Fault-tolerant model predictive control method for virtual synchronous machine current sensor

Abstract

The present invention proposes a virtual synchronous machine current sensor fault-tolerant model predictive control method. In the current reconstruction, the number of phase currents that can be measured by the sub-vector is insufficient or the duration of the sub-vector is too short, which will cause an unmeasurable area and affect the effect of the fault-tolerant control strategy. In order to improve the reliability of the three-level VSG in the case of AC current sensor failure, the present invention designs a dual vector that is not affected by the unmeasurable area. When the selected optimal vector is a dual vector, the three-phase output current of the VSG is reconstructed according to the DC side current. In order to realize the current reconstruction of a single voltage vector, a prediction model of the output current of the LCL type VSG is established. When the selected optimal vector is a single voltage vector, the three-phase output current of the VSG is reconstructed according to the DC side current and the predicted current. The proposed fault-tolerant MPC strategy realizes the reconstruction of the output current when the VSG AC current sensor fails, improves the reliability of the VSG, and realizes its continuous and stable operation.

Description

Technical Field

[0001] The invention relates to the technical field of fault-tolerant model predictive control, and in particular to a virtual synchronous machine current sensor fault-tolerant model predictive control method. Background Art

[0002] With the development of renewable energy generation technology, power electronic equipment has been widely used in power systems. However, renewable energy has the characteristics of random fluctuations, and traditional grid-connected inverters lack inertia and damping, so too high a proportion of renewable energy generation is not conducive to the stable operation of the power grid.

[0003] By simulating the external characteristics of synchronous generators, virtual synchronous generators (VSGs) can provide inertia and damping for power grids, and have received extensive attention in recent years. However, due to the influence of environment, temperature and current overshoot, the failure of VSG AC current sensors will seriously affect its normal and stable operation. Therefore, it is urgent to study fault-tolerant control algorithms for VSG AC current sensor failures. Summary of the invention

[0004] In response to the problem of LCL type VSG AC current sensor failure, the present invention proposes a virtual synchronous machine current sensor fault tolerance model predictive control method, which realizes the reconstruction of the output current when the VSG AC current sensor fails, improves the reliability of the three-level VSG in the case of AC current sensor failure, and realizes its continuous and stable operation.

[0005] The technical solution of the present invention is achieved in this way:

[0006] A virtual synchronous machine current sensor fault tolerance model predictive control method, the steps of which are as follows:

[0007] Step 1: Construct an AC sensor fault criterion. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a dual vector, execute step 2. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a single voltage vector, execute step 3.

[0008] Step 2: According to the collected DC side current i dc Reconstruct VSG output current;

[0009] Step 3: Construct an LCL-type VSG prediction model based on the collected DC side current i dc and VSG prediction model to reconstruct VSG output current;

[0010] Step 4: Calculate the predicted value of the filter capacitor branch voltage based on the reconstructed VSG output current;

[0011] Step 5: Input reference active power and reference reactive power, and calculate reference voltage vector through VSG control strategy;

[0012] Step 6: Calculate the midpoint voltage prediction value, calculate the cost function value of the candidate vector according to the filter capacitor branch voltage prediction value, the reference voltage vector and the midpoint voltage prediction value; and use the vector corresponding to the minimum value of the cost function as the optimal vector;

[0013] Step 7: Output the switch signal corresponding to the optimal vector to control the VSG operation in the next cycle.

[0014] The construction method of the AC sensor fault criterion is:

[0015] According to the structure of VSG, the VSG three-phase output current and the DC side current i are obtained. dc Relationship:

[0016]

[0017] Among them, X a , X b , X c is the correlation coefficient of the three-phase output current of VSG, S a , S b , S c is the three-phase switch state of VSG, u() represents the step function;

[0018]

[0019] Among them, i 1a 、i 1b 、i 1c is the VSG three-phase output current;

[0020] Subtract the DC side current i from the right side of equation (2) dc And take the absolute value, get the DC current calculation error E idc for:

[0021] E idc =|(X a -X c )*i 1a +(X b -X c )*i 1b -i dc | (3);

[0022] Among them, || represents the absolute value;

[0023] Set the margin Y and establish the AC current sensor fault criterion F:

[0024]

[0025] In the step 2, the DC side current i dc The method to reconstruct the VSG output current is;

[0026] When the output vector is a double vector, the DC current i dc The sampling is performed twice under two single vectors, and the corresponding relationship between the sampling value and the VSG output current is shown in Table 1:

[0027] Table 1 i dc Corresponding relationship with VSG output current

[0028] <![CDATA[(X a ,X b ,X c )]]> <![CDATA[i dc Corresponding phase current value]]> (1,0,0) <![CDATA[i a ]]> (0,1,0) <![CDATA[i b ]]> (0,0,1) <![CDATA[i c ]]> (1,1,0) <![CDATA[-i c ]]> (1,0,1) <![CDATA[-i b ]]> (0,1,1) <![CDATA[-i a ]]>

[0029] According to Table 1, we can directly use i dc The two sampling values ​​of VSG two-phase output current are obtained; in addition, in a three-phase three-wire system, the remaining one-phase current value is the opposite value of the sum of the measured two-phase currents; therefore, according to i dc The two sampled values ​​reconstruct the VSG output current.

[0030] The LCL type VSG prediction model is:

[0031]

[0032] Among them, i 1αβ (k+1) is the αβ component of the VSG output current in the k+1 cycle, i 1αβ (k) is the αβ component of the VSG output current in the k cycle, i 2αβ (k) is the αβ component of the VSG grid current in cycle k, u Cαβ (k+1) is the αβ component of the filter capacitor voltage in the k+1 period, u Cαβ (k) is the αβ component of the filter capacitor voltage within the k cycle, u αβ (k) is the αβ component of the inverter output voltage in the k cycle, L 1 is the inverter side filter inductor, R 1 is the parasitic resistance on the inverter side, C is the filter capacitor, T is the sampling period, R 3 is the series resistance of the filter capacitor.

[0033] The DC side current i dc The method of reconstructing the VSG output current using the VSG prediction model is:

[0034]

[0035] Among them, V opt (k-1) is the optimal vector of k-1 period, u Cαβ(k-1) is the αβ component of the filter capacitor voltage within the k-1 period, i 1αβ (k-1) is the αβ component of the VSG output current in the k-1 cycle, i 2αβ (k-1) is the αβ component of the VSG grid current in the k-1 cycle.

[0036] The calculation method of the predicted value of the filter capacitor branch voltage is:

[0037]

[0038] Among them, u pccαβ (k+1) is the αβ component of the predicted value of the filter capacitor branch voltage in the k+1 period, u pccαβ (k) is the αβ component of the predicted value of the filter capacitor branch voltage in period k.

[0039] The method for calculating the reference voltage vector by the VSG control strategy is:

[0040]

[0041] Among them, P ref is the reference active power, Q ref is the reference reactive power, is the angular velocity, θ is the angle, M f is the maximum mutual inductance between the field winding and the stator coil, i f is the excitation current, i is the output current, u ref is the reference voltage.

[0042] The calculation method of the midpoint voltage prediction value is:

[0043]

[0044] Among them, i O (k) is the midpoint current of the k cycle, △U dc (k+1) is the midpoint voltage of the k+1 cycle, △U dc (k) is the midpoint voltage of the k cycle, t a is the time that the zero level of phase a occupies in one cycle, t b is the time that the zero level of phase b occupies in one cycle, t c is the time that the zero level of phase c occupies in one cycle, i 1a (k) is the a-phase output current of VSG in cycle k, i 1b (k) is the b-phase output current of VSG in k cycles, i 1c (k) is the c-phase output current of the VSG in cycle k.

[0045] The cost function is:

[0046]

[0047] Among them, g is the cost function, u refα represents the α-axis component of the VSG reference voltage, u refβ represents the β-axis component of the VSG reference voltage, u pccα (k+2) is the α-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, u pccβ (k+2) is the β-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, △U dc (k+2) is the midpoint voltage of the k+2 period, and λ is the weight coefficient of the midpoint voltage balance term.

[0048] In the fault-tolerant control of the AC-side current sensor of the three-level LCL type VSG, there is an unmeasurable area, which affects the reconstruction of the three-phase output current of the VSG. The present invention proposes a fault-tolerant MPC strategy based on current reconstruction, which eliminates the unmeasurable area and realizes the fault-tolerant operation of the VSG current sensor. Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1) The relationship between the DC side current and the VSG three-phase output current under different switching states was analyzed, and the fault judgment criteria of the VSG AC current sensor were designed to realize the online detection of AC current sensor faults.

[0050] 2) A double vector synthesized from two adjacent vectors was designed, and the three-phase output current of the VSG was reconstructed through the DC side current, thus realizing the current reconstruction of the double vector.

[0051] 3) The relationship between voltage and current in the three-level LCL type VSG is analyzed, and a voltage and current prediction model is established. The three-phase output current of the VSG is reconstructed by predicting the current and measuring the DC current, thus realizing the current reconstruction of a single vector.

[0052] 4) A fault-tolerant MPC strategy for the VSG AC side current sensor is designed to achieve current reconstruction in the entire domain; the proposed method realizes reference voltage tracking control and VSG fault-tolerant operation by predicting voltage and cost function, thereby improving the operation reliability and fault-tolerant capability of VSG. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 This is a diagram of the three-level VSG topology structure and the current sensor position structure of the present invention.

[0055] Figure 2 are candidate vectors of the fault-tolerant MPC method of the present invention; wherein (a) is a single vector and (b) is a double vector.

[0056] Figure 3 This is a control principle diagram of the fault-tolerant MPC method of the present invention.

[0057] Figure 4 Figure 2 is the VSG waveform diagram before and after the AC current sensor fails; (a) is the filter capacitor branch voltage waveform before and after the VSG AC current sensor fails under the conventional VSG control method, and (b) is the grid current waveform before and after the VSG AC current sensor fails under the conventional VSG control method.

[0058] Figure 5 : are the reconstructed current waveform and reconstruction error of the fault-tolerant MPC method of the present invention; wherein, (a) is the reconstructed current waveform, and (b) is the reconstruction error.

[0059] Figure 6 Figure 2 is the waveform before and after the method of the present invention is used when the VSG AC current sensor is in a fault state; (a) is the voltage waveform of the filter capacitor branch, and (b) is the grid-side current and FFT analysis results.

[0060] Figure 7 Figure 2 is the VSG waveform when the active power reference value suddenly changes; (a) is the active power waveform, (b) is the filter capacitor branch voltage waveform, and (c) is the grid-side current waveform.

[0061] Figure 8 Figure 2 is the VSG waveform when the reactive power reference value suddenly changes; (a) is the reactive power waveform, (b) is the filter capacitor branch voltage waveform, and (c) is the grid-side current waveform. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the 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.

[0063] The embodiment of the present invention provides a fault-tolerant model predictive control method for a virtual synchronous machine current sensor. First, a sensor fault criterion is designed to realize online detection of faults. Secondly, two single voltage vectors that can reconstruct currents of different phases are selected to synthesize a dual vector, thereby realizing current reconstruction in the first type of unmeasurable area. Then, an LCL-type VSG output current prediction model is established, and current reconstruction in the second type of unmeasurable area is realized by predicting the current with a single voltage vector. Finally, a fault-tolerant MPC strategy is designed to select the optimal vector output from the dual vector and the single voltage vector, thereby realizing fault-tolerant operation of the VSG current sensor and improving the reliability of the VSG. The specific steps are as follows:

[0064] Step 1: Construct an AC sensor fault criterion. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a dual vector, execute step 2. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a single voltage vector, execute step 3.

[0065] The topology of the three-level VSG is as follows Figure 1 As shown. By simulating the swing equation of the synchronous machine, the VSG has similar characteristics to the synchronous machine. Compared with the traditional grid-connected inverter, the VSG can provide inertia and damping for the grid, which is beneficial to the grid regulation and safe and stable operation.

[0066] exist Figure 1 In the VSG, there is a DC current sensor for overcurrent protection and two AC current sensors. In a three-phase three-wire system, at least two AC current sensors are required to obtain accurate VSG three-phase output current, and the remaining one-phase current is the inverse of the sum of the two-phase measured currents.

[0067] according to Figure 1 , we can get the VSG three-phase output current and DC current i dc The relationship is as follows

[0068]

[0069] Among them, X a , X b , X c is the correlation coefficient of the three-phase output current of VSG, S a , S b , S c is the three-phase switching state of VSG, and u() represents the step function.

[0070]

[0071] Among them, i 1a 、i 1b 、i 1c is the VSG three-phase output current.

[0072] According to equations (1) and (2), the relationship between the VSG switch state and the measured current and its corresponding phase is obtained, which is used to realize AC current sensor fault detection and current reconstruction.

[0073] When the AC current sensor fails, since the accurate fault phase current signal cannot be measured, there is a large error between the measured value of the fault phase and the actual value, resulting in the failure of equation (2). Therefore, it can be judged whether the AC current sensor is faulty according to equation (2).

[0074] Subtract i from the right side of equation (2) dc And take the absolute value, we can get the DC current calculation error E idc for:

[0075] E idc =|(X a -X c )*i 1a +(X b -X c )*i 1b -i dc | (3)

[0076] Here, || represents an absolute value.

[0077] Under ideal conditions, when the AC current sensor works normally, E idc =0; when the AC current sensor fails, E idc ≠0. In actual applications, due to sensor model, environment and other reasons, E idc It may be a non-zero small value, and a margin needs to be set. Therefore, the AC current sensor fault criterion F is designed as:

[0078]

[0079] Where Y is E idc The margin of the present invention is set to 1.

[0080] In formula (4), when E idc When it is less than Y, F=0, indicating that the AC current sensor works normally. idc When it is greater than Y, F=1, indicating that the AC current sensor is faulty and a fault tolerance strategy needs to be used.

[0081] Step 2: According to the collected DC side current i dc Reconstruct VSG output current;

[0082] When the output vector is a double vector, the DC current i dc The sampling is performed twice under two single vectors, and the corresponding relationship between the sampling value and the VSG output current is shown in Table 1:

[0083] Table 1 i dc Corresponding relationship with VSG output current

[0084] <![CDATA[(X a ,X b ,X c )]]> <![CDATA[i dc Corresponding phase current value]]> (1,0,0) <![CDATA[i a ]]> (0,1,0) <![CDATA[i b ]]> (0,0,1) <![CDATA[i c ]]> (1,1,0) <![CDATA[-i c ]]> (1,0,1) <![CDATA[-i b ]]> (0,1,1) <![CDATA[-i a ]]>

[0085] According to Table 1, we can directly use i dc The two sampling values ​​of VSG two-phase output current are obtained; in addition, in a three-phase three-wire system, the remaining one-phase current value is the opposite value of the sum of the measured two-phase currents; therefore, according to i dc The two sampled values ​​reconstruct the VSG output current.

[0086] Step 3: Construct an LCL-type VSG prediction model based on the collected DC side current i dc and VSG prediction model to reconstruct VSG output current;

[0087] According to formula (2), we can know Figure 2 In each 60-degree sector, different long vectors and short vectors correspond to different measured currents of the two phases. Therefore, the long vector and the short vector are used as sub-vectors of the double vector, and the design is greater than T min The duration of the sub-vector can be designed as Figure 2 (b) The double vector combination shown.

[0088] Figure 2 The candidate vectors of the proposed method are given, including 19 single vectors and 30 double vectors, respectively represented by V 0 -V 18 and V V1 -V V30 According to the volt-second balance principle, the dual vector can be obtained according to the duration of the sub-vector:

[0089]

[0090] Among them, T 1 , T 2 is the action time of the two sub-vectors, V 1αβ 、V 2αβ are the αβ components of the two sub-vectors, T s is the sampling period, V Vαβ are the αβ components of the double vector.

[0091] because Figure 2 The dual vectors are synthesized by two sub-vectors that can measure the current of different phases of VSG, and the action time of the sub-vectors is greater than T min , so the dual vector is not affected by the unmeasurable area. After the VSG AC current sensor fails, when the output vector is a dual vector, according to formula (2), the two-phase output current of the VSG is the DC side current i dcTwo sampling values ​​under two different vectors. The phase currents corresponding to the two sampling values ​​are determined according to Table 1. In addition, in a three-phase three-wire system, the sum of the three-phase output currents of the VSG is zero, and the remaining one-phase current value is the opposite value of the sum of the measured two-phase current values. Therefore, the current reconstruction of the dual vector is achieved.

[0092] However, when the output vector is a single vector, through i dc At most, the output current of one phase of VSG can be reconstructed, and the fault-tolerant operation after the AC current sensor failure cannot be achieved. Therefore, the present invention derives the LCL type VSG prediction model. dc And the predicted current reconstructs the VSG three-phase output current.

[0093] according to Figure 1 , the prediction model of LCL-type VSG in the αβ coordinate system can be derived as follows

[0094]

[0095] Among them, i 1αβ (k+1) is the αβ component of the VSG output current in the k+1 cycle, i 1αβ (k) is the αβ component of the VSG output current in the k cycle, i 2αβ (k) is the αβ component of the VSG grid current in cycle k, u Cαβ (k+1) is the αβ component of the filter capacitor voltage in the k+1 period, u Cαβ (k) is the αβ component of the filter capacitor voltage within the k cycle, u αβ (k) is the αβ component of the inverter output voltage in the k cycle, L 1 is the inverter side filter inductor, R 1 is the parasitic resistance on the inverter side, C is the filter capacitor, T is the sampling period, R 3 is the series resistance of the filter capacitor.

[0096] According to formula (6), the VSG output current reconstruction formula can be obtained:

[0097]

[0098] Among them, V opt (k-1) is the optimal vector of k-1 period, u Cαβ (k-1) is the αβ component of the filter capacitor voltage within the k-1 period, i 1αβ (k-1) is the αβ component of the VSG output current in the k-1 cycle, i 2αβ (k-1) is the αβ component of the VSG grid current in the k-1 cycle.

[0099] In formula (7), the optimal vector V selected according to the k-1 cycleopt (k-1) and other historical data to obtain the predicted VSG output current i 1αβ (k) i 1αβ (k) The three-phase predicted current can be obtained after Clark transformation, and the three-phase output current of VSG can be reconstructed to realize the current reconstruction of single voltage vector.

[0100] Step 4: Calculate the predicted value of the filter capacitor branch voltage based on the reconstructed VSG output current; To simulate the characteristics of the synchronous machine, the LCL type VSG filter capacitor branch voltage needs to track the reference voltage signal e of the VSG. According to formula (6), the predicted voltage of the filter capacitor branch is obtained:

[0101]

[0102] Among them, u pccαβ (k+1) is the αβ component of the predicted value of the filter capacitor branch voltage in the k+1 period, u pccαβ (k) is the αβ component of the predicted value of the filter capacitor branch voltage in period k.

[0103] Step 5: Input reference active power and reference reactive power, and calculate reference voltage through VSG control strategy;

[0104] The method for calculating the reference voltage through the VSG control strategy is:

[0105]

[0106] Among them, P ref is the reference active power, Q ref is the reference reactive power, is the angular velocity, θ is the angle, M f is the maximum mutual inductance between the field winding and the stator coil, i f is the excitation current, i is the output current, u ref is the reference voltage.

[0107] Step 6: Calculate the midpoint voltage prediction value, calculate the cost function value of the candidate vector according to the filter capacitor branch voltage prediction value, the reference voltage vector and the midpoint voltage prediction value; and use the vector corresponding to the minimum value of the cost function as the optimal vector;

[0108] The three-level VSG control strategy needs to suppress the DC side capacitor voltage difference. The proposed fault-tolerant MPC method achieves midpoint voltage balance by adding a midpoint voltage balance term to the cost function. The midpoint voltage prediction formula is as follows:

[0109]

[0110] Among them, i O(k) is the midpoint current of the k cycle, △U dc (k+1) is the midpoint voltage of the k+1 cycle, △U dc (k) is the midpoint voltage of the k cycle, t a is the time that the zero level of phase a occupies in one cycle, t b is the time that the zero level of phase b occupies in one cycle, t c is the time that the zero level of phase c occupies in one cycle, i 1a (k) is the a-phase output current of VSG in cycle k, i 1b (k) is the b-phase output current of VSG in k cycles, i 1c (k) is the c-phase output current of the VSG in cycle k.

[0111] Since there is a delay between sampling and control, the output vector used in the kth cycle is further predicted by (6) and (9) to obtain u pccαβ (k+2) and ΔU dc (k+2) to achieve delay compensation. After considering delay compensation, the design cost function g is

[0112]

[0113] Among them, g is the cost function, u refα represents the α-axis component of the VSG reference voltage, u refβ represents the β-axis component of the VSG reference voltage, u pccα (k+2) is the α-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, u pccβ (k+2) is the β-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, △U dc (k+2) is the midpoint voltage of the k+2 period, and λ is the weight coefficient of the midpoint voltage balance term.

[0114] Step 7: Output the switch signal corresponding to the optimal vector to control the VSG operation in the next cycle.

[0115] The control principle diagram of the method proposed by the present invention is as follows: Figure 3 shown.

[0116] After the two AC current sensors of the VSG fail, the VSG three-phase current is reconstructed first. opt When (k-1) is a double vector, the collected DC side current i is directly determined according to its sub-vector dc The corresponding two-phase currents are calculated based on the sum of the three-phase currents being 0 to obtain the reconstructed VSG three-phase output current i 1abc When V opt When (k-1) is a single voltage vector, according to V opt(k-1) The predicted current reconstruction corresponding to i 1abc After the current reconstruction is completed, the predicted value of the filter capacitor branch voltage is calculated by formula (8). In addition, the input set reference active power P ref and reference reactive power Q ref , and calculate the reference voltage vector u through the VSG control strategy refαβ Then, the predicted midpoint voltage is calculated according to (9), and the cost function value of the candidate vector is calculated by (10). The minimum cost function value g min The corresponding vector is the optimal vector V opt Finally, the output V opt The corresponding switching signal can realize continuous and stable operation of the VSG AC current sensor under fault.

[0117] In order to verify the effectiveness of the fault-tolerant MPC strategy proposed in this invention, the following experiments are designed: Figure 5-Figure 8 shown.

[0118] Figure 4 (a) and Figure 4 (b) shows the waveforms of the filter capacitor branch voltage and grid current before and after the VSG AC current sensor fails under the conventional VSG control method. It can be seen that compared with before the fault, u pccαβ and i 2αβ The waveform after the fault is obviously distorted. 2αβ The THD increased from 1.59% to 98.3%, seriously affecting the VSG control performance and grid-connected power quality.

[0119] After the VSG AC current sensor fails, the reconstructed current waveform and reconstruction error of the proposed method are as follows: Figure 5 shown. Figure 5 (a) shows the comparison between the actual three-phase current and the reconstructed current of the inverter. Figure 5 (b) shows the reconstruction error between the actual current and the reconstructed current. Figure 5 It can be seen that the three-phase reconstructed current waveform of the inverter is basically consistent with the actual current waveform, and the maximum absolute value of the difference between the actual current and the reconstructed current is less than 0.8 A, and the reconstruction error is less than 8%. Therefore, under the fault state of the VSG AC current sensor, the proposed fault-tolerant MPC method can accurately reconstruct the VSG three-phase current and be used for VSG fault-tolerant control.

[0120] Figure 6 The waveforms before and after the proposed method is used in the fault state of the VSG AC current sensor are given. Figure 6 (a) Figure 6 In (b), after using the proposed method, i 2αβ and u pccαβ They are close to the ideal sine waveform, and the harmonic components are significantly reduced. Figure 4 , Figure 6 It can be seen that after using the proposed method, i 2αβ and u pccαβ All of them can be restored to the normal operation state before the fault occurred, and the grid-connected current THD is 2.22%, which meets the grid-connected requirements. This shows that the proposed method can effectively realize the fault-tolerant control of VSG AC current sensor.

[0121] In order to verify the dynamic performance of the proposed fault-tolerant MPC method, Figure 7 , Figure 8 The active power reference value P is given respectively ref , reactive power reference value Q ref VSG waveform under mutation conditions.

[0122] like Figure 7 As shown, when P ref After the active power P suddenly changes from 2000W to 3000W, it can track P within 0.8s. ref A new stable state is reached. Figure 8 In, when Q ref After the mutation from 0Var to 1000Var, Q can track Q within 0.4s. ref A new stable state is reached. Figure 7 , Figure 8 It can be seen that when the power reference value changes suddenly, the proposed fault-tolerant MPC strategy can output the filter capacitor voltage and grid current corresponding to the reference power to achieve power tracking.

[0123] 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, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A virtual synchronous machine current sensor fault tolerance model predictive control method, It is characterized in that The steps are as follows: Step 1: Construct an AC sensor fault criterion. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a dual vector, execute step 2. When an AC sensor fault occurs and the optimal vector of the k-1th cycle is a single voltage vector, execute step 3. Step 2: According to the collected DC side current i dc Reconstruct VSG output current; Step 3: Construct an LCL-type VSG prediction model based on the collected DC side current i dc and VSG prediction model to reconstruct VSG output current; Step 4: Calculate the predicted value of the filter capacitor branch voltage based on the reconstructed VSG output current; Step 5: Input reference active power and reference reactive power, and calculate reference voltage vector through VSG control strategy; Step 6: Calculate the midpoint voltage prediction value, calculate the cost function value of the candidate vector according to the filter capacitor branch voltage prediction value, the reference voltage vector and the midpoint voltage prediction value; and use the vector corresponding to the minimum value of the cost function as the optimal vector; Step 7: Output the switch signal corresponding to the optimal vector to control the VSG operation in the next cycle.

2. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 1, It is characterized in that The construction method of the AC sensor fault criterion is: According to the structure of VSG, the VSG three-phase output current and the DC side current i are obtained. dc Relationship: Among them, X a , X b , X c is the correlation coefficient of the three-phase output current of VSG, S a , S b , S c is the three-phase switch state of VSG, u() represents the step function; Among them, i 1a 、i 1b 、i 1c is the VSG three-phase output current; Subtract the DC side current i from the right side of equation (2) dc And take the absolute value, get the DC current calculation error E idc for: E idc =|(X a -X c )*i 1a +(X b -X c )*i 1b -i dc | (3); Among them, || represents the absolute value; Set the margin Y and establish the AC current sensor fault criterion F:

3. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 1, It is characterized in that In the step 2, the DC side current i dc The method to reconstruct the VSG output current is; When the output vector is a double vector, the DC current i dc The sampling is performed twice under two single vectors, and the corresponding relationship between the sampling value and the VSG output current is shown in Table 1: Table 1 i dc Corresponding relationship with VSG output current According to Table 1, we can directly use i dc The two sampling values ​​of VSG two-phase output current are obtained; in addition, in a three-phase three-wire system, the remaining one-phase current value is the opposite value of the sum of the measured two-phase currents; therefore, according to i dc The two sampled values ​​reconstruct the VSG output current.

4. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 1, It is characterized in that The LCL type VSG prediction model is: Among them, i 1αβ (k+1) is the αβ component of the VSG output current in the k+1 cycle, i 1αβ (k) is the αβ component of the VSG output current in the k cycle, i 2αβ (k) is the αβ component of the VSG grid current in cycle k, u Cαβ (k+1) is the αβ component of the filter capacitor voltage in the k+1 period, u Cαβ (k) is the αβ component of the filter capacitor voltage within the k cycle, u αβ (k) is the αβ component of the inverter output voltage in the k cycle, L 1 is the inverter side filter inductor, R 1 is the parasitic resistance on the inverter side, C is the filter capacitor, T is the sampling period, R 3 is the series resistance of the filter capacitor.

5. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 4, It is characterized in that The DC side current i dc The method of reconstructing the VSG output current using the VSG prediction model is: Among them, V opt (k-1) is the optimal vector of k-1 period, u Cαβ (k-1) is the αβ component of the filter capacitor voltage within the k-1 period, i 1αβ (k-1) is the αβ component of the VSG output current in the k-1 cycle, i 2αβ (k-1) is the αβ component of the VSG grid current in the k-1 cycle.

6. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 5, It is characterized in that The calculation method of the predicted value of the filter capacitor branch voltage is: Among them, u pccαβ (k+1) is the αβ component of the predicted value of the filter capacitor branch voltage in the k+1 period, u pccαβ (k) is the αβ component of the predicted value of the filter capacitor branch voltage in period k.

7. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 1, It is characterized in that The method for calculating the reference voltage vector by the VSG control strategy is: Among them, P ref is the reference active power, Q ref is the reference reactive power, is the angular velocity, θ is the angle, M f is the maximum mutual inductance between the field winding and the stator coil, i f is the excitation current, i is the output current, u ref is the reference voltage.

8. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 6, It is characterized in that The calculation method of the midpoint voltage prediction value is: Among them, i O (k) is the midpoint current of the k cycle, △U dc (k+1) is the midpoint voltage of the k+1 cycle, △U dc (k) is the midpoint voltage of the k cycle, t a is the time that the zero level of phase a occupies in one cycle, t b is the time that the zero level of phase b occupies in one cycle, t c is the time that the zero level of phase c occupies in one cycle, i 1a (k) is the a-phase output current of VSG in k cycles, i 1b (k) is the b-phase output current of VSG in k cycles, i 1c (k) is the c-phase output current of the VSG in cycle k.

9. The virtual synchronous machine current sensor fault tolerance model predictive control method according to claim 1, It is characterized in that The cost function is: Among them, g is the cost function, u refα represents the α-axis component of the VSG reference voltage, u refβ represents the β-axis component of the VSG reference voltage, u pccα (k+2) is the α-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, u pccβ (k+2) is the β-axis component of the predicted value of the filter capacitor branch voltage in the k+2 period, △U dc (k+2) is the midpoint voltage of the k+2 period, and λ is the weight coefficient of the midpoint voltage balance term.

Images