CN106982000A - A kind of micro-capacitance sensor data-driven inverter parallel intelligent control method - Google Patents

Abstract

The invention belongs to micro-capacitance sensor control technology field, a kind of micro-capacitance sensor data-driven inverter parallel intelligent control method is disclosed.The problems such as the method overcome the Unmarried pregnancy of conventional inverter control method for parallel presence, i.e., occur parameter drift when there is external interference, to balance the distribution of the bearing power between each inverter.Simultaneously, for the deficiency of existing modulation technique, the PWM inverter control method solved with the reverse line voltage of discrete type, avoid the problem of midpoint of laod unbalance, DC voltage can not find, even if being that can reach poised state according to required line voltage so as to realize in the case of laod unbalance.

Description

A kind of micro-capacitance sensor data-driven inverter parallel intelligent control method

Technical field

The invention belongs to micro-capacitance sensor control technology field, more particularly to a kind of micro-capacitance sensor data-driven inverter parallel intelligence Control method.

Background technology

With the development of micro-capacitance sensor technology, on the one hand, the increase of load requires that micro- source provides bigger power output；It is another Aspect, because renewable energy source distribution relative distribution, small scale, production are discontinuous, and traditional centralized power supply system can not be very These demands are met well, the progress parallel connection of each inverter system can be realized into power distribution, and have good redundancy And reliability.Inverter parallel is that can produce circulation between each inverter the problem of being brought, and makes the output electricity of inverter Can Quality Down, the life-span of shortening power electronic devices.

Existing inverter parallel control method, it is many using voltage x currents it is bicyclic with droop control ring group into three ring knots Structure.Droop control obtains voltage electricity according to the inverter output voltage Current calculation power output measured, then by the sagging formula of PQ Reference value is flowed, controlling switch is carried out by the bicyclic PWM waveform that produces of voltage x current.Droop control is protected by sacrificing voltage level Each inverter power mean allocation is demonstrate,proved, needs to be modeled system when designing controller, has Unmarried pregnancy etc. and ask Topic, when systematic parameter changes, control effect is deteriorated.MFA control algorithm is by introducing pseudo- partial derivative, non- Linear system is linearized, then just with the inputoutput data of system to pseudo- partial derivative real-time update, it is not necessary to right System is modeled, and this complication system for inverter parallel in solution micro-capacitance sensor proposes new resolving ideas.

In addition, traditional inverter control method mainly has SPWM, SVPWM, these methods are all according to three-phase phase voltage It is modulated for modulating wave, the three-phase phase voltage of now system output is in the clear situation of load and DC voltage midpoint It is lower just to can guarantee that balance, and in the application system of reality, because the midpoint of inversion system output three-phase phase voltage and inverter are straight Flow side voltage midpoint be difficult to find and connect, so system output real phase voltage be difficult to find, once laod unbalance or What person's DC voltage midpoint was found has error, and system is adjusted even with advanced control strategy, but because three-phase The reason for having coupling between voltage, resulting three-phase phase voltage remains unbalanced, so line voltage is also unbalanced, I.e. it cannot be guaranteed that the line voltage of three-phase inverter output is desired sinusoidal waveform, although and being connect in practice on carrier surface What is entered is phase voltage, but is substantially line voltage in fact, and it is what characteristic no matter to load, as long as line voltage is sinusoidal waveform It just disclosure satisfy that requirement.

The content of the invention

The problems such as purpose of the present invention is the Unmarried pregnancy for overcoming conventional inverter control method for parallel to exist, i.e., ought have Parameter drift occurs during external interference, it is proposed that a kind of micro-capacitance sensor data driven type inverter parallel control method, to balance Bearing power distribution between each inverter.Meanwhile, for the deficiency of existing modulation technique, propose a kind of discrete type reverse line The PWM inverter control method that voltage is solved, it is to avoid the problem of laod unbalance, the midpoint of DC voltage can not find, so that It is that can reach poised state according to required line voltage even if realizing in the case of laod unbalance.

Micro-capacitance sensor data-driven inverter parallel intelligent control method proposed by the present invention, comprises the following steps：

Step 1, the control parameter initial value for obtaining according to droop control method each inverter, that is, it is defeated when stable Enter output data, i.e.,：Produce the modulating wave U of the line voltage form of PWM control pulses_ab、U_bc、U_caWith output three-phase line voltage E_ab、 E_bc、E_ca, carry out obtaining the DC voltage U under dq coordinate systems after 3s/2r conversion_d、U_qAnd E_d、E_q, then according to the linear of inverter Change initial value φ (1) and incidence matrix P that model obtains pseudo- gradient vector.

Step 1-1, the following coordinate transform of inputoutput data progress to collecting：

In formula, C_ZThe direct transform matrix of dq coordinate systems is tied to for abc coordinates, θ is dq coordinate system d axles and abc coordinate system a axles Angle.

Step 1-2, the initial value φ (1) for obtaining according to the inearized model of single inverter pseudo- gradient vector.

By single inverter inearized model：Δ E=φ Δ U transposition can be obtained：

The pseudo- gradient vector approximately equal of two adjacent groups data, can be obtained when stable because inverter is exported：

So as to obtain pseudo- gradient vector initial value：

In formula, U_d(m-1)、U_d(m)、U_d(m+1) it is respectively value of the line voltage modulating wave in m-1, m and m+1 moment d axle, U_q (m-1)、U_q(m)、U_q(m+1) it is respectively q axle value of the line voltage modulating wave at m-1, m and m+1 moment, E_d(m-1)、E_d(m)、E_d(m + 1) it is respectively d axle value of the output voltage at m-1, m and m+1 moment, E_q(m-1)、E_q(m)、E_q(m+1) it is respectively that output voltage exists The q axle values at m-1, m and m+1 moment.

Step 1-3, incidence matrix P calculated according to the inearized model of inverter parallel.

The pseudo- gradient vector initial value φ (1) of gained is substituted into inverter parallel system inearized model Δ E=φ Δs U-P Δe：

P Δ e=φ (1) Δ U- Δs E (6)

Make equal sign both sides each element of matrix equal to try to achieve incidence matrix P.

In formula, Δ e is itself inverter output voltage and the difference of other inverters, and Δ U inputs the change of modulation voltage Amount, the variable quantity of Δ E output voltages.

Step 2, the pseudo- gradient vector initial value obtained using step 1 utilize the input of inverter parallel system as initial point Output data, criterion function is recognized according to by minimization：

J (φ (k))=| | E^*(k+1)-E(k+1)||²+μ||φ(k)-φ(k-1)||² (7)

In formula, E^*(k+1) it is inverter output voltage reference value, E (k+1) is the actual measurement of inverter output voltage It is worth, μ isThe penalty factor of variable quantity.

Obtain pseudo- gradient vector estimateMore new formula：

I.e.

So as to the estimate of the pseudo- gradient vector of online updating

In formula, and η ∈ (0,2], Δ U_dWith Δ U_qIt is the variable quantity that the k-1 moment inputs modulation voltage under dq coordinate systems, Δ respectively E_dWith Δ E_qIt is the variable quantity of k moment output voltages under dq coordinate systems respectively.

Step 3, the pseudo- gradient vector estimate obtained according to step 2Calculate the three-phase line voltage for producing pwm signal Modulating wave, passes through minimization Controlling object function：

J (U (k))=| | E^*(k+1)-E(k+1)||²+λ||U(k)-U(k-1)||² (10)

So as to obtain inputting modulating wave：

I.e.

ρ ∈ in formula (0,2], λ is the penalty factor of U (k) variable quantities, and P, Δ e (k) are identical with defined in formula (8).

Step 4, the input modulating signal U (k) that is obtained in step 3 is transformed to three phase static abc by 2r/3s coordinates sat Line voltage form U under mark system_abˊ、U_bcˊ、U_caˊ, the control pwm signal of power switch pipe is obtained by reverse line voltage modulated. Comprise the following steps that：

Step 4-1, line voltage U_abˊ correspondence switch conduction times are in each switch periods, to make the area of PWM ripples S_dWith sine wave area S_lEqual, area represents power switch pipe VT1, VT6 conducting to be positive, and area is that negative indication VT3, VT4 are led Logical, power switch pipe ON time is designated as t_ab, then have：

S_l=U '_ab(k)T (13)

S_d=V_dct_ab (14)

Wherein, T is switch periods, U_abˊ (k) is the line voltage modulated signal between k moment A, B phases, V_dcFor DC side electricity Pressure.

Step 4-2, line voltage U_abˊ replaces with U respectively_bcˊ and U_caˊ, U can be tried to achieve by substituting into formula (15)_bcˊ and U_caˊ pairs The switch conduction times t answered_bcAnd t_ca。

Step 4-3, by phase by the waveform of each cycle neutral voltage be divided into 6 regions, i.e. the firstth area, the secondth area, 3rd area, the 4th area, the 5th area and the 6th area, are illustrated to the switch combination in each region and ON time.(definition is opened Off status P is open-minded for the upper bridge arm of inverter, and on off state O is open-minded for the lower bridge arm of inverter)

Step 4-3-1, in the firstth area, be easy to get t_ab>0、t_bc>0、t_ca<0, i.e. power switch pipe VT1, VT6 is simultaneously turned on t_ab, VT2, VT3 simultaneously turn on t_bc, VT1, VT2 simultaneously turn on t_ca.2 switching tubes can not be simultaneously turned on above and below same bridge arm, and Always there is the work of 3 switching tubes at each moment, therefore switch combination is VT1, VT6, VT2 (POO) while opening t_ab, VT1, VT3, VT2 (PPO) is while open t_bc, remaining time T-t_ab-t_bcWith switch combination VT4, VT6, VT2 (OOO) or VT1, VT3, VT5 (PPP) supplement.

Step 4-3-2, in the secondth area, be easy to get t_ab<0、t_bc>0、t_ca<0, i.e. power switch pipe VT3, VT4 is simultaneously turned on t_ab, VT2, VT3 simultaneously turn on t_bc, VT1, VT2 simultaneously turn on t_ca.Therefore, switch combination is that VT1, VT3, VT2 (PPO) are opened simultaneously Logical t_ca, VT4, VT3, VT2 (OPO) are while open t_ab, remaining time T-t_ab-t_caWith switch combination VT4, VT6, VT2 (OOO) or Person VT1, VT3, VT5 (PPP) are supplemented.

Step 4-3-3, in the 3rd area, be easy to get t_ab<0、t_bc>0、t_ca>0, i.e. power switch pipe VT1, VT6 is simultaneously turned on t_ab, VT2, VT3 simultaneously turn on t_bc, VT1, VT2 simultaneously turn on t_ca.Therefore, switch combination is that VT4, VT3, VT2 (OPO) are opened simultaneously Logical t_bc, VT4, VT3, VT5 (OPP) are while open t_ca, remaining time T-t_ab-t_bcWith switch combination VT4, VT6, VT2 (OOO) or Person VT1, VT3, VT5 (PPP) are supplemented.

Step 4-3-4, in the 4th area, be easy to get t_ab<0、t_bc<0、t_ca>0, i.e. power switch pipe VT4, VT3 is simultaneously turned on t_ab, VT6, VT5 simultaneously turn on t_bc, VT5, VT4 simultaneously turn on t_ca.Therefore, switch combination is that VT4, VT3, VT5 (OPO) are opened simultaneously Logical t_ab, VT4, VT6, VT5 (OOP) are while open t_bc, remaining time T-t_ab-t_bcWith switch combination VT4, VT6, VT2 (OOO) or Person VT1, VT3, VT5 (PPP) are supplemented.

Step 4-3-5, in the 5th area, be easy to get t_ab<0、t_bc<0、t_ca>0, i.e. power switch pipe VT1, VT6 is simultaneously turned on t_ab, VT6, VT5 simultaneously turn on t_bc, VT5, VT4 simultaneously turn on t_ca.Therefore, switch combination is that VT1, VT6, VT5 (POP) are opened simultaneously Logical t_ab, VT4, VT6, VT5 (OOP) are while open t_ca, remaining time T-t_ab-t_caWith switch combination VT4, VT6, VT2 (OOO) or Person VT1, VT3, VT5 (PPP) are supplemented.

Step 4-3-6, in the 6th area, be easy to get t_ab<0、t_bc<0、t_ca>0, i.e. power switch pipe VT1, VT6 is simultaneously turned on t_ab, VT6, VT5 simultaneously turn on t_bc, VT2, VT1 simultaneously turn on t_ca.Therefore, switch combination is that VT1, VT6, VT5 (POP) are opened simultaneously Logical t_bc, VT1, VT6, VT2 (POO) are while open t_ca, remaining time T-t_bc-t_caWith switch combination VT4, VT6, VT2 (OOO) or Person VT1, VT3, VT5 (PPP) are supplemented.

Step 4-4, when have two switch combination time sums be more than switch periods when, i.e. T-t_ab-t_bc<0, take

Step 5, the time data in 6 regions obtained by calculating by data conversion sent into ratio in the PWM to DSP Compared with register, the relation of the value in the value and comparand register of the counter inside PWM module, by comparative result send to PWM module, generation pwm pulse delivers to driving and power amplification unit triggers each power switch pipe.

Step 6, repeat step 2 to step 5 obtain the pseudo- gradient vector and line voltage modulated signal of subsequent time, then pass through Reverse line voltage modulated obtains pwm switching signal, until obtaining preferable output voltage.

Solving pseudo- gradient vector estimateIdentification criterion function in, for the inverter output voltage number of collection According to E (k+1) selection, the validity of data is determined using intelligent coordinated analytic approach, i.e., whether is met so that system tends towards stability Requirement.The intelligent determination method of E (k+1) data is as follows：

Δ E=φ Δ U-P Δs e is substituted into formula (11), σ (k+1)=E is made^*(k+1)-E (k+1) can be obtained,

Construct energy function

In formula, σ (k+1) is the difference of inverter actual output voltage and reference voltage, and L is positive definite real symmetric matrix.So After filter out and can makeFor the output data of negative definite.

When having inverter to be incorporated to or cutting off, the initial value of pseudo- gradient vector is reacquired according to traditional droop control method φ (1) and incidence matrix P.

In data transmission procedure, when there is shortage of data, delay even mistake, it can be carried out by existing data Estimate and made up.It is following (by j-th of inverter exemplified by the data that the r moment transmits break down) that data estimate scheme：

In the data drive control module of i-th of inverter, existing previous moment r-1 output data E is utilized_j (r-1) E is estimated_j(r) value.Because the model that each inverter is obtained by the linearisation of model-free adaption algorithm is approximate, institute Also approximate with the change of pseudo- gradient vector, the pseudo- gradient vector by the use of i-th of inverter r moment is used as j-th inverter Pseudo- gradient vector, solves E_j(r)。

By

Obtain the estimation output voltage of j-th of inverter of r moment

In formula,For the pseudo- gradient vector at i-th of inverter r moment, P is incidence matrix, and Δ e (r) is the r moment itself The difference of inverter output voltage and other inverters.

Beneficial effects of the present invention are that instant invention overcomes the Unmarried pregnancy that conventional inverter control method for parallel is present The problems such as, balance the bearing power distribution between each inverter.Simultaneously, it is to avoid laod unbalance, the midpoint of DC voltage The problem of can not find, even if being that can reach balance according to required line voltage so as to realize in the case of laod unbalance State.

Brief description of the drawings

Fig. 1 is data driven type microgrid inverter parallel-connection structure block diagram of the invention.

Fig. 2 is the intelligent data drive control block diagram with reverse line voltage pwm maker of the invention.

Fig. 3 is specific data-driven module control block diagram of the invention.

Fig. 4 is implementation process figure of the invention.

Fig. 5 is the implementation process figure of the MFA control algorithm of the present invention.

Fig. 6 is coordinate transform schematic diagram of the invention.

Fig. 7 is the discrete waveform schematic diagram of the three-phase line voltage of the present invention.

Fig. 8 is the enlarged drawing of first area of the present invention.

Fig. 9 is line voltage S in first area of the present invention_ab=S_dThe PWM waveform of isoboles and correspondence switching tube.

Figure 10 is inverter three-phase line voltage output waveform of the invention.

Figure 11 is inverter line voltage U of the invention_ab、U_bcOutput waveform.

Embodiment

The driving microgrid inverter of distributed data proposed by the present invention intelligent control method in parallel is applied to 3 In the parallel system of 30kW inverters, as illustrated in fig. 1 and 2, micro-capacitance alternating current bus voltage is 380V, each circuit control device All use and propose data-driven intelligent control method, control block diagram as shown in figure 3, and each inverter is all total by communication Line carries out data sharing.Specific control is realized that step is as follows according to Fig. 4 and Fig. 5：

Step 1, the control parameter initial value for obtaining according to droop control method each inverter, i.e.,：Produce PWM control arteries and veins The modulating wave U of the line voltage form of punching_ab、U_bc、U_caWith output three-phase line voltage E_ab、E_bc、E_ca, by dsp chip TMS320F28377D A/D sampling modules are converted to data signal, and sample frequency is 10kHz, and 3s/2r conversion is carried out in dsp The DC voltage U under dq coordinate systems is obtained afterwards_d、U_qAnd E_d、E_q, pseudo- gradient vector φ is obtained according to the inearized model of inverter (1) incidence matrix P.

Step 1-1, the inputoutput data progress coordinate transform using formula (1) and (2) to collecting.To stablize the moment Exemplified by t=0.08s, θ=wt=2 π × 50 × 0.08=8 π are obtained：

Step 1-2, the initial value φ (1) for obtaining according to the inearized model of single inverter pseudo- gradient vector.With reference to public affairs Formula (3)-(5) and by step 1-1 coordinate transform inputoutput datas, take different time datas by 30 computings and then ask flat , so as to obtain pseudo- gradient vector initial value：

Step 1-3, incidence matrix P calculated according to the inearized model of inverter parallel.

The pseudo- gradient vector initial value φ (1) of gained is substituted into inverter parallel system inearized model

Make equal sign both sides each element of matrix equal to try to achieve incidence matrix P.

Step 2, the pseudo- gradient vector initial value obtained using step 1 utilize the input of inverter parallel system as initial point Output data, criterion function is recognized according to by minimization：

J (φ (k))=| | E^*(k+1)-E(k+1)||²+μ||φ(k)-φ(k-1)||² (26)

In formula, output line voltage virtual value reference value is set to E^*(k+1)=380V, E (k+1) are that inverter output voltage has The actual measured value of valid value, μ=1.2.

Obtain pseudo- gradient vector estimateMore new formula：

So as to the estimate of the pseudo- gradient vector of online updating

In formula, k=20 is taken, η=1 is tried to achieve according to the inputoutput data of previous moment and pseudo- gradient vector estimate

Step 3, the pseudo- gradient vector estimate obtained according to step 2Calculate the three-phase line voltage for producing pwm signal Modulating wave, passes through minimization Controlling object function：

J (U (k))=| | E^*(k+1)-E(k+1)||²+λ||U(k)-U(k-1)||² (29)

So as to obtain inputting modulating wave：

ρ=1, λ=1.5 in formula,WithBe decomposition of the inverter output reference voltage under dq coordinate systems andE_dAnd E_qIt is that inverter exports decomposition of the measurement voltage under dq coordinate systems, and E during k=19_d= 310.2117, E_q=-0.4666.

Then

Step 4, the input modulating signal U (k) that is obtained in step 3 by 2r/3s coordinate transforms to three phase static abc seats Mark is line voltage form U that is lower and being converted into_abˊ、U_bcˊ、U_caˊ, as shown in fig. 6,

The control pwm signal of power switch pipe is obtained by reverse line voltage modulated.Comprise the following steps that：

Step 4-1, line voltage U_abˊ correspondence switch conduction times are in each switch periods, to make the area of PWM ripples S_dWith sine wave area S_lEqual, area represents power switch pipe VT1, VT6 conducting to be positive, and area is that negative indication VT3, VT4 are led Logical, power switch pipe ON time is designated as t_ab, then have：

S_l=U '_ab(k)T (33)

S_d=V_dct_ab (34)

In formula, T=0.02s is switch periods, V_dc=600V is DC voltage.

Step 4-2, line voltage U_abˊ replaces with U respectively_bcˊ and U_caˊ, U can be tried to achieve by substituting into formula (15)_bcˊ and U_caˊ pairs The switch conduction times t answered_bc=0.0963T and t_ca=0.4959T.

Step 4-3, by phase the waveform of each cycle neutral voltage is divided into 6 regions, as shown in fig. 7, i.e. first Area, the secondth area, the 3rd area, the 4th area, the 5th area and the 6th area, say to the switch combination in each region and ON time It is bright.(defining the upper bridge arm that on off state P is inverter open-minded, on off state O is that the lower bridge arm of inverter is open-minded)

The time size t calculated according to step 4-2_ab<0、t_bc>0、t_ca>0 understands that now line voltage is in the 3rd area, such as Fig. 8 With 9, switch conduction times are solved, power switch pipe VT1, VT6 is obtained and simultaneously turns on t_ab, VT2, VT3 are simultaneously turned on t_bc, VT1, VT2 simultaneously turn on t_ca.Therefore, switch combination is VT4, VT3, VT2 (OPO) while opening t_bc=0.0963T, VT4, VT3, VT5 (OPP) are while open t_ca=0.4959T, remaining time T-t_ab-t_bc=0.4078T switch combination VT4, VT6, VT2 (OOO) or VT1, VT3, VT5 (PPP) are supplemented.

Step 5, the switch conduction times data now obtained by calculating by data conversion are sent into the PWM to DSP In comparand register, the relation of the value in the value and comparand register of the counter inside PWM module, by comparative result Send to PWM module, generation pwm pulse delivers to power amplification unit and triggers each power switch pipe.

Step 6, repeat step 2 to step 5 obtain the pseudo- gradient vector and line voltage modulated signal of subsequent time, then pass through Reverse line voltage modulated obtains pwm switching signal, until obtaining preferably exporting three-phase voltage, as shown in FIG. 10 and 11.

Solving pseudo- gradient vector estimateIdentification criterion function in, for the inverter output voltage number of collection According to E (k+1) selection, the validity of data is determined using intelligent coordinated analytic approach, i.e., whether is met so that system tends towards stability Requirement.The intelligent determination method of E (k+1) data is as follows：

Δ E=φ Δ U-P Δs e is substituted into formula (11), σ (k+1)=E is made^*(k+1)-E (k+1) can be obtained,

Construct energy function

In formula, σ (k+1) is the difference of inverter actual output voltage and reference voltage, and L is positive definite real symmetric matrix.So After filter out and can makeFor the output data of negative definite.

When having inverter to be incorporated to or cutting off, the initial value of pseudo- gradient vector is reacquired according to traditional droop control method φ (1) and incidence matrix P.

In data transmission procedure, when there is shortage of data, delay even mistake, it can be carried out by existing data Estimate and made up.It is following (by the 2nd inverter exemplified by the data that the r moment transmits break down) that data estimate scheme：

In the data drive control module of the 1st inverter, existing previous moment r-1 output data E is utilized₂ (r-1) E is estimated₂(r) value.Because the model that each inverter is obtained by the linearisation of model-free adaption algorithm is approximate, institute Also approximate with the change of pseudo- gradient vector, the pseudo- gradient vector by the use of the 1st inverter r moment is used as the 2nd inverter Pseudo- gradient vector, solves E₂(r)。

By

It can obtain the estimation output voltage of j-th of inverter of r moment

In formula,For the pseudo- gradient vector at the 1st inverter r moment, P is incidence matrix, and Δ e (r) is the r moment itself The difference of inverter output voltage and other inverters.

Claims (8)

1. a kind of micro-capacitance sensor data-driven inverter parallel intelligent control method, it is characterised in that comprise the following steps：

Step 1, the control parameter initial value for obtaining according to droop control method each inverter, produce the line that PWM controls pulse The modulating wave U of voltage form_ab、U_bc、U_caWith output three-phase line voltage E_ab、E_bc、E_ca, dq coordinates are obtained after carrying out 3s/2r conversion DC voltage U under system_d、U_qAnd E_d、E_q, the initial value φ of pseudo- gradient vector is then obtained according to the inearized model of inverter And incidence matrix P (1)；

Step 1-1, the following coordinate transform of inputoutput data progress to collecting：

In formula, C_ZThe direct transform matrix of dq coordinate systems is tied to for abc coordinates, θ is the folder of dq coordinate system d axles and abc coordinate system a axles Angle；

Step 1-2, the initial value φ (1) for obtaining according to the inearized model of single inverter pseudo- gradient vector；

By single inverter inearized model：Δ E=φ Δ U transposition are obtained：

${\mathrm{\&Delta;E}}_{(1\&times;2)}^T={\mathrm{\&Delta;U}}_{(1\&times;2)}^T{\mathrm{\&phi;}}_{(1\&times;2)}^T---\left(3\right)$

The pseudo- gradient vector approximately equal of two adjacent groups data, is obtained when stable because inverter is exported：

$\left(\begin{array}{cc}{\mathrm{\&Delta;E}}_{d1}& {\mathrm{\&Delta;E}}_{q1}\\ {\mathrm{\&Delta;E}}_{d2}& {\mathrm{\&Delta;E}}_{q2}\end{array}\right)=\left(\begin{array}{cc}{\mathrm{\&Delta;U}}_{d1}& {\mathrm{\&Delta;U}}_{q1}\\ {\mathrm{\&Delta;U}}_{d2}& {\mathrm{\&Delta;U}}_{q2}\end{array}\right){\left(\begin{array}{cc}{\mathrm{\&phi;}}_{11}& {\mathrm{\&phi;}}_{12}\\ {\mathrm{\&phi;}}_{21}& {\mathrm{\&phi;}}_{22}\end{array}\right)}^T---\left(4\right)$

So as to obtain pseudo- gradient vector initial value：

$\begin{array}{c}\mathrm{\&phi;}\left(1\right)=\left(\begin{array}{cc}{\mathrm{\&phi;}}_{11}& {\mathrm{\&phi;}}_{12}\\ {\mathrm{\&phi;}}_{21}& {\mathrm{\&phi;}}_{22}\end{array}\right)={\left({\left(\begin{array}{cc}{\mathrm{\&Delta;U}}_{d1}& {\mathrm{\&Delta;U}}_{q1}\\ {\mathrm{\&Delta;U}}_{d2}& {\mathrm{\&Delta;U}}_{q2}\end{array}\right)}^{-1}\left(\begin{array}{cc}{\mathrm{\&Delta;E}}_{d1}& {\mathrm{\&Delta;E}}_{q1}\\ {\mathrm{\&Delta;E}}_{d2}& {\mathrm{\&Delta;E}}_{q2}\end{array}\right)\right)}^T\\ =\left({\left(\begin{array}{cc}{U}_d\left(m+1\right)-U_d\left(m\right)& U_d\left(m\right)-U_d\left(m-1\right)\\ U_q\left(m+1\right)-U_q\left(m\right)& U_q\left(m\right)-U_q\left(m-1\right)\end{array}\right)}^{-1}{\left(\begin{array}{cc}{E}_d\left(m+1\right)-E_d\left(m\right)& E_d\left(m\right)-E_d\left(m-1\right)\\ E_q\left(m+1\right)-E_q\left(m\right)& E_q\left(m\right)-E_q\left(m-1\right)\end{array}\right)}^T\right)\end{array}---\left(5\right)$

In formula, U_d(m-1)、U_d(m)、U_d(m+1) it is respectively value of the line voltage modulating wave in m-1, m and m+1 moment d axle, U_q(m- 1)、U_q(m)、U_q(m+1) it is respectively q axle value of the line voltage modulating wave at m-1, m and m+1 moment, E_d(m-1)、E_d(m)、E_d(m+1) Respectively output voltage is in the d axle values at m-1, m and m+1 moment, E_q(m-1)、E_q(m)、E_q(m+1) it is respectively output voltage in m- 1st, the q axle values at m and m+1 moment；

Step 1-3, incidence matrix P calculated according to the inearized model of inverter parallel；

The pseudo- gradient vector initial value φ (1) of gained is substituted into inverter parallel system inearized model Δ E=φ Δ U-P Δs e：

P Δ e=φ (1) Δ U- Δs E (6)

Make equal sign both sides each element of matrix equal to try to achieve incidence matrix P；

In formula, Δ e is itself inverter output voltage and the difference of other inverters, and Δ U inputs the variable quantity of modulation voltage, Δ The variable quantity of E output voltages；

Step 2, the pseudo- gradient vector initial value φ (1) obtained with step 1 utilize the input of inverter parallel system for initial point Output data, criterion function is recognized by minimization：

J (φ (k))=| | E^*(k+1)-E(k+1)||²+μ||φ(k)-φ(k-1)||² (7)

In formula, E^*(k+1) it is inverter output voltage reference value, E (k+1) is the actual measured value of inverter output voltage, and μ isThe penalty factor of variable quantity；

Obtain pseudo- gradient vector estimateMore new formula：

$\begin{array}{c}\left(\begin{array}{cc}{\widehat{\mathrm{\&phi;}}}_{11}& {\widehat{\mathrm{\&phi;}}}_{12}\\ {\widehat{\mathrm{\&phi;}}}_{21}& {\widehat{\mathrm{\&phi;}}}_{22}\end{array}\right)\left(k\right)=\left(\begin{array}{cc}{\widehat{\mathrm{\&phi;}}}_{11}& {\widehat{\mathrm{\&phi;}}}_{12}\\ {\widehat{\mathrm{\&phi;}}}_{21}& {\widehat{\mathrm{\&phi;}}}_{22}\end{array}\right)\left(k-1\right)+\\ \frac{\mathrm{\&eta;}{\left(\begin{array}{c}{\mathrm{\&Delta;U}}_d\\ {\mathrm{\&Delta;U}}_q\end{array}\right)}^T\left(k-1\right)}{\mathrm{\&mu;}+\left|\right|\left(\begin{array}{c}{\mathrm{\&Delta;U}}_d\\ {\mathrm{\&Delta;U}}_q\end{array}\right)\left(k-1\right)|{|}^2}\left(\left(\begin{array}{c}{\mathrm{\&Delta;E}}_d\\ {\mathrm{\&Delta;E}}_q\end{array}\right)\left(k\right)-\left(\begin{array}{cc}{\widehat{\mathrm{\&phi;}}}_{11}& {\widehat{\mathrm{\&phi;}}}_{12}\\ {\widehat{\mathrm{\&phi;}}}_{21}& {\widehat{\mathrm{\&phi;}}}_{22}\end{array}\right)\left(k-1\right)\left(\begin{array}{c}{\mathrm{\&Delta;U}}_d\\ {\mathrm{\&Delta;U}}_q\end{array}\right)\left(k-1\right)-P\mathrm{\&Delta;}e\left(k\right)\right)\end{array}---\left(8\right)$

So as to the estimate of the pseudo- gradient vector of online updating

In formula, and η ∈ (0,2], Δ U_dWith Δ U_qIt is the variable quantity that the k-1 moment inputs modulation voltage under dq coordinate systems, Δ E respectively_dWith ΔE_qIt is the variable quantity of k moment output voltages under dq coordinate systems respectively；

Step 3, the pseudo- gradient vector estimate obtained according to step 2Calculate the three-phase line voltage modulation for producing pwm signal Ripple, passes through minimization Controlling object function：

J (U (k))=| | E^*(k+1)-E(k+1)||²+λ||U(k)-U(k-1)||² (9)

So as to obtain inputting modulating wave：

$\left(\begin{array}{c}{U}_d\\ U_q\end{array}\right)\left(k\right)=\left(\begin{array}{c}{U}_d\\ U_q\end{array}\right)(k-1)+\frac{\mathrm{\&rho;}{\left(\begin{array}{cc}{\widehat{\mathrm{\&phi;}}}_{11}& {\widehat{\mathrm{\&phi;}}}_{12}\\ {\widehat{\mathrm{\&phi;}}}_{21}& {\widehat{\mathrm{\&phi;}}}_{22}\end{array}\right)}^T\left(k\right)}{\mathrm{\&lambda;}+\left|\right|\left(\begin{array}{cc}{\widehat{\mathrm{\&phi;}}}_{11}& {\widehat{\mathrm{\&phi;}}}_{12}\\ {\widehat{\mathrm{\&phi;}}}_{21}& {\widehat{\mathrm{\&phi;}}}_{22}\end{array}\right)\left(k\right)|{|}^2}\left(\left(\begin{array}{c}{E}_d^{*}\\ E_q^{*}\end{array}\right)\right(k+1)-\left(\begin{array}{c}{E}_d\\ E_q\end{array}\right)(k)-P\mathrm{\&Delta;}e(k\left)\right)---\left(10\right)$

ρ ∈ in formula (0,2], λ is the penalty factor of U (k) variable quantities, and P, Δ e (k) are identical with formula (6)；

Step 4, the input modulating signal U (k) that is obtained in step 3 by 2r/3s coordinates is transformed to three phase static abc coordinate systems Under line voltage form U_abˊ、U_bcˊ、U_caˊ, the control pwm signal of power switch pipe is obtained by reverse line voltage modulated；Specifically Step is as follows：

Step 4-1, line voltage U_abˊ correspondence switch conduction times are in each switch periods, to make the area S of PWM ripples_dWith Sine wave area S_lEqual, area represents power switch pipe VT1, VT6 conducting to be positive, and area turns on for negative indication VT3, VT4, work( Rate switching tube ON time is designated as t_ab, then have：

S_l=U '_ab(k)T (11)

S_d=V_dct_ab (12)

$t_{ab}=\frac{U_{ab}^{\&prime;}\left(k\right)}{V_{dc}}T---\left(13\right)$

Wherein, T is switch periods, U_abˊ (k) is the line voltage modulated signal between k moment A, B phases, V_dcFor DC voltage；

Step 4-2, line voltage U_abˊ replaces with U respectively_bcˊ and U_caˊ, U can be tried to achieve by substituting into formula (15)_bcˊ and U_caˊ is corresponding to be opened Close ON time t_bcAnd t_ca；

Step 4-3, by phase the waveform of each cycle neutral voltage is divided into 6 regions, i.e. the firstth area, the secondth area, the 3rd Area, the 4th area, the 5th area and the 6th area, calculate the switch combination in each region and ON time；

Step 5, the time data in 6 regions obtained by calculating by data conversion sent into relatively posting in the PWM to DSP Storage, the relation of the value in the value and comparand register of the counter inside PWM module sends comparative result to PWM Module, generation pwm pulse delivers to driving and power amplification unit triggers each power switch pipe；

Step 6, repeat step 2 to step 5 obtain the pseudo- gradient vector and line voltage modulated signal of subsequent time, then by reverse Line voltage modulation obtains pwm switching signal, until obtaining preferable output voltage.

2. control method according to claim 1, it is characterised in that the switch in each region is calculated described in step 4-3 Combination and the specific method of ON time are：Define on off state P open-minded for the upper bridge arm of inverter, on off state O is inversion The lower bridge arm of device is open-minded；

1) in the firstth area, t_ab>0、t_bc>0、t_ca<0, i.e. power switch pipe VT1, VT6 simultaneously turns on t_ab, VT2, VT3 are led simultaneously Logical t_bc, VT1, VT2 simultaneously turn on t_ca；2 switching tubes can not be simultaneously turned on above and below same bridge arm, and each moment always has 3 to open It is respectively P, O, O while opening t for VT1, VT6, VT2 on off state to close pipe work, i.e. switch combination_ab, VT1, VT3, VT2's On off state is respectively P, P, O while opening t_bc, remaining time T-t_ab-t_bcWith switch combination VT4, VT6, VT2 on off state Respectively O, O, O or VT1, VT3, VT5 on off state are respectively P, P, P supplement；

When there is two switch combination time sums to be more than switch periods, i.e. T-t_ab-t_bc<0, then take

$\begin{array}{c}{t_{ab}}^{\&prime;}=\frac{t_{ab}}{t_{ab}+t_{bc}}T\\ {t_{bc}}^{\&prime;}=\frac{t_{bc}}{t_{ab}+t_{bc}}T\end{array}---\left(14\right)$

2) repeat step 1) obtain switch combination and the ON time in 2-6 regions.

3. control method according to claim 1 or 2, it is characterised in that the pseudo- gradient vector estimation of solution described in step 2 ValueIdentification criterion function in, for the inverter output voltage data E (k+1) of collection selection, using intelligent coordinated Analytic approach determines the validity of data：

Δ E=φ Δ U-P Δs e is substituted into formula (11), σ (k+1)=E is made^*(k+1)-E (k+1) can be obtained,

$\mathrm{\&sigma;}(k+1)=(\frac{\mathrm{\&rho;}}{\mathrm{\&lambda;}+\left|\right|\widehat{\mathrm{\&phi;}}\left(k\right)|{|}^2}{\left({\widehat{\mathrm{\&phi;}}}^T\right(k\left)\right)}^{-1}-\widehat{\mathrm{\&phi;}}(k\left)\right)\mathrm{\&Delta;}U\left(k\right)-P\mathrm{\&Delta;}e\left(k\right)---\left(15\right)$

Construct energy function

$V\left(\mathrm{\&sigma;}\right)=\frac{1}{2}{\mathrm{\&sigma;}}^{T}L\mathrm{\&sigma;}---\left(16\right)$

In formula, σ (k+1) is the difference of inverter actual output voltage and reference voltage, and L is positive definite real symmetric matrix；Then sieve Selecting can make V (σ) derivative be negative output data.

4. control method according to claim 1 or 2, it is characterised in that when thering is inverter to be incorporated to or cutting off, step 1, The initial value φ (1) and incidence matrix P of pseudo- gradient vector are reacquired according to droop control method.

5. control method according to claim 3, it is characterised in that when thering is inverter to be incorporated to or cutting off, step 1, root The initial value φ (1) and incidence matrix P of pseudo- gradient vector are reacquired according to droop control method.

6. the control method according to claim 1 or 2 or 5, in data transmission procedure, when there is shortage of data, delay When even wrong, i.e., it can not obtain the output voltage data E of inverter parallel system so that step 3 can not be carried out, then by Some data are estimated：

If the output voltage data E that j-th of inverter is transmitted at the r moment_j(r) break down, in the data of i-th of inverter In drive control module, existing previous moment r-1 output data E is utilized_j(r-1) E is estimated_j(r) value；According in step 3 Formula (10) solve E_j(r)；

${\mathrm{\&Delta;U}}_j(r-1)=\frac{\mathrm{\&rho;}{\widehat{\mathrm{\&phi;}}}_i^T(r-1)}{\mathrm{\&lambda;}+\left|\right|{\widehat{\mathrm{\&phi;}}}_i(r-1)|{|}^2}\left({E_j}^{*}\right(r)-E_j(r-1)-P\mathrm{\&Delta;}e(r-1\left)\right)---\left(19\right)$

By inverter parallel system inearized modelObtain the r moment The estimation output voltage of j inverter：

$\begin{array}{c}{E}_j\left(r\right)=E_j(r-1)+{\mathrm{\&Delta;E}}_j\left(r\right)-P\mathrm{\&Delta;}e(r-1)\\ =E_j\left(r-1\right)+{\widehat{\mathrm{\&phi;}}}_i(r-1)\mathrm{\&Delta;}U\left(r-1\right)-P\mathrm{\&Delta;}e(r-1)\end{array}---\left(20\right)$

In formula,For the pseudo- gradient vector estimate at i-th of inverter r-1 moment, P is incidence matrix, and Δ e (r) is r The difference of moment itself inverter output voltage and other inverters.

7. control method according to claim 3, even wrong when there is shortage of data, delay in data transmission procedure Mistake, i.e., can not obtain the output voltage data E of inverter parallel system so that step 3 can not be carried out, then pass through existing number According to being estimated：

If the output voltage data E that j-th of inverter is transmitted at the r moment_j(r) break down, in the data of i-th of inverter In drive control module, existing previous moment r-1 output data E is utilized_j(r-1) E is estimated_j(r) value；According in step 3 Formula (10) solve E_j(r)；

${\mathrm{\&Delta;U}}_j(r-1)=\frac{\mathrm{\&rho;}{\widehat{\mathrm{\&phi;}}}_i^T(r-1)}{\mathrm{\&lambda;}+\left|\right|{\widehat{\mathrm{\&phi;}}}_i(r-1)|{|}^2}\left({E_j}^{*}\right(r)-E_j(r-1)-P\mathrm{\&Delta;}e(r-1\left)\right)---\left(19\right)$

By inverter parallel system inearized modelObtain the r moment The estimation output voltage of j inverter：

$\begin{array}{c}{E}_j\left(r\right)=E_j(r-1)+{\mathrm{\&Delta;E}}_j\left(r\right)-P\mathrm{\&Delta;}e(r-1)\\ =E_j(r-1)+{\widehat{\mathrm{\&phi;}}}_i\left(r-1\right)\mathrm{\&Delta;}U\left(r-1\right)-P\mathrm{\&Delta;}e(r-1)\end{array}---\left(20\right)$

In formula,For the pseudo- gradient vector estimate at i-th of inverter r-1 moment, P is incidence matrix, and Δ e (r) is r The difference of moment itself inverter output voltage and other inverters.

8. control method according to claim 4, even wrong when there is shortage of data, delay in data transmission procedure Mistake, i.e., can not obtain the output voltage data E of inverter parallel system so that step 3 can not be carried out, then pass through existing number According to being estimated：

If the output voltage data E that j-th of inverter is transmitted at the r moment_j(r) break down, in the data of i-th of inverter In drive control module, existing previous moment r-1 output data E is utilized_j(r-1) E is estimated_j(r) value；According in step 3 Formula (10) solve E_j(r)；

${\mathrm{\&Delta;U}}_j(r-1)=\frac{\mathrm{\&rho;}{\widehat{\mathrm{\&phi;}}}_i^T(r-1)}{\mathrm{\&lambda;}+\left|\right|{\widehat{\mathrm{\&phi;}}}_i(r-1)|{|}^2}\left({E_j}^{*}\right(r)-E_j(r-1)-P\mathrm{\&Delta;}e(r-1\left)\right)---\left(19\right)$

By inverter parallel system inearized modelObtain the r moment The estimation output voltage of j inverter：

$\begin{array}{c}{E}_j\left(r\right)=E_j(r-1)+{\mathrm{\&Delta;E}}_j\left(r\right)-P\mathrm{\&Delta;}e(r-1)\\ =E_j(r-1)+{\widehat{\mathrm{\&phi;}}}_i\left(r-1\right)\mathrm{\&Delta;}U\left(r-1\right)-P\mathrm{\&Delta;}e(r-1)\end{array}---\left(20\right)$

In formula,For the pseudo- gradient vector estimate at i-th of inverter r-1 moment, P is incidence matrix, and Δ e (r) is r The difference of moment itself inverter output voltage and other inverters.