Model predictive control method based on discrete space vector modulation

Abstract

The invention discloses a model prediction control method based on discrete space vector modulation, and belongs to the field of power electronics and power transmission, and the method comprises the steps: calculating the number of times n + m through a cost function, and discretizing a two-level voltage vector plane space into 2n + m-1 selectable virtual voltage vectors; the optimal voltage vector search method is a sequence of phase angle first and amplitude later; the phase angle and the amplitude need to be searched for n times and m times of cost function calculation respectively, n > = 2, and m > = 1; comparing two candidate voltage vectors in each iteration, and selecting the candidate voltage vector with a smaller cost function to the next step; firstly, the amplitude is fixed to be (0.5 + 1 / 2m + 1) [mu] m, and an optimal voltage vector phase angle theta opt is searched; secondly, a fixed phase angle is theta opt, and an optimal voltage vector amplitude uopt is searched; each phase of bridge arm switching pulse signal is generated by the output voltage vector, and a voltage waveform is output. Compared with an existing method, under the same cost function calculation times, the optimal vector search precision can be greatly improved, the harmonic content of the output waveform is reduced, and the control performance is improved.

Description

Technical Field

[0001] The present invention belongs to the field of power electronics and electric drive, and more specifically, relates to a model predictive control method based on discrete space vector modulation. Background Art

[0002] The traditional model predictive control (MPC) method calculates the cost function of 8 fixed voltage vectors (VV), applies the VV with the smallest cost function in the control cycle, further generates the switching pulse signal of each phase bridge arm, and outputs the voltage waveform. However, there is a large deviation between the fixed 8 VVs and the optimal VV, which will generate additional harmonic components and deteriorate the control performance.

[0003] Therefore, in order to improve the above defects, domestic and foreign scholars have conducted relevant research and proposed MPC methods based on dual vectors and three vectors. However, this multi-vector-based MPC method faces a large amount of calculation in each control cycle, and the computing power of microcomputers is limited, which may cause the sum of calculation and action time to exceed the control cycle, and the optimal control cannot be achieved. Further, an MPC method based on discrete space vector modulation was proposed, which discretizes the entire space vector into several points, each point corresponds to an optional virtual VV, and uses the traversal method to find the virtual VV closest to the optimal VV and use it as the application VV. As the degree of discreteness increases, the error between the application VV and the optimal VV gradually decreases, but it also brings a large amount of calculation. Therefore, it is necessary to further improve the search accuracy of the optimal VV, reduce the harmonic content, and improve the control performance under limited calculation. Summary of the invention

[0004] In view of the defects of the prior art, the purpose of the present invention is to provide a model predictive control method based on discrete space vector modulation, aiming to improve the search accuracy of the optimal VV with limited computational effort, reduce the harmonic content and improve the control performance.

[0005] To achieve the above object, the present invention provides a method for searching an optimal voltage vector, comprising the following steps:

[0006] S1-1, fixed voltage vector VV amplitude is (1-1 / 2 m+1 ) m ; Among them, the entire converter output voltage vector space takes the origin as the center, A circular plane space with radius u dc is the DC bus voltage; according to the calculation times n+m of the cost function, m is the amplitude search times, n is the phase angle search times, n≥2, m≥1, 2 n+m-1 discrete points, corresponding to 2 n+m-1A selectable virtual voltage vector;

[0007] S1-2, determine the angle difference Δθ=2π / 2 according to the phase angle search times n n-1 ; Where n≥2;

[0008] S1-3, the phase angle selection range is [0, 2π], and the phase angles of the two candidate VVs V1 and V2 are θ(V1) = Δθ / 2 = 2π / 2 n and θ(V2)=2π-Δθ / 2=2π-2π / 2 n ; Calculate the cost function g respectively, select the candidate VV with the smaller cost function and record it as V opt1 ;

[0009] S1-4, if V opt1 If V1 is V2, V2 is discarded and a candidate voltage vector V3 is added. The phase angle selection range is [0, π]. The phase angles of the two candidate voltage vectors V1 and V3 are θ(V1) = Δθ / 2 = 2π / 2. n and θ(V3)=π-Δθ / 2=π-2π / 2 n ; If V opt1 If V2 is the candidate voltage vector, V1 is discarded and a candidate voltage vector V3 is added. The phase angle selection range is [π, 2π]. The phase angles of the two candidate voltage vectors V2 and V3 are θ(V2)=2π-Δθ / 2=2π-2π / 2 n and θ(V3)=π+Δθ / 2=π+2π / 2 n ; Calculate the cost function g respectively, select the candidate VV with the smaller cost function and record it as V opt2 ;

[0010] S1-5, iterative calculation according to the rule of S1-4, that is, the candidate VV phase angle of each round of comparison is the phase angle selection range ±Δθ / 2, and the phase angle selection range of this round of iteration is divided into two halves according to the angle bisector, and the half area with a smaller cost function is retained;

[0011] S1-6, after n-1 rounds of iteration, the phase angle selection range is reduced from the initial 2π to 2π / 2 n-1 , maximum phase angle error ε θ =π / 2 n-1 , the optimal phase angle of the candidate voltage vector is θ(V opt(n-1) );

[0012] S2-1. Fixed candidate VV V opt(n-1) The phase angle is θ(V opt(n-1) );

[0013] S2-2, determine the amplitude difference Δu=1 / 2 of the candidate VV according to the amplitude search times m m u m ;

[0014] S2-3, the amplitude selection range is [0, u m ], increase the candidate voltage vector V n+1 , two candidates VV V opt(n-1) and V n+1 The amplitudes are u(V opt(n-1) )=u m -1 / 2Δu=(1-1 / 2 m+1 ) m and u(V n+1 )=1 / 2Δu=1 / 2 m+1 u m , calculate the cost function g respectively, select the candidate VV with the smaller cost function and record it as V optn ; Among them, V opt(n-1) The cost function g(V opt(n-1) ) has been calculated and there is no need to repeat the calculation, so this step only calculates one cost function g(V n+1 );

[0015] S2-4, if V optn V opt(n-1) , then discard V n+1 , increase the candidate voltage vector V n+2 , the amplitude selection range is [1 / 2u m ,u m ], two candidates VV V opt(n-1) and V n+2 The amplitudes are u(V opt(n-1) )=u m -1 / 2Δu=(1-1 / 2 m+1 ) m and u(V n+2 )=1 / 2u m -1 / 2Δu=(1 / 2-1 / 2 m+1 ) m ; If V optn V n+1 , then discard V opt(n-1) , increase the candidate voltage vector V n+2 , the amplitude selection range is [0, 1 / 2u m ], two candidates VV V n+1 and V n+2 The amplitudes are u(V n+1 )=1 / 2Δu=1 / 2 m+1 u m and u(V n+2 )=1 / 2u m -1 / 2Δu=1 / 2u m -1 / 2 m+1 u m; Calculate the cost function g respectively, select the candidate VV with the smaller cost function and record it as V optn+2 ;

[0016] S2-5, iterative calculation is performed according to the rule of S2-4, that is, the candidate VV amplitude of each round of comparison is the amplitude selection range ±Δu / 2, and the amplitude selection range of this round of iteration is divided into two halves according to the bisector, and the half area with a smaller cost function is retained;

[0017] S2-6, after m rounds of iterations, the amplitude selection range is changed from the initial u m Reduce to 1 / 2 m u m , the maximum amplitude error ε u =1 / 2 m+1 u m , the optimal amplitude of the candidate voltage vector is u(V opt(n+m-1) ), thereby obtaining the optimal voltage vector.

[0018] Furthermore, the cost function g is expressed as:

[0019] g=[i 1αref -i 1α (k+1)] 2 +[i 1βref -i 1β (k+1)] 2

[0020] Among them, k is the current moment, k+1 is the next moment, and i 1αβ =[i 1α ,i 1β ] T is the primary current, i 1αβref =[i 1αref ,i 1βref ] T is the primary reference current of αβ axis.

[0021] The present invention also provides a model predictive control method based on discrete space vector modulation, comprising the following steps:

[0022] (1) Calculate the αβ axis reference current based on the sampled phase current and speed;

[0023] (2) Substituting the αβ-axis reference current and the sampled αβ-axis predicted current into the cost function, obtaining the optimal voltage vector according to the above-mentioned optimal voltage vector search method, and applying it to the converter at the next moment;

[0024] (3) Generate bridge arm switch pulse signals for each phase based on the optimal voltage vector and output voltage waveform.

[0025] The present invention also provides an electronic device, comprising: a computer-readable storage medium and a processor;

[0026] The computer-readable storage medium is used to store executable instructions;

[0027] The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the above method.

[0028] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to execute the above method.

[0029] The present invention also provides a computer program product, comprising a computer program or instructions, characterized in that the computer program or instructions implement the above method when executed by a processor.

[0030] Compared with the prior art, the above technical solutions conceived by the present invention can achieve the following

[0031] Beneficial effects:

[0032] (1) The present invention proposes a method for discretizing space vectors. is the radius of the circle, and the origin is the center of the circle. This method can discretize the entire space into 2 n+m-1 points, corresponding to 2 n+m-1 A selectable virtual voltage vector, which provides more vector options compared to the traditional discrete method;

[0033] (2) The present invention proposes a search method that searches for phase angle first and amplitude second. During the search iteration process, only one candidate voltage vector is added in each round. By comparing two of them, it is ensured that an accurate voltage vector can be obtained. Compared with traditional search methods, this method can obtain a more accurate voltage vector within a certain limit on the number of cost function calculations. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The present invention provides a linear induction motor model prediction control flow chart based on discrete space vector modulation.

[0035] Figure 2 Schematic diagram of a VV discrete space vector provided by an embodiment of the present invention.

[0036] Figure 3 This is a theoretical analysis of the optimal VV search method provided by an embodiment of the present invention.

[0037] Figure 4 These are the execution steps of the optimal VV search method provided by the embodiment of the present invention.

[0038] Figure 5 It is an overall control block diagram of a linear induction motor model predictive current control method based on discrete space vector modulation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0040] The present invention takes a linear induction motor as an example to explain in detail a linear induction motor model prediction control process based on discrete space vector modulation, such as Figure 1 As shown, the specific steps include:

[0041] According to the equivalent circuit model of the linear induction motor, a prediction model of the primary current and secondary flux is established.

[0042] Specifically, according to the classic T-type equivalent circuit model proposed by J. Duncan, the expressions of primary current and secondary flux in the αβ axis are:

[0043]

[0044] Among them, i 1αβ =[i 1α ,i 1β ] T is the primary current, ψ 2αβ =[ψ 1α ,ψ 1β ] T is the secondary magnetic flux, u 1αβ =[u 1α ,u 1β ] T is the primary voltage, L1=L m +L 1σ , L2=L m +L 2σ , ω2=v2π / τ, R1, R2, L1, L2, ω2, v2 and τ represent primary resistance, secondary resistance, primary inductance, secondary inductance, secondary angular frequency, speed and pole pitch respectively.

[0045] Based on formula (1), the prediction model of primary current and secondary flux can be expressed as:

[0046]

[0047] Among them, T s is the control cycle, (k) is the current moment, and (k+1) is the next moment.

[0048] The motor phase current and speed signal are sampled by the current sensor and speed sensor respectively, and the primary reference current of the αβ axis is further calculated.

[0049] Specifically, the q-axis primary reference current is obtained by subtracting the given speed from the sampling speed and performing PI proportional integration; the secondary flux amplitude and phase angle are obtained by observing the sampling speed and sampling current, and the d-axis primary reference current is obtained by subtracting the given secondary flux amplitude from the observed secondary flux amplitude and performing PI proportional integration; the dq-axis primary reference current is further transformed into the αβ-axis primary reference current through Park transformation.

[0050] The voltage vector with the smallest cost function is selected as the optimal voltage vector and applied to the converter at the next moment.

[0051] Specifically, after the predicted current and the reference current are obtained, the sum of the squares of the difference between the primary current reference value and the predicted value is defined as the cost function g.

[0052] g=[i 1αref -i 1α (k+1)] 2 +[i 1βref -i 1β (k+1)] 2 (3)

[0053] In the traditional two-level model predictive current control strategy, u in equation (2) 1αβ There are only 8 VVs to choose from, and the error between the actual optimal VV and the selected VV is large, which makes the current harmonic content larger, produces additional thrust fluctuations, and affects the control performance.

[0054] In order to increase the number of VVs available for selection, the present invention improves the model predictive current control and designs a two-level discrete space vector method, such as Figure 2 As shown in the figure, n is the number of cost function calculations when searching for phase angle, and m is the number of cost function calculations when searching for amplitude.

[0055] When n=2, m=1, the entire space is discretized into 4 points, the maximum phase angle error is π, and the maximum amplitude error is 1 / 2u m .

[0056] When n=3, m=2, the entire space is discretized into 16 points, the maximum phase angle error is π / 2, and the maximum amplitude error is 1 / 4u m .

[0057] When n=4, m=3, the entire space is discretized into 64 points, the maximum phase angle error is π / 4, and the maximum amplitude error is 1 / 8u m .

[0058] …

[0059] According to n and m, the entire space can be discretized into 2 n+m-1 points, corresponding to 2 n+m-1 selectable virtual voltage vectors (n≥2, m≥1), with a maximum phase angle error of ε θ =π / 2 n-1 , the maximum amplitude error is ε u =1 / 2 m+1 u m .

[0060] If all virtual VVs are traversed to find the optimal VV, the amount of calculation will be very huge. Therefore, an optimal voltage vector search method is designed. Figure 3 The principle of searching for the optimal VV in a discrete space vector plane is described.

[0061] Figure 3 In, V opt To find the optimal VV, if the search order of amplitude first and phase angle second is adopted, firstly, the phase angle is assumed to be fixed to 0, and |OB| is found as the amplitude with the minimum cost function; secondly, the amplitude is fixed to |OB|, V OD VV is found as the minimum cost function. Obviously, V OD and V opt There is an error between ε=V DA , this error occurs only when the assumed phase angle when searching for the amplitude is equal to V opt The phase angles are the same and are 0, but this is almost impossible.

[0062] On the contrary, if the search order is phase angle first and amplitude second, the amplitude is first assumed to be fixed at u m ,θ OE The phase angle with the minimum cost function is found; secondly, the phase angle is fixed to θ OE ,u OF The VV with the minimum cost function is found. Theoretically, when there are enough virtual VVs, the error ε = u FA can approach 0. Therefore, searching the phase angle first and then the amplitude is the way to find the optimal VV.

[0063] Figure 4 The execution steps of the optimal VV search method when n=4 and m=3 are given. The optimal VV is assumed to be V opt , the specific steps are:

[0064] 1. Search for the optimal VV phase angle, with a fixed VV amplitude of 15 / 16um , Δθ=π / 4;

[0065] S1, the phase angle selection range is [0, 2π], the phase angles of the two candidate VVs are θ(V1) = π / 8 and θ(V2) = -π / 8, respectively, calculate the cost functions g(V1) and g(V2), select the candidate VV with the smaller cost function and record it as V1 = V opt1 ;

[0066] S2, the phase angle selection range is [0, π], the phase angles of the two candidate VVs are θ(V1) = π / 8 and θ(V3) = π-π / 8, calculate the cost function g(V3), select the candidate VV with the smaller cost function and record it as V1 = V opt2 ;

[0067] S3, the phase angle selection range is [0, π / 2], the phase angles of the two candidate VVs are θ(V1) = π / 8 and θ(V4) = π / 2-π / 8, the cost function g(V4) is calculated, and the candidate VV with the smaller cost function is selected and recorded as V4 = V opt3 ;

[0068] 2. Search for the optimal VV amplitude, fix the VV phase angle to θ(V opt3 ), Δu=1 / 8u m ;

[0069] S4, amplitude selection range is [0, u m ], the amplitudes of the two candidate VV are u(V opt3 )=15 / 16u m and u(V5)=1 / 16u m , calculate the cost function g(V5), select the candidate VV with the smaller cost function and record it as V4=V opt4 .

[0070] S5, amplitude selection range is [1 / 2u m ,u m ], the amplitudes of the two candidate VV are u(V opt4 )=15 / 16u m and u(V6)=9 / 16u m , calculate the cost function g(V6), select the candidate VV with the smaller cost function and record it as V6=V opt5 .

[0071] S6, amplitude selection range is [1 / 2u m , 3 / 4u m ], the amplitudes of the two candidate VV are u(V opt5 )=9 / 16u m and u(V7)=11 / 16u m, calculate the cost function g(V7), select the candidate VV with the smaller cost function and record it as V7=V opt6 .

[0072] The optimal VV search method is designed based on the prediction model and cost function. After iterative search n+m-1 times, the number of cost function calculations is n+m, and 2 n+m-1 A selectable virtual voltage vector, the selected V optn+m-1 With V opt Compared with the conventional traversal search for the optimal VV method, the amount of calculation is greatly reduced.

[0073] Finally, the obtained optimal VV is sent to the PWM modulator to generate the corresponding PWM signal.

[0074] The overall control block diagram of the model predictive control method based on discrete space vector modulation provided by the present invention is as follows: Figure 5 shown.

[0075] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for searching an optimal voltage vector, characterized in that: The following steps are involved: S1-1, fixed voltage vector amplitude is (1-1 / 2 m+1 ) m ; Where m is the number of amplitude searches, m≥1, u m is the radius of the circular plane space of the converter output voltage vector; S1-2, determine the angle difference Δθ=2π / 2 according to the phase angle search times n n-1 ; Where n≥2; S1-3, the phase angle selection range is [0, 2π], and the phase angles of the two candidate voltage vectors V1 and V2 are θ(V1) = Δθ / 2 = 2π / 2 n and θ(V2)=2π-Δθ / 2=2π-2π / 2 n ; Calculate the cost function g respectively, select the candidate voltage vector with the smaller cost function and record it as V opt1 ; S1-4, if V opt1 If V1 is V2, V2 is discarded and a candidate voltage vector V3 is added. The phase angle selection range is [0, π]. The phase angles of the two candidate voltage vectors V1 and V3 are θ(V1) = Δθ / 2 = 2π / 2. n and θ(V3)=π-Δθ / 2=π-2π / 2 n ; If V opt1 If V2 is the candidate voltage vector, V1 is discarded and the candidate voltage vector V3 is added. The phase angle selection range is [π, 2π]. The phase angles of the two candidate voltage vectors V2 and V3 are θ(V2) = 2π-Δθ / 2 = 2π-2π / 2 n and θ(V3)=π+Δθ / 2=π+2π / 2 n ; Calculate the cost function g respectively, select the candidate voltage vector with the smaller cost function and record it as V opt2 ; S1-5, iterative calculation is performed according to the rule of S1-4, that is, the candidate voltage vector phase angle for each round of comparison is the phase angle selection range ±Δθ / 2, and the phase angle selection range of this round of iteration is divided into two halves according to the angle bisector, and the half area with a smaller cost function is retained; S1-6, after n-1 rounds of iteration, the phase angle selection range is reduced from the initial 2π to 2π / 2 n-1 , maximum phase angle error ε θ =π / 2 n-1 , the optimal phase angle of the candidate voltage vector is θ(V opt(n-1) ); S2-1. Fixed candidate voltage vector V opt(n-1) The phase angle is θ(V opt(n-1) ); S2-2, determine the amplitude difference Δu=1 / 2 of the candidate voltage vector according to the amplitude search times m m u m ; S2-3, the amplitude selection range is [0, u m ], increase the candidate voltage vector V n+1 , two candidate voltage vectors V opt(n-1) and V n+1 The amplitudes are u(V opt(n-1) )=u m -1 / 2Δu=(1-1 / 2 m+1 ) m and u(V n+1 )=1 / 2Δu=1 / 2 m+1 u m , calculate the cost function g respectively, select the candidate voltage vector with the smaller cost function and record it as V optn ; S2-4, if V optn V opt(n-1) , then discard V n+1 , increase the candidate voltage vector V n+2 , the amplitude selection range is [1 / 2u m ,u m ], two candidate voltage vectors V opt(n-1) and V n+2 The amplitudes are u(V opt(n-1) )=u m -1 / 2Δu=(1-1 / 2 m+1 ) m and u(V n+2 )=1 / 2u m -1 / 2Δu=(1 / 2-1 / 2 m+1 ) m ; If V optn V n+1 , then discard V opt(n-1) , increase the candidate voltage vector V n+2 , the amplitude selection range is [0, 1 / 2u m ], two candidate voltage vectors V n+1 and V n+2 The amplitudes are u(V n+1 )=1 / 2Δu=1 / 2 m+1 u m and u(V n+2 )=1 / 2u m -1 / 2Δu=1 / 2u m -1 / 2 m+1 u m ; Calculate the cost function g respectively, select the candidate voltage vector with the smaller cost function and record it as V optn+2 ; S2-5, iterative calculation is performed according to the rule of S2-4, that is, the amplitude of the candidate voltage vector for each round of comparison is the amplitude selection range ±Δu / 2, and the amplitude selection range of this round of iteration is divided into two halves according to the bisector, and the half area with a smaller cost function is retained; S2-6, after m rounds of iterations, the amplitude selection range is changed from the initial u m Reduce to 1 / 2 m u m , the maximum amplitude error ε u =1 / 2 m+ 1 u m , the optimal amplitude of the candidate voltage vector is u(V opt(n+m-1) ), thereby obtaining the optimal voltage vector.

2. The method for searching the optimal voltage vector according to claim 1, characterized in that: The cost function g is expressed as: g=[i 1αref -i 1α (k+1)] 2 +[i 1βref -i 1β (k+1)] 2 Among them, k is the current moment, k+1 is the next moment, and i 1αβ =[i 1α ,i 1β ] T is the primary current, i 1αβref =[i 1αref ,i 1βref ] T is the primary reference current of αβ axis.

3. The method for searching the optimal voltage vector according to claim 1, characterized in that: In step S2-3, V opt(n-1) The cost function g(V opt(n-1) ) is equal to the cost function g(V opt(n-1) ).

4. A model predictive control method based on discrete space vector modulation, characterized in that: The following steps are involved: (1) Calculate the αβ axis reference current based on the sampled phase current and speed; (2) Substituting the αβ-axis reference current and the sampled αβ-axis predicted current into the cost function, obtaining the optimal voltage vector according to the optimal voltage vector search method according to any one of claims 1 to 3, and applying it to the converter at the next moment; (3) Generate bridge arm switch pulse signals for each phase based on the optimal voltage vector and output voltage waveform.

5. An electronic device, characterized in that: include: A computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium and execute the method according to claim 4.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to cause a processor to execute the method according to claim 4.

7. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the method according to claim 4 is implemented.

Images