) is the error of the average value of the absolute value of the midpoint voltage at time k+1, Δk u (k+1) is the error of the weight factor of the midpoint voltage balance term at time k+1; For a nonlinear system, there exists a time-varying partial derivative φ such that the system satisfies the following equation: Δu av (k+1)=φ(k)∆k u (k) In the formula, φ(k) is the partial derivative value at time k, Δk u (k) is the error of the weight factor of the midpoint voltage balance term at time k; According to the model-free adaptive control theory, the cost function of the pseudo partial derivative is set, and the partial derivative is derived to obtain the optimal value of the partial derivative; According to the model-free adaptive control theory, the cost function of the weight factor of the midpoint voltage balance term is set, and the weight factor of the midpoint voltage balance term is derived after substituting the optimal value of the partial derivative to obtain the optimal value of the weight factor of the midpoint voltage balance term.

2. The data-driven weight factor design method according to claim 1, characterized in that: The cost function of the pseudo partial derivative is set to: J(φ(k))=|u av (k)-u av (k-1)-φ(k)Δk u (k-1) 2 +μφ(k)-φ(k-1) 2 Where μ>0 is the weight factor of the pseudo partial derivative; The derivative of the partial derivative is obtained as: Let the derivative of the partial derivative be equal to zero, and find the optimal value of the partial derivative: Where φ(k) is the partial derivative value at time k, φ(k-1) is the partial derivative value at time k-1, and Δu av (k) is the error of the average value of the absolute value of the midpoint voltage at time k, Δk u (k-1) is the error of the weight factor of the midpoint voltage balance term at time k-1, and η is the adjustment factor of the partial derivative.

3. The data-driven weight factor design method according to claim 1, characterized in that: The cost function for setting the weight factor of the midpoint voltage balance term is: Where, λ>0 is the weight factor of the midpoint voltage balance term; The derivative of the weight factor of the midpoint voltage balance term is: Let the derivative of the weight factor of the midpoint voltage balance term be equal to zero, and find the optimal value of the weight factor of the midpoint voltage balance term: Where ρ is the adjustment factor of the weight factor of the midpoint voltage balance term, k u (k) is the weight factor of the midpoint voltage balance term at time k, k u (k-1) is the weight factor of the midpoint voltage balance term at time k-1, φ(k) is the partial derivative value at time k, u av (k) is the average value of the absolute value of the midpoint voltage at time k, It is the reference value of the average value of the absolute value of the midpoint voltage at time k+1.

4. A T-type three-level energy storage converter model predictive control method, characterized in that: The method comprises: Collecting state variables of the T-type three-level energy storage converter at time k; the state variables include grid current, grid-side voltage and capacitor voltage; A discrete mathematical model of a T-type three-level energy storage converter in a two-phase stationary α-β coordinate system is established, and the reference value of the grid current at time k+1 is obtained according to the Lagrange extrapolation theorem. Adopting the data-driven weight factor design method according to any one of claims 1 to 3 to optimize the weight factor of the midpoint voltage balance item online; Based on the voltage vector conversion rule, determining a candidate voltage vector; A cost function is constructed according to the predicted values ​​of the grid current and midpoint voltage and the reference value of the grid current at time k+1, and the cost function value of each candidate voltage vector is calculated by the cost function, and the voltage vector corresponding to the minimum cost function value is selected as the optimal voltage vector of the kth control cycle; The optimal voltage vector and the gate driving signal corresponding to the action time thereof are applied to the power electronic semiconductor device of the T-type three-level energy storage converter.

5. The T-type three-level energy storage converter model predictive control method according to claim 4, characterized in that: The discrete mathematical model of the T-type three-level energy storage converter in the two-phase static α-β coordinate system is: In the formula, L and R are the inductance value and equivalent resistance of the filter inductor respectively, T s is the sampling period; i α (k), i β (k) is the grid current sampling value in the α-β coordinate system at time k; i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1; e α (k), e β (k) is the grid voltage sampling value in the α-β coordinate system at time k; u α (k),u β (k) is the AC side output voltage of the converter in the α-β coordinate system at time k, u o (k),u o (k+1) are the DC side midpoint voltage values ​​at time k and k+1 respectively; i o (k) is the midpoint current value of the DC side at time k; C1 and C2 are the capacitance values ​​of the upper and lower bus capacitors of the DC side respectively; T s is the control period of the control system.

6. The T-type three-level energy storage converter model predictive control method according to claim 5, characterized in that: According to the Lagrange extrapolation theorem, the reference value of the grid current at time k+1 is obtained as: In the formula, are the reference values ​​of the α-axis grid current at the time k+1, k, k-1, and k-2 respectively, are the reference values ​​of the β-axis grid current at moments k+1, k, k-1, and k-2 respectively.

7. The T-type three-level energy storage converter model predictive control method according to claim 1, characterized in that: The voltage vector conversion rule is: only select the voltage vector adjacent to the optimal voltage vector at the previous moment and satisfy the change of the phase voltage and line voltage of the three-level energy storage converter cannot exceed u dc The voltage vector required by the voltage change of / 2 is taken as the candidate voltage vector, and the zero voltage vector only considers the "000" state.

8. The T-type three-level energy storage converter model predictive control method according to claim 7, characterized in that: Based on the voltage vector conversion rule, the candidate voltage vector is determined as: The optimal voltage vector for the k-1 cycle is: V1, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V5, V6, V7, V8, V9, V 10 、V 11 、V 12 、V 13 、V 14 、V 15 ; The optimal voltage vector for the k-1 cycle is: V2, and the corresponding candidate voltage vector for the k cycle is: None; The optimal voltage vector for the k-1 cycle is: V3, and the corresponding candidate voltage vector for the k cycle is: None; The optimal voltage vector for the k-1 cycle is: V4, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V6, V7, V 14 、V 15 、V 16 、V 21 、V 22 ; The optimal voltage vector for the k-1 cycle is: V5, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V5, V6, V 14 、V 16 、V 21 、V 22 ; The optimal voltage vector for the k-1 cycle is: V6, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V5, V6, V8, V9, V 16 、V 17 、V 23 ; The optimal voltage vector for the k-1 cycle is: V7, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V4, V7, V8, V 16 、V 17 、V 23 ; The optimal voltage vector for the k-1 cycle is: V8, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V6, V7, V8, V 10 、V 11 、V 17 、V 18 、V 24 ; The optimal voltage vector for the k-1 cycle is: V9, and the corresponding candidate voltage vectors for the k-1 cycle are: V1, V6, V9, V 10 、V 17 、V 18 、V 24 ; The optimal voltage vector for k-1 period is: V 10 , the corresponding k-cycle candidate voltage vectors are: V1, V8, V9, V 10 、V 12 、V 13 、V 18 、V 19 、V 25 ; The optimal voltage vector for k-1 period is: V 11 , the corresponding k-cycle candidate voltage vectors are: V1, V8, V 11 、V 12 、V 18 、V 19 、V 25 ; The optimal voltage vector for k-1 period is: V 12 , the corresponding k-cycle candidate voltage vectors are: V1, V 10 、V 11 、V 12 、V 14 、V 15 、V 19 、V 20 、V 26 ; The optimal voltage vector for k-1 period is: V 13 , the corresponding k-cycle candidate voltage vectors are: V1, V 10 、V 13 、V 14 、V 19 、V 20 、V 26 ; The optimal voltage vector for k-1 period is: V 14 , the corresponding k-cycle candidate voltage vectors are: V1, V4, V5, V 12 、V 13 、V 14 、V 20 、V 21 、V 27 ; The optimal voltage vector for k-1 period is: V 15 , the corresponding k-cycle candidate voltage vectors are: V1, V4, V 12 、V 15 、V 20 、V 21 、V 27 ; The optimal voltage vector for k-1 period is: V 16 , the corresponding k-cycle candidate voltage vectors are: V4, V5, V6, V7, V 16 、V 22 、V 23 ; The optimal voltage vector for k-1 period is: V 17 , the corresponding k-cycle candidate voltage vectors are: V6, V7, V8, V9, V 17 、V 23 、V 24 ; The optimal voltage vector for k-1 period is: V 18 , the corresponding k-cycle candidate voltage vectors are: V8, V9, V 10 、V 11 、V 18 、V 24 、V 25 ; The optimal voltage vector for k-1 period is: V 19 , the corresponding k-cycle candidate voltage vector is: V 10 、V 11 、V 12 、V 13 、V 19 、V 25 、V 26 ; The optimal voltage vector for k-1 period is: V 20 , the corresponding k-cycle candidate voltage vector is: V 12 、V 13 、V 14 、V 15 、V 20 、V 26 、V 27 ; The optimal voltage vector for k-1 period is: V 21 , the corresponding k-cycle candidate voltage vectors are: V4, V5, V 14 、V 15 、V 21 、V 22 、V 27 ; The optimal voltage vector for k-1 period is: V 22 , the corresponding k-cycle candidate voltage vectors are: V4, V5, V 16 、V 21 、V 22 ; The optimal voltage vector for k-1 period is: V 23 , the corresponding k-cycle candidate voltage vectors are: V6, V7, V 16 、V 17 、V 23 ; The optimal voltage vector for k-1 period is: V 24 , the corresponding k-cycle candidate voltage vectors are: V8, V9, V 17 、V 18 、V 24 ; The optimal voltage vector for k-1 period is: V 25 , the corresponding k-cycle candidate voltage vector is: V 10 、V 11 、V 18 、V 19 、V 25 ; The optimal voltage vector for k-1 period is: V 26 , the corresponding k-cycle candidate voltage vector is: V 12 、V 13 、V 19 、V 20 、V 26 ; The optimal voltage vector for k-1 period is: V 27 , the corresponding k-cycle candidate voltage vector is: V 13 、V 14 、V 20 、V 21 、V 27 .

9. The T-type three-level energy storage converter model predictive control method according to claim 4 or 8, characterized in that: The cost function is: In the formula, is the grid current reference value in the α-β coordinate system at time k+1, i α (k+1), i β (k+1) is the predicted value of the grid current in the α-β coordinate system at time k+1, k u (k) is the optimal value of the weight factor at time k, u o (k+1) is the predicted value of the midpoint voltage at time k+1.

10. A T-type three-level energy storage converter model predictive control device, characterized in that: The device comprises: A collection module, used to collect state variables of the T-type three-level energy storage converter at time k; the state variables include grid current, grid-side voltage and capacitor voltage; The discrete mathematical model module is used to establish a discrete mathematical model of the T-type three-level energy storage converter in a two-phase static α-β coordinate system, and obtain the reference value of the grid current at time k+1 according to the Lagrange extrapolation theorem; A weight factor online optimization module, which adopts the data-driven weight factor design method according to any one of claims 1 to 3 to online optimize the weight factor of the midpoint voltage balance item; A candidate voltage vector selection module, used for determining a candidate voltage vector based on a voltage vector conversion rule; The optimal voltage vector selection module is used to construct a cost function according to the predicted values ​​of the grid current and the midpoint voltage and the reference value of the grid current at time k+1, calculate the cost function value of each candidate voltage vector through the cost function, and select the voltage vector corresponding to the minimum cost function value as the optimal voltage vector of the kth control cycle; The action module is used to apply the optimal voltage vector and the gate drive signal corresponding to the action time thereof to the power electronic semiconductor device of the T-type three-level energy storage converter.

Images