CN110011359A - A Parameter Identification Method of Grid-connected Inverter Under Finite Set Model Predictive Control - Google Patents

Abstract

The invention discloses a grid-connected inverter parameter identification method under finite set model prediction control. The relationship between the real value and the nominal value of the parameter can identify the real inductance and resistance in real time, and build an accurate prediction model, thereby improving the quality of the grid-connected current and improving the robustness of the finite set model predictive control.

Description

A Parameter Identification Method of Grid-connected Inverter Under Finite Set Model Predictive Control

technical field

The invention belongs to the technical field of grid-connected inverter control, and more particularly relates to a parameter identification method under simplified model predictive control, which is used to improve the quality of the grid-connected current waveform of the inverter and the finite set model predictive control. Parameter robustness of down grid-connected inverters.

Background technique

Model Predictive Control (MPC) is a computer control algorithm that originated in the late 1970s. Its concept is intuitive, easy to model, and does not require accurate models and complex control parameter design. Problems such as nonlinearity and uncertainty have very good results, and it is easy to add constraints, fast dynamic response, and strong robustness.

In view of its huge advantages, FCS-MPC, one of MPCs, has been widely used and developed in power electronic converters, motor drives, power systems and other related fields in the past decade. There are many challenges, such as a large amount of online calculation, if the calculation times out, the switching action is not the optimal switching action at that moment, which reduces the effect of predictive control, and when there is a modeling error, it is necessary to follow the traditional FCS-MPC The optimal switching function combination selected by the algorithm has lost its optimality when it is implemented in the three-phase inverter, resulting in a deviation between the actual response curve of the system controlled variable and the optimal response curve determined by the FCS-MPC algorithm. affect the control performance of the system. However, the current identification methods for the mismatch of the model predictive control parameters of grid-connected inverters are all based on the traditional model predictive current control. The risk of timeout, and most of the current parameter identification schemes only identify the inductance, ignoring the resistance mismatch of the filter inductance, when the resistance fluctuates in a large range, the steady-state error of the grid-connected current will increase, and the current quality will decrease .

SUMMARY OF THE INVENTION

In order to avoid the above-mentioned shortcomings of the prior art, the present invention provides a parameter identification method of a grid-connected inverter under the predictive control of a finite set model, in order to identify the inductance and its parasitic resistance in real time by introducing two prediction voltage error equations , to obtain an accurate model and reduce the calculation amount of the predictive control, thereby improving the parameter robustness of the inverter model predictive control and improving the quality of the inverter grid-connected current waveform.

The present invention adopts the following technical scheme for solving the technical problem:

The characteristics of a method for identifying parameters of a grid-connected inverter under the predictive control of a finite set model of the present invention are that the method is carried out according to the following steps:

Step 1: Use equation (1) to construct the nominal discrete model of the simplified grid-connected inverter under the finite set model predictive control when the parameters are mismatched at time k:

In formula (1), is the optimal output voltage of the grid-connected inverter in the two-phase αβ static coordinate system at time k, e _{α, β} ^k is the grid voltage in the two-phase αβ static coordinate system at time k, r′ is the nominal resistance, and L′ is Nominal inductance; i _{α, β_ref} ^k+1 is the reference current in the two-phase αβ stationary coordinate system at time k+1, i _{α, β} ^k is the grid-connected current in the two-phase αβ stationary coordinate system at time k, Ts is the sampling time interval;

Use equation (2) to construct the real discrete model of the simplified grid-connected inverter under the finite set model predictive control when the actual parameters are running at time k:

In formula (2), V _{α, β} ^k is the actual output voltage of the grid-connected inverter in the two-phase αβ stationary coordinate system at time k, i _{α, β} ^k+1 is the two-phase αβ stationary coordinate system at the time k+1. The actual current, r is the actual resistance, L is the actual inductance;

The optimal output voltage of the grid-connected inverter in the two-phase αβ stationary coordinate system at time k-1 is obtained from equations (1) and (2) respectively and the actual output voltage V _α,β ^{k-1 of the grid-connected inverter in the actual two-phase αβ stationary coordinate system at time k-1} ;

Step 2: Use equation (3) to obtain the difference ΔV _α,β ^{k-1 between the optimal output voltage and the actual voltage in the two-phase αβ stationary coordinate system at time k-1} :

In formula (3), i _{α, β_ref} ^k is the reference current in the two-phase αβ stationary coordinate system at time k, i _{α, β} ^k-1 is the grid-connected current in the two-phase αβ stationary coordinate system at time k-1;

Using equation (4), the difference Δδ _α,β ^k-1 of the optimal voltage estimation error in the two-phase αβ stationary coordinate system at time k-1 and time k-2 is obtained:

In formula (4), i _{α, β_ref} ^k-1 are the reference currents in the two-phase αβ stationary coordinate system at time k-1, and i _{α, β} ^k-2 are the parallels in the two-phase αβ stationary coordinate system at the time k-2. network current;

The identification expression L _{α,β_est} ^k of the actual inductance L in the two-phase αβ stationary coordinate system at time k is constructed by formula (5):

Step 3: Use formula (6) to obtain the estimated formula R _{α,β_est} k of the actual resistance r in the two-phase αβ stationary coordinate system at time ^k :

Step 4: Use the α-axis inductance estimation formula shown in equation (7) to estimate the actual inductance L, and obtain the α-axis inductance estimated value L _{α_est} k at time ^k :

In formula (7), L _{α_est} ^k-1 is the estimated value of the α-axis inductance at time k-1, Δδ _α ^k-1 is the difference between the estimation errors of the optimal α-axis voltage at time k-1 and time k-2, i _αref ^k , i _αref ^k-1 are the α-axis reference currents at time k, time k-1, respectively, i _α ^k , i _α ^k-1 , i _α ^k-2 are time k, time k-1, k- The α-axis grid-connected current at time 2, V _α ^k-2 and V _α ^k-1 are the output voltages of the α-axis grid-connected inverter at time k-2 and time k-1, respectively;

Step 5: Use equation (8) to estimate the actual resistance r, and obtain the estimated resistance value R _est k at time ^k :

In formula (8), θ is the grid phase angle at time k-1, variable A=[1π/4,3π/4]∪[5π/4,7π/4], variable B=[0,1π/4] ∪[3π/4,5π/4]∪[7π/4～π];

Step 6: take the estimated value of the α-axis inductance L _{α_est} ^{k at the time k} as the nominal inductance L′, take the estimated resistance value R _est ^{k at the time k} as the nominal resistance r′ and substitute it into the formula (1), Get the nominal discrete model at time k;

Step 7: Substitute k with k+1 and enter into formula (1) to obtain a two-step prediction model;

Step 8: Obtain the optimal output voltage of the grid-connected inverter at time k+1 according to the two-step prediction model;

Step 9: According to the optimal output voltage, the value function optimization method is used to obtain the grid-connected inverter switch tube action signals S _a ^k+1 , S _b ^k+1 , S _c ^{k+1 at time k+1} , and output the switch tube action signals S _a ^k+1 , S _b ^k+1 , S _c ^k+1 of the grid-connected inverter after a delay of one cycle to realize the switching of the grid-connected inverter at time k+1 action;

Step 10: At time k+1, after assigning k+1 to k, return to step 1 for execution.

Compared with the prior art, the beneficial effects of the present invention are reflected in:

1. On the basis of the simplified finite set model predictive control, the present invention identifies the inductance and the parasitic resistance of the inductance by constructing a double voltage error equation, completely eliminating the parameter mismatch and improving the grid-connected current. It reduces the overall calculation amount of the finite set model predictive control based on parameter identification, thus reducing the risk of calculation timeout in one cycle.

2. The present invention performs parameter identification on the basis of the simplified finite set model predictive control model in step 1, which avoids the huge overall calculation amount caused by the parameter identification based on the traditional predictive current control, and greatly reduces the overall calculation It reduces the computational burden of the controller;

3. In the present invention, by constructing a dual voltage error equation in steps 2 and 3, the parasitic resistance of the inductance is identified while the inductance is identified, the parameter mismatch is completely eliminated, and the robustness of the finite set model predictive control is improved. Awesome.

4. In the present invention, the singularity problem of the estimated inductance and estimated resistance expressions in steps 2 and 3 is analyzed in detail in steps 4 and 5, and a simple and effective solution is proposed, which avoids the problems in previous studies. The complexity of the analysis of the parameter-estimated singularity problem.

Description of drawings

Fig. 1 is the structural block diagram of the grid-connected inverter system in the present invention;

Fig. 2 is the principle diagram of singularity processing in the inductance estimation of the present invention;

Fig. 3 is the principle diagram of singularity processing in the resistance estimation of the present invention;

Fig. 4 is the effect diagram of parameter estimation before and after adding parameter identification according to the present invention;

Fig. 5 is the effect diagram of the grid-connected current before and after adding parameter estimation according to the present invention;

Fig. 6 is the FFT analysis diagram of the grid-connected current before adding parameter estimation according to the present invention;

FIG. 7 is an FFT analysis diagram of the grid-connected current after parameter estimation is added in the present invention.

Detailed ways

Referring to FIG. 1 , in this embodiment, a method for parameter identification of grid-connected inverters under simplified finite set model predictive control is to construct a simplified model predictive control mismatch parameter prediction model and an actual model based on accurate parameters, According to the above model construction, by constructing a dual voltage error equation, the real inductance and resistance are identified in real time, an accurate prediction model is constructed, the optimal output voltage is calculated, two-step prediction, and finally the optimal inverter is obtained through the optimization of the value function. switch tube action. Specifically, follow the steps below:

Step 1: Use equation (1) to construct a nominal discrete model based on calculating the optimal voltage of the grid-connected inverter under the finite set model predictive control when the parameters are mismatched at time k:

In formula (1), is the optimal output voltage of the grid-connected inverter in the two-phase αβ static coordinate system at time k, e _{α, β} ^k is the grid voltage in the two-phase αβ static coordinate system at time k, r′ is the nominal resistance, and L′ is Nominal inductance; i _{α, β_ref} ^k+1 is the reference current in the two-phase αβ stationary coordinate system at time k+1, i _{α, β} ^k is the grid-connected current in the two-phase αβ stationary coordinate system at time k, Ts is the sampling time interval;

Using Equation (2) to construct a real discrete model based on calculating the optimal voltage of the grid-connected inverter under the finite set model predictive control when the actual parameters are running at time k:

In formula (2), V _{α, β} ^k is the actual output voltage of the grid-connected inverter in the two-phase αβ stationary coordinate system at time k, i _{α, β} ^k+1 is the two-phase αβ stationary coordinate system at the time k+1. The actual current, r is the actual resistance, L is the actual inductance;

The optimal output voltage of the grid-connected inverter in the two-phase αβ stationary coordinate system calculated at time k-1 is obtained from equations (1) and (2) respectively And the actual output voltage V _{α, β} k-1 of the grid-connected inverter in the actual two-phase αβ stationary coordinate system at the time ^k-1 , but the actual output voltage V _{α, β} ^k-1 can be determined by the inverse of the time k-1. The switching action of the inverter and the DC voltage of the inverter are directly obtained; the prediction model based on the optimal voltage calculation only needs one calculation of the optimal voltage, which is greatly reduced compared to the 8 calculations of the predicted current of the traditional predicted current model. Therefore, the parameter identification of the simplified model predictive control based on calculating the optimal voltage can greatly reduce the overall calculation amount.

Step 2: Calculate the optimal output voltage of the inverter calculated in Step 1 Difference with the actual output voltage V _{α, β} ^k-1 : Using equation (3), the difference between the optimal output voltage and the actual voltage in the two-phase αβ stationary coordinate system at time k-1 is obtained:

In formula (3), i _{α, β_ref} ^k is the reference current in the two-phase αβ stationary coordinate system at time k, i _{α, β} ^k-1 is the grid-connected current in the two-phase αβ stationary coordinate system at time k-1;

Further construct the difference between the optimal voltage estimation errors before and after the time: Δδ _α,β ^k-1 =ΔV _α,β ^k-1 -ΔV _α,β ^k-2 , use formula (4) to obtain the time k-1 and k-2 The difference between the optimal voltage estimation errors in the two-phase αβ stationary coordinate system at time:

In formula (4), i _{α, β_ref} ^k-1 are the reference currents in the two-phase αβ stationary coordinate system at time k-1, and i _{α, β} ^k-2 are the parallels in the two-phase αβ stationary coordinate system at the time k-2. network current;

In formula (4), there are generally:

Therefore, ignoring the resistance term in equation (4), the identification expression L _{α,β_est} ^k of the actual inductance L in the two-phase αβ stationary coordinate system at time k shown in equation (6) is obtained:

Step 3: Substitute the formula (6) in step 2 into the voltage error expression (3), and obtain the actual resistance r and the nominal resistance r in the two-phase αβ static coordinate system at the time k-1 as shown in the formula (7). The relational formula J _α,β of ', as shown in formula (7):

The estimated formula R _{α,β_est} ^k of the actual resistance r in the two-phase αβ stationary coordinate system at time k is obtained from equation (7):

Step 4: When calculating the inductance calculation formula (6) calculated in step 2, if the numerator term is 0, the estimated value has a peak value, which is regarded as a singular point. To obtain a stable estimated value, the singular point must be avoided. It is difficult to directly analyze the singularity of the inductor. According to the inverter switching vector relationship in Figure 2, the voltage vector relationship shown in equation (9) can be obtained:

In formula (9), V _{α, β} ^k-2 , V _{α, β} ^k-1 are the inverter output voltages in the two-phase αβ stationary coordinate system at time k-2, k-1, respectively, V _{grid_α, β} ^{k -2} , V _{grid_α, β} ^k-1 are the grid voltages in the two-phase αβ stationary coordinate system at time k-2, k-1, respectively.

When the sampling frequency is large enough, the magnitudes of the grid voltages at two adjacent moments are approximately equal, and equation (9) can be transformed into:

From the formula (10), the form of the singularity in the formula (6) in step 2 can be converted into the difference between the two switching actions before and after, which greatly simplifies the judgment process of the singularity. From the inverter switch vector diagram 2, it can be seen that the singularity of the α-axis inductance estimation formula only exists when the switching actions of the inverter are identical at two adjacent moments, and the singularity of the β-axis inductance estimation formula is not only in the above cases. Exist, when the inverter switching action is between 000 and 100, between 001 and 101, between 010 and 110, and between 111 and 011, it also exists, the number of singular points is more than the α-axis inductance estimation formula, so the α-axis The response speed of the inductance estimation is faster than the β-axis inductance estimation formula, so the α-axis inductance estimation formula is used to estimate the actual inductance.

Use the α-axis inductance estimation formula shown in equation (11) to estimate the actual inductance L, and obtain the α-axis inductance estimated value L _{α_est} k at time ^k :

In formula (11), L _{α_est} ^k-1 is the estimated value of the α-axis inductance at time k-1, Δδ _α ^k-1 is the difference between the estimation errors of the optimal α-axis voltage at time k-1 and time k-2, i _αref ^k , i _αref ^k-1 are the α-axis reference currents at time k, time k-1, respectively, i _α ^k , i _α ^k-1 , i _α ^k-2 are time k, time k-1, k- The α-axis grid-connected current at time 2, V _α ^k-2 and V _α ^k-1 are the output voltages of the α-axis grid-connected inverter at time k-2 and time k-1, respectively;

Step 5: Step 3 The resistance estimation formula (8) also has a singularity, but the singularity is simple in form, that is, the zero-crossing point of the two-axis current, and the current phase of the two axes in the two-phase αβ static coordinate system differs by 90°. The singularities of the resistance estimation formula are also out of phase by 90°. By switching the estimation axis, it is easy to avoid the estimated singularity of each axis, that is, within a certain range of the singularity of the α-axis estimation formula, use the β-axis resistance estimation formula to estimate, stop the α-axis estimation, and reach the β-axis singularity Within a certain range, use the α-axis resistance estimation formula to estimate, and stop the β-axis estimation. However, the selection of the range near the singularity of the two axes will affect the effect of the estimation. In theory, the farther away from the singularity, the smaller the shock of the estimated value, and the better the estimation effect.

Step 3 In formula (7), r' is the actual applied value of the resistance in the prediction model at time k-1, and the estimated value of r is the applied value at time k. When the sampling frequency is high enough, it can be considered that the current before and after the current The amplitudes at the time are equal, and the formula (7) can be approximated as the voltage on the resistor at time k-1 minus the voltage on the resistor at time k, to obtain (12):

J _α,β ≈V _r ^k-1 -V _r ^k (12)

In formula (14), V _r ^k-1 and V _r ^k are the voltages on the resistor r at time k-1 and k, respectively, and both sides are divided by the sampling period Ts to obtain formula (13):

J _α,β /Ts≈(V _r ^k-1 -V _r ^k )/Ts=-ΔV _r /Ts (13)

In Equation (13), ΔV _r is the difference between the voltages on the resistor r at time k and k-1.

It can be seen that J _{α, β} lag V _r ^k-1 90°, and the voltage on the resistor r is in phase with the current, so J _{α, β} also lags i _{α, β} ^k-1 90°, then J _α /i _α ^{k- 1} can be expressed as (14):

J _α /i _α ^k-1 =J _{_α} sin(θ-90)/I _{_α} ×sin(θ)=J _{_α} /I _{_α} tan(θ-90) (14)

In formula (14), J _α is the α-axis component of the voltage difference at the adjacent time on the resistor r, i _α ^k-1 is the α-axis component of the grid-connected current at the time k-1, and _{J_α} is the voltage on the resistance at the adjacent time. The difference amplitude α-axis component, _{I_α} is the amplitude of the α-axis component of the grid-connected current, and θ is the grid phase angle at time k-1.

Similarly, J _β /i _β ^k-1 can be expressed as formula (15):

J _β /i _β ^k-1 ==J _{_β} /I _{_β} tanθ (15)

In formula (15), J _β is the β-axis component of the voltage difference between adjacent moments on the resistor r, i _β ^k-1 is the β-axis component of the grid-connected current at the time k-1, and J _{_β} is the voltage on the resistor at the adjacent moment. Difference β-axis component amplitude, _{I_β} is the amplitude of the β-axis component of grid-connected current.

From the graphs of equations (14) and (15) in Figure 3, it can be seen that near π/2, 3π/2, the estimated value of the α-axis resistance has a smaller oscillation, while in the vicinity of π and 2π, the estimated value of the β-axis resistance has a larger oscillation. Therefore, the constant amplitude oscillation point is taken as the critical point of the two-axis identification and conversion. In general, the estimation of the inductance is relatively stable.

Therefore, Equation (14) is equal to Equation (15), and Equation (16) is obtained:

|J _{_α} /I _{_α} tan(θ-90)|=|J _{_β} /I _{_β} tanθ| (16)

The critical angle of the two-axis resistance identification transformation in the two-phase αβ stationary coordinate system is obtained as: π/4, 3π/4, 5π/4, 7π/4, and the actual resistance r is estimated by using the formula (17), and the time k is obtained. The resistance estimate R _est ^k of :

In formula (17), θ is the grid phase angle at time k-1, variable A=[1π/4,3π/4]∪[5π/4,7π/4], variable B=[0,1π/4] ∪[3π/4,5π/4]∪[7π/4～π]; the resistance estimate obtained by this method has the smallest overall oscillation amplitude and periodic oscillation, so moving average filtering can be used to obtain a smooth resistance estimate curve.

Step 6: Take the estimated value of the α-axis inductance L _{α_est} ^{k at time k} as the nominal inductance L′, and take the estimated resistance value R _est ^{k at time k} as the nominal resistance r′ and substitute it into Equation (1) to obtain the Nominal discrete model;

Step 7: Substitute k with k+1 and enter into formula (1) to obtain a two-step prediction model;

Step 8: Obtain the optimal output voltage of the grid-connected inverter at time k+1 according to the two-step prediction model;

Step 9: According to the optimal output voltage, the value function optimization method is used to obtain the switch tube action signals S _a ^k+1 , S _b ^k+1 , S _c ^{k+1 of the grid-connected inverter at the time k+1} , and in After a period of delay, output the switch tube action signals S _a ^k+1 , S _b ^k+1 , S _c ^k+1 of the grid-connected inverter to realize the switching action of the grid-connected inverter at time k+1;

Step 10: At time k+1, after assigning k+1 to k, return to step 1 for execution.

In order to verify the validity of the parameter identification method proposed by the present invention, a three-phase grid-connected inverter model with a rated capacity of 18kw is built in Matlab/simulink. The nominal values of the inductance and resistance are 1mH and 0.5Ω respectively. The actual inductance and resistance is 5mH, 0.3Ω. Figure 4 shows the effect of inductance and resistance identification before and after adding parameter identification. The parameter identification algorithm is added at 0.5s. The nominal value before 0.5s and the actual inductance value after 0.5s are identified. As can be seen from Figure 4, identifying the inductance and resistance can quickly and accurately track the actual value of the inductance resistance. Figure 5 shows the grid-connected current waveform before and after adding parameter identification. The parameter identification algorithm is added at 0.1s. It can be seen that the three-phase grid-connected current is seriously distorted before 0.1s, and the current waveform is greatly improved after 0.1s. Figure 6 is the FFT analysis result of the grid-connected current before adding the parameter identification. The grid-connected current amplitude obviously deviates from the current command value of 40A, the current waveform is seriously distorted, and the THD value is too high. Figure 7 shows the result of FFT analysis of the grid-connected current after adding parameter identification. The deviation of the amplitude of the grid-connected current is significantly improved, the current waveform is sinusoidal, and the THD value is greatly reduced. The parameter identification based on the simplified model predictive control proposed by the present invention can well improve the grid-connected current amplitude deviation and current distortion caused by parameter mismatch, and greatly improve the robustness of the finite set model predictive control.

Claims (1)

1. the gird-connected inverter parameter identification method under a kind of finite aggregate Model Predictive Control, characterized in that the method be by What following steps carried out:

Step 1: when using formula (1) building k moment parameter mismatch, simplified gird-connected inverter under finite aggregate Model Predictive Control Nominal discrete model:

In formula (1), V_α,β ^k*It is the optimal output voltage of gird-connected inverter under k moment two-phase α β rest frame, e_α,β ^kFor the k moment Network voltage under two-phase α β rest frame, r ' are nominal resistance, and L ' is nominal inductance；i_{α,β_ref} ^k+1For k+1 moment two-phase α Reference current under β rest frame, i_α,β ^kFor the grid-connected current under k moment two-phase α β rest frame, Ts is the sampling time Interval；

When being run using formula (2) building k moment actual parameter, simplified gird-connected inverter is true under finite aggregate Model Predictive Control Real discrete model:

In formula (2), V_α,β ^kIt is the actual output voltage of gird-connected inverter under k moment two-phase α β rest frame, i_α,β ^k+1When for k+1 The actual current under two-phase α β rest frame is carved, r is actual resistance, and L is actual inductance；

Obtain the optimal output voltage of gird-connected inverter under k-1 moment two-phase α β rest frame respectively by formula (1) and formula (2) V_α,β ^k-1*And under k-1 moment actual two-phase α β rest frame gird-connected inverter actual output voltage V_α,β ^k-1；

Step 2: obtaining the difference Δ of optimal output voltage and virtual voltage under k-1 moment two-phase α β rest frame using formula (3) V_α,β ^k-1:

In formula (3), i_α,β__ref ^kFor the reference current under k moment two-phase α β rest frame, i_α,β ^k-1It is quiet for k-1 moment two-phase α β The only grid-connected current under coordinate system；

The difference Δ of the optimal voltage evaluated error under k-1 moment and k-2 moment two-phase α β rest frame is obtained using formula (4) δ_α,β ^k-1:

In formula (4), i_{α,β_ref} ^k-1For the reference current under k-1 moment two-phase α β rest frame, i_α,β ^k-2For k-2 moment two-phase α β Grid-connected current under rest frame；

The identification expression formula L of actual inductance L under k moment two-phase α β rest frame is constructed using formula (5)_{α,β_est} ^k:

Step 3: obtaining the estimator R of actual resistance r under k moment two-phase α β rest frame using formula (6)_{α,β_est} ^k:

Step 4: actual inductance L being estimated using α axle inductance estimator shown in formula (7), obtains the α axle inductance at k moment Estimated value L_{α_est} ^k:

In formula (7), L_{α_est} ^k-1For the α axle inductance estimated value at k-1 moment, Δ δ_α ^k-1It is optimal for k-1 moment and the α axis at k-2 moment The difference of voltage evaluated error, i_αref ^k, i_αref ^k-1Respectively k moment, the α axis reference current at k-1 moment, i_α ^k, i_α ^k-1, i_α ^k-2Respectively For k moment, k-1 moment, the α axis grid-connected current at k-2 moment, V_α ^k-2, V_α ^k-1The α axis at respectively k-2 moment, k-1 moment is grid-connected inverse Become the output voltage of device；

Step 5: actual resistance r being estimated using formula (8), obtains the resistance estimated value R at k moment_est ^k:

In formula (8), θ is the power grid phase angle at k-1 moment, variables A=[1 π/4,3 π/4] ∪ [5 π/4,7 π/4], variable B=[0,1 π/4] ∪ [π/4 3 π/4,5] ∪ [7 π/4~π]；

Step 6: by the α axle inductance estimated value L at the k moment_{α_est} ^kAs nominal inductance L ', the resistance at the k moment is estimated Value R_est ^kIt as nominal resistance r ' and substitutes into formula (1), obtains the nominal discrete model at k moment；

Step 7: being substituted into formula (1) after k is replaced with k+1, obtain two-staged prediction model；

Step 8: the optimal output voltage of k+1 moment gird-connected inverter is obtained according to the two-staged prediction model；

Step 9: according to the optimal output voltage, utility value function optimization method obtains obtaining the parallel network reverse at k+1 moment Device switching tube action signal S_a ^k+1,S_b ^k+1,S_c ^k+1, and gird-connected inverter switching tube action signal is exported after a cycle that is delayed S_a ^k+1,S_b ^k+1,S_c ^k+1, realize to gird-connected inverter the k+1 moment switch motion；

Step 10: at the k+1 moment, after k+1 is assigned to k, return step 1 is executed.