A control method of energy storage converter of a neutral point clamped three-level inverter

Abstract

The application provides a control method for a midpoint clamping type three-level inverter energy storage converter, which comprises an inverter control mode based on model prediction, a rectifier control mode based on model prediction and a reactive power compensation control mode based on model prediction; the energy storage converter based on the midpoint clamping type converter determines a specific working mode according to the electric quantity state of an energy storage module, a current period and the voltage of the energy storage module, wherein the inverter control mode obtains a reference current through given power calculation I* , and predicts a limited set of switch states; the rectifier control mode obtains a reference current through PI control I* ; the reactive power compensation control mode obtains a reference current through load current calculation I* ; finally, the optimal switch amount is selected by introducing a value function, and the optimal gating pulse of the energy storage converter is given; compared with the prior art, the control simplicity of the energy storage converter is improved by introducing the model prediction control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy storage converter PCS, and particularly relates to an energy storage converter control method of a neutral point clamped three-level inverter. BACKGROUND

[0002] The power conversion system (PCS) as the core of the energy storage system realizes the energy transmission between the energy storage device and the power grid. The PCS can control the bidirectional flow of energy between the alternating current power grid and the energy storage device through the rectification function and the inverter function of AC / DC. The conventional PCS control method often adopts PI control, so it is necessary to set corresponding PI parameters for different application environments, which adds the problem of design complexity, resulting in that the energy storage device usually cannot have the function of reactive power compensation in use. Moreover, the conventional PI control method has poor followability, and has the problem of slow response, which cannot make the inverter quickly output the required output value.

[0003] The inverter topology can select a neutral point clamped (NPC) three-level inverter, which has the advantages of half DC bus voltage stress of the power tube, less equivalent capacitance than other three-level inverter topologies, strong theoretical analysis uniformity, and can be directly extended to more level inverter topologies, etc. The model prediction control (MPC) is introduced to achieve the purposes of optimizing algorithm design, saving parameter setting process and improving dynamic performance, so as to solve the problems of slow response, complex algorithm, more links, complex parameter setting, etc. caused by the conventional PCS control method. SUMMARY

[0004] The present application relates to the field of energy storage converter PCS, and particularly relates to an energy storage converter control method of a neutral point clamped three-level inverter.

[0005] The present application relates to the field of energy storage converter PCS, and particularly relates to an energy storage converter control method of a neutral point clamped three-level inverter.

[0006] The model prediction-based inversion control mode is an inversion mode, the energy storage is discharged, the reference current I* is calculated through the given power and is brought into the value function, the optimal switch state under the limited set is predicted, and the optimal gating pulse of the energy storage converter is given;

[0007] The model prediction-based rectification control mode is a rectification mode, the energy storage is charged, the reference current I* is obtained through PI control, the optimal switch value is selected by bringing the reference current I* into the value function, and the optimal gating pulse of the energy storage converter is selected based on the optimal gating pulse of the energy storage converter.

[0008] The model prediction-based reactive power compensation control mode is a reactive power compensation mode, compensates for the reactive current generated by the load in the same system, calculates the reference current I* through the load current, selects the optimal switch value by bringing the reference current I* into the value function, and gives the optimal gating pulse of the energy storage converter.

[0009] Further, the specific working mode is determined according to the state of charge of the energy storage module, the current period and the voltage of the energy storage module, and is specifically as follows:

[0010] Step 2.1: Determine whether it is in a non-electricity low valley period or a set non-energy storage power supply period;

[0011] Step 2.2: If it is determined to be no, determine whether the voltage of the PCS energy storage device is less than the set value U1;

[0012] Step 2.3: If it is determined to be yes in step 2.2, the rectification mode is adopted to charge the energy storage device;

[0013] Step 2.4: Detect the state of charge SOC of the energy storage module, if greater than or equal to 95%, stop charging, otherwise continue charging;

[0014] Step 2.5: If it is determined to be no in step 2.2, the reactive power compensation mode is adopted to improve the power factor of the power grid;

[0015] Step 2.6: If it is determined to be yes in step 2.1, determine whether the voltage of the PCS energy storage device is less than the set value U2, U2 should be set to be greater than U1;

[0016] Step 2.7: If it is determined to be yes in step 2.6, the reactive power compensation mode is adopted to improve the power factor of the power grid;

[0017] Step 2.8: If it is determined to be no in step 2.6, the inversion mode is adopted to discharge the power grid at a constant power.

[0018] Further, the value function is used to select the optimal switch value, through the two-step time delay compensation optimization algorithm, the switch value at k+2 time is selected as the switch state when the minimum value of the value function is selected, and is specifically as follows:

[0019] The mathematical model of the three-phase voltage type inverter is shown in the following formula (1):

[0020]

[0021] In the formula, L is the filter inductance in series with the inverter, I a , I b , I c are the a, b, c phase currents passing through the filter inductance, u ca , u cb , u cc are the output voltages of the filter capacitors of each phase, V a , V b , V c are the voltages output by each phase of the inverter, and R is the line impedance.

[0022] The αβ axis current voltage equation is as follows:

[0023]

[0024] In the formula, the subscript αβ variable is the value of the corresponding variable after αβ coordinate transformation, and the specific transformation method is shown in the following formulas (3) and (4):

[0025]

[0026] After discretization of formula (2), it is shown in the following formula (5):

[0027]

[0028] According to Kirchhoff's current law, the current i o flowing out of the neutral point of the NPC three-level inverter DC side can be written as:

[0029] i o =i c1 -i c2 (6)

[0030] Since the energy storage voltage V dc is changing, the 27 vectors u n will change dynamically according to the sampling voltage value of the energy storage side at time k, and the amplitude is 0, ±V dc / 3, ±2V dc / 3; when the specific switching state S n is selected according to the principle of NPC grid-connected inverter, the inverter outputs the corresponding 27 voltage vectors u n , the current i c1 , i c2 on the DC side capacitor is expressed by the following formula:

[0031]

[0032] I a +I b +I c =0 (10)

[0033] where, V c1 , V c2 is the DC side capacitor voltage, I a , I b , I c is the a, b, c phase current through the filter inductance; the above formula (7) ~ (10) into formula (6), the expression of midpoint voltage and three-phase current is:

[0034]

[0035] where ΔV c is V c1 -V c2 , denoted as the midpoint voltage;

[0036] The power frequency AC is selected as the grid-connected source, and the grid voltage period is much larger than the sampling period, so formula (11) is discretized by forward difference method:

[0037]

[0038] By sampling the midpoint voltage value ΔV c (k), the switch state S n and the three-phase current I abc , the next moment of midpoint voltage value ΔV c (k+1) is analyzed and predicted;

[0039] The value function selects the output current as the controlled object of the target function, which is specifically expressed as shown in formula (13), and the optimization algorithm of two-step delay compensation adopts formula (14):

[0040] g i1 = |I * (k+1)-I(k+1)| (13)

[0041] g i2 = |I * (k+2)-I(k+2)| (14)

[0042] Where I*(k+1) is the given output current at k+1 moment, and formula (13) is selected as the constraint term of current to join the value function;

[0043] Another control target of the target function aims to keep the DC side capacitor voltage difference ΔV cThe control constant is always 0, and the constraint term of the midpoint voltage is added to the target function by a two-step delay compensation, as shown in the following formula (15):

[0044]

[0045] Combining formula (14) with formula (15), the final target function used is:

[0046] g 2min = |I * (k+2)-I(k+2)|+λ|ΔV c (k+2)| (16)

[0047] The value function is designed to use the voltage V * (k) required to reach the given current I * (k+1) at the given time, by selecting the optimal V * (k) so that the output current I(k+1) is close to the given current I * (k+1) at the same time, by formula (5), replacing I(k+1) with I * (k+1), and replacing V(k) with V * (k), rearranging to the following formula:

[0048]

[0049] Formula (17) shows that the closer the output voltage V(k) of the three-level inverter at time k is to the predicted voltage V * (k), the smaller the difference between the output current I(k+1) and the predicted current I * (k+1), and the smaller the value of the constraint term (13) in the target function; combining formula (17) and formula (5), and simplifying, we get:

[0050]

[0051] By formula (18), the value function formula (16) is adopted as:

[0052] g 1min = |V * (k+1)-V(k+1)|+λ|ΔV c (k+2)| (19)

[0053] By substituting the 27 voltage vectors, g 1min is minimized when the switch state at time k+2 is obtained.

[0054] Further, the inverter mode calculates the reference current I* by the given power, predicts the optimal switch state under the finite set, and gives the optimal gating pulse to the energy storage converter, as follows:

[0055] Step 4.1, the current reference value i is obtained by Clarke transformation based on the voltage measurement value u(k), the given active power and reactive power values of grid-connected * (k), while measuring the grid-connected sampling current i(k);

[0056] Step 4.2, the current reference value at k+1 time is predicted by phase-shifting the current reference value * (k+1);

[0057] Step 4.3, the current reference value at k+2 time is taken as the reference current by using the two-step time-delay compensation optimization algorithm in a phase-shifting manner * (k+2);

[0058] Step 4.4, the output voltage of the neutral-point-clamped three-level converter is divided into 27 different voltage vectors, each voltage vector has a different switching sequence, and the 27 output voltage vectors u i (k) have a total of 3 different values, which are -0.5Vdc, 0, 0.5Vdc, wherein Vdc is the average value V avg calculated by dynamically tracking the voltage value of the energy storage module on the primary side, and a certain sampling frequency is set;

[0059] Step 4.5, the predicted value i(k+2) of the sampling current at k+2 time under the 27 groups of voltage vectors u i (k) screened out in step 4.4 is calculated according to the grid-connected current sampling value, the DC side capacitor voltage difference value AVc(k) at k time, and the switching state sequence Sn to predict the DC side capacitor voltage difference value AVc(k+2) at k+2 time;

[0060] Step 4.6, the reference voltage reference value u * (k+1) at k+1 time is calculated according to the reference current predicted value i * (k+2) obtained in step 4.3, and the 27 groups of voltage predicted values u(k+1) at k+1 time are calculated based on the 27 groups of predicted values i(k+2) of the sampling current, which avoids the prediction of the instantaneous current i(k+2) at k+2 time, and instead uses the voltage u * (k+1) required to reach the given current i * (k+2);

[0061] Step 4.7, the 27 groups of voltage predicted values u(k+1) calculated in step 4.6 and the reference voltage predicted value u * (k+1) at k+2 time calculated in step 4.6 are used to measure the voltage predicted value u(k+1) and the reference predicted value u *(k+1) and the DC side capacitor voltage difference value AVc(k+2) at the time k+2 in step 4.5, to construct the value function g 1min And according to the value function, the most suitable voltage prediction value u(k+1) and the switching sequence Sn in 27 groups are calculated;

[0062] Step 4.8, according to step 4.7, the optimal gating pulse of the energy storage converter of the midpoint clamped three-level inverter is given.

[0063] Further, the current given value i * (k) is calculated based on the given active power and reactive power of the grid-connected in step 4.1, specifically including:

[0064] First, the grid-connected voltage u c is obtained by Clarke transformation to get u cα , u cβ , and then the current I* α , I* β is calculated by formula as shown in formula (20), (21):

[0065]

[0066] Then I* α , I* β is obtained by inverse Clarke transformation to get i * (k) as the current at time k.

[0067] Further, in the rectification mode, the reference current value is calculated by PI control of the voltage outer loop, specifically the given energy storage side voltage reference value V ref and the measured value V dc is obtained by PI control to get the reference current I* d , and the reference current I* q in dq coordinate system is set to 0, I* d , I* q is combined with I* obtained by inverse Clarke transformation and brought into the MPC value function g 1min to get the optimal gating pulse of the converter, specifically as follows:

[0068] Step 6.1, in the charging rectification mode, the primary side charging adopts voltage outer loop control and current inner loop model predictive control to dynamically track the primary side energy storage module voltage value to calculate its average value V avg , which is subtracted from the reference value V ref and the difference is sent to the voltage outer loop PI controller to output the model predictive control current reference value i * (k);

[0069] Step 6.2, the optimization algorithm of time delay compensation in two steps, through the phase shift method to give the current value i * (k+2) as the reference current;

[0070] Step 6.3, the output voltage of the midpoint clamped three-level converter is divided into 27 different state voltage vectors, each voltage vector has a different switching sequence, and the 27 output voltage vectors u i (k) has 3 different values, -0.5Vdc, 0, 0.5Vdc, where Vdc is the average value V avg， Set a certain sampling frequency;

[0071] Step 6.4, according to the grid-connected current sampling value, calculate the 27 groups of voltage vectors u i (k) of the sampling current at k+2 time, through the DC side capacitor voltage difference value ΔVc(k) at k time, the switching state sequence Sn predicts the DC side capacitor voltage difference value ΔVc(k+2) at k+2 time;

[0072] Step 6.5, according to the reference current prediction value i * (k+2) obtained in step 6.2, calculate the reference voltage given value u * (k+1), based on the 27 groups of prediction values i(k+2) of the sampling current, calculate the 27 groups of voltage prediction values y(k+1) at k+1 time, which can avoid the prediction of the instantaneous current i(k+2) at k+2 time, and instead use the voltage u * (k+1) required to reach the given current i * (k+2);

[0073] Step 6.6, according to the 27 groups of voltage prediction values u(k+1) calculated in step 6.5 and the reference voltage prediction value u * (k+1) calculated in step 6.5, through the measurement voltage prediction value u(k+1), the reference prediction value u * (k+1) and the DC side capacitor voltage difference value ΔVc(k+2) at k+2 time in step 6.4, construct the value function And according to the value function, calculate the most suitable voltage prediction value u(k+1) and switching sequence Sn from the 27 groups;

[0074] Step 6.7, according to the step 6.6, the energy storage converter of the inverter is given a gating pulse.

[0075] Further, in the reactive power compensation mode, the PCS outputs a value equivalent to the opposite value of the reactive component of the load current, so that the current vector has only the reactive component; sample the load current Ir I is obtained by Clarke transformation of the phase wt of the grid-connected voltage. rd I rq Then in I* d =0、I* q =-I rq When performing the inverse Park transformation on the value of the reference current in the dq coordinate system, a phase angle compensation is provided, and the compensated i * (k+2) is substituted into the value function as a reference current, and g is selected... 1min The switching state at the minimum value is the optimal turn-on pulse, as detailed below:

[0076] Step 7.1, Measure the load current i at time k. r (k);

[0077] Step 7.2: Predict the load current i at time k+2 using a phase-shifting method. r (k+2);

[0078] Step 7.3: Calculate the grid-connected voltage phase wt using a phase-locked loop;

[0079] Step 7.4, load current i r (k+2) Based on the grid-connected voltage phase wt, the i in the dq coordinate system is obtained through Park transformation. r d、i r q, where i r q is the reactive current value;

[0080] Step 7.5, take i r d, -i r q is the reference current at time k+2, and i is obtained through the inverse Park transform. * (k+2), as the reference current prediction value, based on i * (k+2) Calculate the phase shift caused by the LC filter at time k+2, and compensate for i. * (k+2) yields the predicted reference current value i * (k+2);

[0081] Step 7.6: Divide the output voltage of the midpoint clamped three-level converter into 27 voltage vectors with different states. Each voltage vector has a different switching sequence. The 27 output voltage vectors u i (k) There are a total of 3 different values: -0.5Vdc, 0, and 0.5Vdc, where Vdc is the average value calculated by dynamically tracking the voltage of the primary-side energy storage module. avg Set a certain sampling frequency;

[0082] Step 7.7. Calculate the 27 groups of voltage vectors u based on the grid-connected current sampling value in step 7.6 i (k) The predicted value of the sampling current at time k+2, i(k+2), is predicted by the DC side capacitor voltage difference value ΔVc(k) at time k and the switching state sequence Sn;

[0083] Step 7.8. In this step, the reference voltage given value u * (k+1) is calculated based on the 27 groups of predicted values of the sampling current i(k+2) in step 7.5, and the 27 groups of voltage predicted values u(k+1) at time k+1 are calculated, which avoids the prediction of the instantaneous current i(k+2) at time k+2, and instead uses the voltage u * (k+1) required to reach the given current i * (k+2). * (k+1);

[0084] Step 7.9. In this step, the 27 groups of voltage predicted values u(k+1) calculated in step 7.8 and the reference voltage predicted value u * (k+1) at time k+2 are used to measure the voltage predicted value u(k+1), the reference predicted value u * (k+1), and the DC side capacitor voltage difference value ΔVc(k+2) at time k+2 in step 7.7 to construct a value function and calculate the most suitable voltage predicted value u(k+1) and switching sequence Sn from the 27 groups according to the value function;

[0085] Step 7.10. According to step 7.9, the energy storage converter gating pulse is given to the inverter.

[0086] Further, in step 7.5, the phase shift caused by the LC filter at time k+2 is calculated based on i * (k+2), which compensates for i * (k+2), specifically:

[0087] The phase angle difference of the grid-connected current at time k+2 is calculated based on i * (k+2), and i * (k+2) is regarded as a constant current source, and u c (k+2) is regarded as a constant voltage source, where u c (k+2) is based on u c (k) obtained by compensating the phase angle of the sampling period, and the value of the grid-connected current is calculated by Kirchhoff's law, which is i *The phase angle difference wt* of (k+2) and I2(k+2) is the phase angle difference caused by the LC filter, and the value of wt* after compensation of wt is brought into the inverse Park calculation to obtain compensated I* α , I* β , I* α , I* β The reference current I* is obtained through the inverse Clarke calculation and brought into the value function g 1min .

[0088] Advantages:

[0089] The inverter topology of the application selects a neutral point clamped (NPC) three-level inverter, which has the advantages of half DC bus voltage stress of the power tube, less equivalent capacitor compared with other three-level inverters, strong theoretical analysis uniformity, and direct extension to more level inverter topologies, etc. The model prediction control (MPC) is introduced to achieve the purposes of optimization algorithm design, parameter setting process omission and dynamic performance improvement, so as to solve the problems of slow response, complex algorithm, more links, complex parameter setting, etc. caused by the traditional PCS control method.

[0090] The application sets a fixed value function to simplify the design of the control mode, so that the energy storage converter has the additional reactive power compensation function in addition to the traditional energy storage and discharge functions. Meanwhile, the MPC control of the application selects the optimal solution at the next moment to provide extremely fast followability for the energy storage converter control. The model prediction control method of the application further optimizes the output effect of the energy storage converter by using the two-step delay compensation optimization algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0091] Figure 1 It is the topology structure diagram of the three-level PCS control method based on model prediction of the application.

[0092] Figure 2 It is each vector diagram of the NPC grid-connected inverter used in the application.

[0093] Figure 3 It is the equivalent diagram of the LC filter at k+2 moment.

[0094] Figure 4 It is the working mode schematic diagram of the application.

[0095] Figure 5 It is the active power and reactive power diagram of the PCS output under the rated power in the inverter mode of the application.

[0096] Figure 6 This is a diagram showing the change of DC side voltage under the rated DC side voltage Vref in rectification mode according to the present invention.

[0097] Figure 7 This is the load grid-connected current diagram before compensation in the reactive power compensation mode of this invention.

[0098] Figure 8 This is the grid-connected current diagram after compensation in reactive power compensation mode according to the present invention.

[0099] Figure 9 This is a graph showing the change of grid-connected current value over time in this invention. Detailed Implementation

[0100] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0101] This invention discloses a control method for an energy storage converter in a midpoint clamped three-level inverter, including a model-predictive inverter control method, a model-predictive rectification control method, and a model-predictive reactive power compensation control method. The energy storage converter based on the midpoint clamped converter determines the specific operating mode according to the energy storage module's charge status, the current time period, and the energy storage module's voltage.

[0102] The storage strategy of this invention is as follows: Figure 4 As shown, the energy storage converter based on the midpoint clamping converter determines its specific operating mode according to the energy storage module's state of charge, the current time period, and the energy storage module's voltage, switching between three operating modes, specifically:

[0103] Step 2.1: Determine whether it is outside of off-peak electricity hours or outside of the designated energy storage power supply period.

[0104] Step 2.2: If the determination is negative, then determine whether the voltage of the PCS energy storage device is less than the set value U1.

[0105] Step 2.3: If the determination in step 2.2 is yes, then the rectification mode is used to charge the energy storage device.

[0106] Step 2.4: Detect the SOC (State of Charge) of the energy storage module. If it is greater than or equal to 95%, stop charging; otherwise, continue charging.

[0107] Step 2.5: If the determination in step 2.2 is negative, then the reactive power compensation mode is adopted to improve the power factor of the power grid.

[0108] Step 2.6: If the judgment in step 2.1 is yes, then determine whether the voltage of the PCS energy storage device is less than the set value U2. U2 should be set to be greater than U1.

[0109] Step 2.7: If step 2.6 is yes, then the reactive power compensation mode is adopted to improve the power factor of the power grid.

[0110] Step 2.8: If step 2.6 is no, then the inverter mode is adopted to discharge the power storage at a constant power to the power grid.

[0111] The inverter control mode based on model prediction is the inverter mode, the power storage discharges, the reference current I* is calculated through the given power and is brought into the value function, the optimal switching state under the limited set is predicted, and the optimal gating pulse of the power storage converter is given.

[0112] The rectifier control mode based on model prediction is the rectifier mode, the power storage charges, the reference current I* is obtained through PI control, and the optimal switching quantity is selected by bringing the reference current I* into the value function based on the optimal gating pulse of the power storage converter.

[0113] The reactive power compensation control mode based on model prediction is the reactive power compensation mode, the reactive current generated by the load in the same system is compensated, the reference current I* is calculated through the load current, and the optimal switching quantity is selected by bringing the reference current I* into the value function, and the optimal gating pulse of the power storage converter is given.

[0114] The value function is constructed as follows:

[0115] Combined with the mathematical model of the three-phase voltage type inverter, the system modeling can be carried out. From Figure 1 , the mathematical model of the three-phase voltage type inverter can be analyzed as shown in the following formula (1):

[0116]

[0117] In the formula, L is the filter inductance in series with the inverter, I a , I b , I c are the currents flowing through each phase of the filter inductance, u ca , u cb , u cc are the output voltages of each phase of the filter capacitor, V a , V b , V c are the voltages output by each phase of the inverter, and R is the line impedance.

[0118] The αβ-axis current and voltage equation is as follows:

[0119]

[0120] In the formula, the variables with subscript αβ are the values of the corresponding variables after αβ coordinate transformation. The specific transformation method is shown in the following formula (3) and (4):

[0121]

[0122] Since the MPC method involves k time and k+1 time, the constraint relationship between the controlled variable x(k) and the predicted variable x(k+1) can be expressed by discretization, as shown in the following formula (5):

[0123]

[0124] According to the Kirchhoff's current law Figure 1 The current i flowing out of the O node in the formula (5) can be written as:

[0125] i o = i o = -i c1 (6)

[0126] Since the energy storage voltage V c2 is variable, the 27 vectors u dc will change dynamically according to the sampling voltage value at k time, and the amplitude is 0, ±V n / 3, ±2V dc / 3. When the specific switching state S dc is selected according to the NPC grid-connected inverter principle, the inverter outputs 27 voltage vectors u n corresponding to Figure 2 , the current i n on the DC side capacitor, i c1 , i c2 can be expressed by the following formula:

[0127]

[0128] I a +I b +I c =0 (10)

[0129] In the above formula, V c1 , V c2 are the DC side capacitor voltages, I a , I b , I c are the current of each phase of the three-phase current I abc . Substituting the above formula (7)-(10) into formula (6), the expression of the midpoint voltage and the three-phase current is obtained as:

[0130]

[0131] In the above formula, ΔV c is V c1 -V c2 , which is called the midpoint voltage.

[0132] The present application selects power frequency alternating current as the grid-connected source. The grid voltage period is much larger than the sampling period used above, so formula (11) can be discretized into formula (12) by forward difference method:

[0133] Through the above formula, the midpoint voltage value ΔV c (k) at time k can be obtained by sampling the midpoint voltage value ΔV n (k) at time k. abc The three-phase current I c (k) at time k.

[0134] The conventional control method of the three-level inverter usually takes the grid-connected voltage, output current and midpoint voltage of the DC side as the controlled quantity. As described in the previous section, the voltage and current double-loop control method. For MPC, the above controlled quantities can be used as the control target of the objective function, and are added to the final objective function through different constraint terms and weight coefficients. The objective function of the present application mainly selects the output current as the controlled object of the objective function. The specific expression is shown in formula (13) as follows. In order to compensate for the errors caused by the switching time of components and the calculation time of the algorithm, a two-step delay compensation method is used, as shown in formula (14):

[0135] g i1 = |I * (k+1)-I(k+1)|

[0136] g i2 = |I * (k+2)-I(k+2)|

[0137] In the above formula, I*(k+1) is the given output current at time k+1. Formula (13) and formula (14) can both constrain the output current value and achieve current tracking control. Since model predictive control is usually calculated in the αβ coordinate system, in order to facilitate the design and operation in practical application, formula (13) is selected as the current constraint term added to the value function.

[0138] Another control target of the objective function is to keep the DC side capacitor voltage difference ΔV c (k) constant at 0. To achieve this purpose, the midpoint voltage constraint term is added to the objective function designed in the present application by a two-step delay compensation method, as shown in formula (15):

[0139]

[0140] Combining formula (14) and formula (15), the objective function used in the present application is:

[0141] g 2min = |I* (k+2)-I(k+2)|+λ|ΔV c (k+2)| (16)

[0142] The design value function can be used to achieve a given current I. * The voltage V required for (k+1) * (k). By selecting the optimal V * (k), so that the output current I(k+1) is close to the given current I at the same time. * (k+1), through equation (5), using I * (k+1) replaces I(k+1), V * Replacing V(k) with (k) and rearranging them, we get the following equation:

[0143]

[0144] Equation (17) shows that the voltage V(k) output by the three-level inverter at time k is similar to the predicted voltage V. * The closer (k) is to the predicted current I(k+1), the closer the output current I(k+1) is to the predicted current I. * The smaller the difference (k+1), the smaller the value of constraint term (13) in the objective function. Combining equations (17) and (5), we can simplify to obtain:

[0145] Through equation (18), the value function equation (16) can be adopted as:

[0146] g 1min =|V * (k+1)-V(k+1)|+λ|ΔV c (k+2)| (19)

[0147] Inverter mode:

[0148] Inverter mode such as Figure 1 As shown, the specific operation is based on the grid-connected voltage u c To calculate the rated power P ref Q ref The rated current under the specified conditions. Specifically, the method involves first setting the grid-connected voltage u... c u along the αβ axis is obtained using the Clarke transform. cα u cβ Then, the current I* is calculated using the formulas shown in equations (20) and (21). α 、I* β .

[0149]

[0150] Then I* α 、I* βThe I* is obtained by inverse Clarke transformation and is brought into the MPC value function above to predict the optimal switching state under the limited set, and the optimal gating pulse of the energy storage converter is given.

[0151] Rectification mode:

[0152] The rectification mode is as shown in Figure 1 , and the given energy storage side voltage reference value V ref and the measured value V dc are obtained by PI control to obtain the reference current I* d , and the reference current I* q in the dq coordinate system is set to 0, i.e. the reactive component is 0, I* d , I* q Combined with the I* obtained by inverse Clarke transformation and brought into the MPC value function above, the optimal gating pulse of the converter is obtained.

[0153] Reactive power compensation mode:

[0154] The reactive power compensation mode is as shown in Figure 1 , and the load current I r is first sampled, and I rd , I rq are obtained by Clarke transformation with the phase wt of the grid-connected voltage, wherein I rq is the reactive component of the load current, which provides reactive power. Therefore, the PCS energy storage inverter provides a grid-connected current with a d-axis value of 0 and a q-axis value of -I rq in the dq coordinate system, which can compensate for the reactive current component of the load and improve the power factor. Considering that the inverter is in series with an LC filter circuit, which will cause phase shift, a phase angle compensation is provided when I* d = 0, I* q = -I rq current is inverse Park transformed.

[0155] Figure 3 The equivalent diagram of the LC filter at k+2 time is given, in which I * (k+2) can be regarded as a constant current source, and u c (k+2) can be regarded as a constant voltage source. Wherein u c (k+2) can be obtained based on u c (k) at k time by compensating the phase angle of the sampling period. The value of the grid-connected current is calculated by Kirchhoff's law, and the phase angle difference wt* between I * (k+2) and I2(k+2) is obtained, i.e. the phase angle difference caused by the LC filter. The value after compensating wt is brought into the inverse Park calculation above, and the compensated I* α , I* βI* is calculated by inverse Clarke, and is brought into the value function above. α I* is calculated by inverse Clarke, and is brought into the value function above. β I* is calculated by inverse Clarke, and is brought into the value function above.

[0156] Simulation experiment

[0157] In this embodiment, a three-level PCS energy storage converter is built by using Simulink tool in MATLAB. The DC power is inverted to three-phase voltage by three-level circuit after passing through DC bus capacitor, and the smooth three-phase sinusoidal voltage is output by the virtual model predictive control method based on three-level converter. The electrical parameter settings in the simulation process are as follows:

[0158]

[0159] Considering that in actual application, due to large voltage and long charging and discharging time, it is not conducive to simulate the charging and discharging process, this embodiment selects a smaller value for simulation. As shown in FIG. 6, in the inverter mode, the output quickly realizes the rated power output, which embodies the rapidity of model prediction. As shown in FIG. 7, in the rectifier mode, the DC side voltage is quickly raised to the set DC bus voltage reference value, achieving the effect of charging the PCS by rectification. As shown in FIG. 8, the load outputs about 4A of reactive current, which affects the power factor of the system, and as shown in FIG. 9, in the reactive power compensation mode, the reactive current is successfully reduced to 0, which improves the power factor of the system. As can be seen from FIG. 10, after the reference current I* is given, the PCS quickly outputs the corresponding current I, which embodies the good follow-up of model predictive control. Figure 5 Figure 6 Figure 7 Figure 8 Figure 9

[0160] The above embodiments are only for illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and implement it, and cannot limit the protection scope of the present application. Any equivalent transformation or modification made according to the spirit and essence of the present application should be covered within the protection scope of the present application.​​​​​

Claims

1. A method of controlling an energy storage converter of a neutral point clamped three-level inverter, characterized by, The control method comprises a model prediction-based inversion control mode, a model prediction-based rectification control mode and a model prediction-based reactive compensation control mode; the energy storage converter based on the midpoint clamped converter determines a specific working mode according to the state of charge of the energy storage module, a current period and the voltage of the energy storage module, and the working mode is as follows: The model prediction-based inversion control mode is an inversion mode, the energy storage is discharged, a reference current I* is calculated through a given power, and the reference current I* is brought into a value function to predict an optimal switch state in a limited set and give the energy storage converter an optimal gating pulse; The model prediction-based rectification control mode is a rectification mode, the energy storage is charged, a reference current I* is obtained through PI control, the optimal switch quantity is selected by bringing the reference current I* into the value function, and the optimal gating pulse is given to the energy storage converter; The model prediction-based reactive compensation control mode is a reactive compensation mode, reactive currents generated by loads in the same system are compensated, a reference current I* is calculated through the load current, the optimal switch quantity is selected by bringing the reference current I* into the value function, and the optimal gating pulse is given to the energy storage converter; The specific working mode is determined according to the state of charge of the energy storage module, the current period and the voltage of the energy storage module, and is as follows: Step 2.1: whether it is in a non-electricity low valley period or a set non-energy storage power supply period is judged; Step 2.2: if it is judged as no, whether the voltage of the PCS energy storage device is less than a set value U1 is judged; Step 2.3: if it is judged as yes in step 2.2, the rectification mode is adopted to charge the energy storage device; Step 2.4: the state of charge SOC of the energy storage module is detected, if it is greater than or equal to 95%, the charging is stopped, otherwise the charging is continued; Step 2.5: if it is judged as no in step 2.2, the reactive compensation mode is adopted to improve the power factor of the power grid; Step 2.6: if it is judged as yes in step 2.1, whether the voltage of the PCS energy storage device is less than a set value U2 is judged, U2 should be set to be greater than U1; Step 2.7: if it is judged as yes in step 2.6, the reactive compensation mode is adopted to improve the power factor of the power grid; Step 2.8: if it is judged as no in step 2.6, the inversion mode is adopted to discharge the power grid with a constant power.

2. The control method of the energy storage converter of the neutral point clamped three-level inverter according to claim 1, characterized in that, The value function is used to select the optimal switch quantity, through a two-step time delay compensation optimization algorithm, the switch state at the time k+2 when the minimum value of the value function is selected is taken as the switch quantity at the time k+2, and the specific expression is as follows: The mathematical model of the three-phase voltage type inverter is shown in the following formula (1): where L is the filter inductance in series with the inverter, I a , I b , I c is the phase a, b, c current through the filter inductance, u ca , u cb , u cc is the output voltage of the filter capacitor for each phase, V a , V b , V c is the voltage output of the inverter for each phase, and R is the line impedance. The αβ-axis current voltage equation is as follows: In the formula, the variable with subscript αβ is the value of the corresponding variable after αβ coordinate transformation, and the specific transformation mode is shown in the following formula (3) and (4): After the formula (2) is discretized, it is shown in the following formula (5): According to the Kirchhoff's current law, the current i flowing out of the neutral point NPC of the three-level inverter on the DC side o Can be written as: i o = i c1 - i c2 (6) Since the energy storage measurement V dc is changing, 27 vector u n Will change, the amplitude of the sampling obtained from the energy storage side voltage value at time k is 0, ±V dc / 3, ±2V dc / 3; when the specific switch state S n Is selected according to the principle of NPC grid-connected inverter, the corresponding 27 voltage vectors u n Is outputted by the inverter, the current i c1 On the DC side capacitor, i c2 It can be expressed by the following formula: I a +I b +I c =0 (10) wherein V c1 , V c2 is the DC side capacitor voltage, I a , I b , I c is the a, b, c phase current passing through the filter inductor; substituting the above equations (7)-(10) into equation (6), the expression of the midpoint voltage and three-phase current is obtained as: where ΔV c is V c1 - V c2 , denoted as the midpoint voltage; The power frequency alternating current is selected as the grid-connected source, the grid voltage period is much larger than the sampling period, and therefore the formula (11) is discretized into the following formula (12) through the forward difference method: By sampling the midpoint voltage value ΔV at time k c (k), the switching state S n and the three-phase current I abc , the next time midpoint voltage value ΔV c (k+1) is predicted and analyzed The value function selects the output current as the controlled object of the target function, and the specific expression is shown in the following formula (13), and the two-step time delay compensation optimization algorithm adopts the formula (14): g i1 = |I * (k+1)-I(k+1)| (13) g i2 = |I * (k+2)-I(k+2)| (14) In the formula, I*(k+1) is the given output current at the time k+1, and the formula (13) is selected as the constraint term of the current to be added to the value function. Another control target of the objective function aims to keep the DC side capacitor voltage difference AV c The constraint term of the midpoint voltage is added to the objective function by a two-step delay compensation method, and the control is always 0, as shown in the following formula (15): Combining equation (14) with equation (15), the final adopted objective function is: g 2min = |I * (k+2)-I(k+2)|+λ|ΔV c (k+2)| (16) The value function is designed to use the voltage V * (k+1) required to reach the given current I * (k) by selecting the optimum V * (k) such that the output current I(k+1) approaches the given current I * (k+1) at the same time, by substituting I * (k+1) for I(k+1) and V * (k) for V(k) in equation (5) and rearranging to give Equation (17) shows that the closer the output voltage V(k) of the three-level inverter at time k to the predicted voltage V * (k), the smaller the difference between the output current I(k+1) and the predicted current I * (k+1), and the smaller the value of the constraint term (13) in the objective function. By combining equation (17) and equation (5), we get: Through equation (18), the value function equation (16) is adopted as: g 1min = |V * (k+1)-V(k+1)|+λ|ΔV c (k+2)| (19) The 27 voltage vectors are brought in such a way that g 1min The minimum is the switching state at the instant k+2.

3. The control method of the energy storage converter of the neutral point clamped three-level inverter according to claim 2, characterized in that, The inverter mode calculates the reference current I* through the given power, predicts the optimal switch state under the limited set, and gives the optimal gating pulse to the energy storage converter, as follows: Step 4.1 The current reference values i are derived from the voltage measurements u(k), the given active and reactive power values of the grid connection by Clarke transformation * (k), while measuring the grid-connection sample current i(k); Step 4.2, the current given value at k+1 time is predicted by phase shift way * (k+1); Step 4.3, the optimization algorithm of time delay compensation by two steps, the current given value i * (k+2) as the reference current; Step 4.4, the output voltage of the midpoint clamped three-level converter is divided into 27 different state voltage vectors, each voltage vector has a different switching sequence, 27 output voltage vectors u i (k) There are 3 different values in total, which are -0.5Vdc, 0, 0.5Vdc, where Vdc is the average value of the dynamic tracking primary side energy storage module voltage value avg , set a certain sampling frequency; Step 4.

5. Calculate the 27 groups of voltage vectors u selected in step 4.4 according to the grid-connected current sampling value i (k) The predicted value i(k+2) of the sampling current at k+2 time is predicted by the DC side capacitor voltage difference value AVc(k) at k time, the switch state sequence Sn, the DC side capacitor voltage difference value AVc(k+2) at k+2 time. Step 4.

6. The reference current prediction value i(k+2) is obtained according to step 4.3 * (k+2) The reference voltage given value u(k+1) at k+1 time is calculated * (k+1) The 27 groups of voltage prediction values u(k+1) at k+1 time are calculated based on the 27 groups of prediction values i(k+2) of the sampling current, which avoids the prediction of the instantaneous current i(k+2) at k+2 time, and instead uses the voltage u(k+1) required to reach the given current i * (k+2) at k+2 time * (k+1) Step 4.

7. According to the 27 groups of voltage prediction values u(k+1) calculated in step 4.6 and the reference voltage prediction value u(k+1) at k+2 calculated in step 4.6, the most suitable voltage prediction value u(k+1) and the switching sequence Sn in the 27 groups are calculated by measuring the voltage prediction value u(k+1), the reference prediction value u(k+1) and the DC side capacitor voltage difference value ΔVc(k+2) at k+2 in step 4.5, constructing the value function g * (k+1), and calculating the most suitable voltage prediction value u(k+1) and the switching sequence Sn in the 27 groups according to the value function g * (k+1), and calculating the most suitable voltage prediction value u(k+1) and the switching sequence Sn in the 27 groups according to the value function g 1min (k+1), and calculating the most suitable voltage prediction value u(k+1) and the switching sequence Sn in the 27 groups according to the value function g Step 4.8, according to step 4.7, the optimal gating pulse is given to the energy storage converter of the neutral point clamped three-level inverter.

4. The control method of the energy storage converter of the three-level neutral point clamped inverter according to claim 3, characterized in that, The step 4.1 of calculating the current given value i based on the values of the given active power and reactive power of the grid * (k), specifically comprising: The grid voltage u c The u cα , u cβ in the αβ axis is obtained by Clarke transformation α , I* β : where P ref is the rated active power, Q ref is the rated reactive power; and I* α , I* β are the active and reactive currents, respectively, at the kth time instant. * is obtained by inverse Clarke transformation as i*(k) at the kth time instant.

5. The control method of the energy storage converter of the three-level neutral point clamped inverter according to claim 2, characterized in that, In the rectification mode, the reference current value is calculated by the voltage outer loop of PI control, specifically, the given energy storage side voltage reference value V ref and the measured value V dc The reference current I* d is obtained by PI control, and the reference current I* q in the dq coordinate system is set to 0, I* d , I* q Combined with I* obtained by inverse Clarke transformation, the optimal pulse of the converter is obtained by bringing it into the MPC value function g 1min , specifically as follows: Step 6.1, in the charging rectification mode, the primary side charging adopts voltage outer loop control, current inner loop model predictive control, and dynamically tracks the primary side energy storage module voltage value to calculate its average value V avg , and subtracts it from the reference value V ref , and sends the difference to the voltage outer loop PI controller to output the model predictive control current reference value i * (k); Step 6.2, the optimization algorithm of time delay compensation with two steps, the current given value i * (k+2) as the reference current; Step 6.3, the output voltage of midpoint clamped three-level converter is divided into 27 different state voltage vectors, each voltage vector has different switching sequence, 27 output voltage vectors u i (k) There are 3 different values in total, -0.5Vdc, 0, 0.5Vdc, where Vdc is the average value of the dynamic tracking of the primary side energy storage module voltage value avg， Set a certain sampling frequency; Step 6.

4. Calculate the 27 groups of voltage vectors u selected in step 6.3 according to the grid-connected current sampling value i (k) The predicted value i(k+2) of the sampling current at k+2 time is predicted by the DC side capacitor voltage difference value AVc(k) at k time, the switch state sequence Sn, the DC side capacitor voltage difference value AVc(k+2) at k+2 time. Step 6.

5. The reference current prediction value i(k+2) obtained from step 6.2 * (k+2) is used to calculate the reference voltage given value u(k+1) at k+1 time * (k+1) based on the 27 groups of prediction values i(k+2) of the sampling current, it is known that the prediction of the instantaneous current i(k+2) at k+2 time is avoided, and the voltage u * (k+2) required to reach the given current i * (k+1) is used; Step 6.

6. According to the 27 voltage prediction values u(k+1) calculated in step 6.5 and the reference voltage prediction value u * (k+1) calculated in step 6.5, the value function is constructed by measuring the voltage prediction value u(k+1), the reference prediction value u * (k+1) and the DC side capacitor voltage difference value AVc(k+2) at k+2 in step 6.4 and the most suitable voltage prediction value u(k+1) and switch sequence Sn in the 27 groups are calculated according to the value function. Step 6.7, according to step 6.6, the gating pulse is given to the energy storage converter of the inverter.

6. The control method of the energy storage converter of the three-level neutral point clamped inverter according to claim 2, characterized in that, The reactive compensation mode, through the PCS output equivalent to the load current of the opposite value of the reactive component, so that the current vector and only reactive component; sampling load current I r , I rd , I rq , then in I* d = 0, I* q =-I rq As the reference current in the dq coordinate system, the inverse Park transformation provides a phase angle compensation, and the compensated i * (k+2) as the reference current into the value function, by selecting g 1min The minimum value of the switch state is the optimal pulse, as follows: Step 7.1, measuring load current i at time k r (k); Step 7.2, predicting the load current i at time instant k+2 by phase-shifted approach r (k+2); Step 7.3, the grid-connected voltage phase wt is calculated through the phase-locked loop; Step 7.4, load current i r (k+2) Based on the grid voltage phase wt, through Park transformation, i in dq coordinate system is obtained r d, i r q, where i r q is the reactive current value; Step 7.5, take i r d, -i r q is the reference current at time k+2, obtained by inverse Park transformation of i * (k+2) as a reference current prediction, based on i * (k+2) to calculate the phase offset caused by the LC filter at time k+2, to compensate i * (k+2) to obtain the reference current prediction i * (k+2). Step 7.6, the output voltage of the midpoint clamped three-level converter is divided into 27 different state voltage vectors, each voltage vector has a different switching sequence, 27 output voltage vectors u i (k) There are 3 different values in total, -0.5Vdc, 0, 0.5Vdc, where Vdc is the average value of the dynamically tracked primary side energy storage module voltage value avg , set a certain sampling frequency; Step 7.

7. Calculate the 27 groups of voltage vectors u selected in step 7.6 according to the grid-connected current sampling value i (k) The predicted value i(k+2) of the sampling current at k+2 time is predicted by the DC side capacitor voltage difference value AVc(k) at k time, the switch state sequence Sn, the DC side capacitor voltage difference value AVc(k+2) at k+2 time. Step 7.8, in this step, the reference current prediction value i * (k+2) is calculated according to the reference voltage given value u * (k+1) at k+1 moment based on the 27 groups of prediction values i(k+2) of the sampling current, it is known that the prediction of the instantaneous current i(k+2) at k+2 moment is avoided, and the voltage u * (k+1) required to reach the given current i * (k+2) is used; Step 7.9, in this step, according to the 27 groups of voltage prediction value u(k+1) and reference voltage prediction value u(k+1) at k+2 time calculated in step 7.8, measured voltage prediction value u(k+1), reference prediction value u(k+1) and DC side capacitor voltage difference value ΔVc(k+2) at k+2 time in step 7.7, the value function is constructed * (k+1), over-measured voltage prediction value u(k+1), reference prediction value u * (k+1) and DC side capacitor voltage difference value ΔVc(k+2) at k+2 time in step 7.7, the value function is constructed and the most suitable voltage prediction value u(k+1) and switch sequence Sn in 27 groups are calculated according to the value function; Step 7.10, according to step 7.9, the gating pulse is given to the energy storage converter of the inverter.

7. The control method of the energy storage converter of the three-level neutral point clamped inverter according to claim 6, characterized in that, The step 7.5 is based on i * (k+2) calculates the phase offset caused by the LC filter at the k+2 moment, compensates i * (k+2), specifically: The value of the grid-connected current at time k+2 is obtained by * the phase angle difference between i * (k+2) and i c (k+2) is regarded as a constant voltage source, where u c (k+2) is based on u c (k) obtained by compensating the phase angle of the sampling period, the value of the grid-connected current is calculated by Kirchhoff's law, and i * (k+2) and the phase angle difference wt* between I2(k+2) is the phase angle difference brought by the LC filter. The value after compensating wt is brought into the inverse Park calculation to obtain the compensated I* α , I* β , I* α , I* β The reference current I* is obtained by inverse Clarke calculation, which is brought into the value function g 1min .

Images