Inverter control method based on feature recognition model

Abstract

The invention discloses an inverter control method based on a feature recognition model. The method comprises the following steps: acquiring an input voltage VPV, an input current IPV, an environment temperature Ttemp, a bus voltage Vbus and a real-time power Pac in real time; performing an inverter time-sharing control strategy; constructing a power prediction model for soft start in an output capability intensifying period; main influence factors on power fluctuation are extracted through principal component analysis, and power corresponding to power is output through an inverter efficiency equivalent model, a photovoltaic cell equivalent model and a boost circuit mathematical model; and performing active shutdown in an output capability weakening period. According to the method, main influence factors on power fluctuation can be extracted through principal component analysis, so that power reference is output, the adjustment requirement of a back-end control strategy is met, and the power generation efficiency is remarkably improved. According to the invention, soft start in an output intensifying period and shutdown protection control requirements in an output weakening period are met, frequent start and stop are basically avoided, the damage risk of switching elements is reduced, and the operation control of the inverter is more reliable and stable.

Description

Technical Field

[0001] The invention relates to an inverter control method based on a feature recognition model, and belongs to the technical field of inverter control. Background Art

[0002] As the mainstream equipment in industrial and commercial power stations, high-power string inverters integrate ACSPS and DCSPS to ensure that the equipment does not stop. The advantage is that it is conducive to the monitoring of equipment and power stations, but it will conflict with the traditional control logic. For example, for a T-type three-level topology string inverter, when the PV energy is low, the MPPT tracking accuracy will be affected due to sampling accuracy limitations and device losses. On the one hand, it will cause bus instability and cause overmodulation. On the other hand, it will also cause the divergence of the neutral point balance controller, resulting in neutral point imbalance, which will have a great impact on the life of the bus capacitor and the stability of the switching device. Summary of the invention

[0003] The purpose of the present invention is to solve the deficiencies of the above-mentioned prior art. In view of the fact that the output characteristics of photovoltaic cells of string inverters are very soft under extremely low input power conditions, which leads to poor MPPT accuracy, and further causes bus instability and neutral point control divergence, an inverter control method based on a feature recognition model is proposed.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] The inverter control method based on the feature recognition model includes the following steps:

[0006] S1 data acquisition, real-time acquisition of input voltage V PV 、Input current I PV 、Ambient temperature T temp , bus voltage V bus , real-time power P ac ;

[0007] S2 according to the input voltage V PV Change trend, slope CIK, input voltage V PV The fluctuation difference CIvar is used to implement the inverter time-sharing control strategy;

[0008] Output capability becomes stronger, input voltage V PV When the trend is upward and CIK>0;

[0009] Output plateau period, input voltage V PV Stable and CIvar tends to 0;

[0010] Output capacity weakening period, input voltage V PV When the trend is down and CIK is less than 0;

[0011] During the period when S3 output capacity becomes stronger, a power prediction model is constructed to perform soft start;

[0012] S4 uses principal component analysis to extract the main factors affecting power fluctuations;

[0013] When the main influencing factor is the input voltage V PV 、Input current I PV When the inverter efficiency equivalent model is used, the output power P ref1 ; When the main influencing factor is the input voltage V PV 、Ambient temperature T temp When the photovoltaic cell equivalent model is used, the output power P ref2 ; When the main influencing factor is the input voltage V PV , bus voltage V bus When the boost circuit mathematical model is used, the output power P ref3 ;

[0014] S5 automatically shuts down during the period of output capacity decline.

[0015] Preferably, in step S4, a data set X is generated, where X={V PV , I PV , V bus , T temp}, find the mean CIave and standard deviation CIvar of each data vector, the sampling average

[0016] Construct the covariance matrix V: n is the number of sampling points;

[0017] Solve for the eigenvalue λ of V i and the eigenvector ω i ;

[0018] Variance Contribution and cumulative variance contribution As the evaluation criteria, the eigenvector corresponding to the cumulative variance contribution rate of more than 80% was selected as the principal component;

[0019] The eigenvalue is Λ=[λ 1 ,λ 2 ,λ 3 , ..., λ m ] corresponds to the eigenvector W m ,

[0020] W m =[ω 1 ,ω 2 ,ω 3 ,......,ω m ]T , m<4, principal component

[0021] Principal component reconstruction dataset

[0022] Preferably, in step S4, when the inverter efficiency equivalent model is output;

[0023] P ref1 =E AC =E DC ·η=V DC I DC η;

[0024] Among them, E AC is the inverter output power; E DC is the DC input power; η is the comprehensive conversion efficiency.

[0025] Preferably, in step S4, when the photovoltaic cell equivalent model is output;

[0026]

[0027] Where, I is the output current; V is the output voltage; I pv is the photocurrent; I o Saturation current; R s Series equivalent resistance; R p Shunt resistance; n ideality factor; k Boltzmann constant; T ambient temperature; T n Rated temperature; q charge element; Eg 0 Semiconductor band gap energy; N s Number of modules connected in series;

[0028] P ref2 =V·I.

[0029] Preferably, in step S4, when the boost circuit mathematical model is output;

[0030]

[0031] V busref -10V≤V bus ≤V busref +10V;

[0032] Where D is the duty cycle of the boost circuit, Va, Vb, and Vc are the three-phase voltage peaks of the three-phase grid;

[0033] P ref3 =max(V PV I PV ).

[0034] Preferably, in the step S2, at power P ref1 , power P ref2、 Power P ref3 When output respectively, after performing difference processing with the real-time power P ac , power control correction is performed to output a control pulse.

[0035] Preferably, in the step S2, a data set X is generated, X = {VPV, IPV, Vbus, Ttemp};

[0036] Slope

[0037] The maximum value data set DImax = max{X 1 , X 2 ,......, X n};

[0038] The minimum value data set CImin = min{X 1 , X 2 ,......, X n};

[0039] Average value

[0040] Standard deviation

[0041] First derivative

[0042] Second derivative

[0043] is the sampling average value, and n is the number of sampling points.

[0044] Preferably, in the step S3, the energy that the PV battery can provide is predicted through the capacity of the bus capacitor:

[0045]

[0046] E C is the capacity that the bus capacitor can store, C is the capacitance value, and U is the bus voltage;

[0047] Set the grid connection threshold Pb. When E C > Pb, the inverter is allowed to self-check and connect to the grid.

[0048] Preferably, the power of the photovoltaic cell Ppv is calculated according to Vpv and Ipv. When Ppv < P1, enter the active shutdown procedure;

[0049] P1 = PR + Paux + Pfan + Prly + Pboost, where PR is the module loss, Paux is the sps power, Pfan is the fan power, Prly is the relay power, and Pboost is the boost circuit switching loss;

[0050] When Vpv < a1, increase the dummy load boost_duty to simulate the increase of the dummy load, and continuously monitor Vpv and Ipv;

[0051] When Vpv < a2, increasing the dummy load boost_duty can no longer maintain the stability of the BUS voltage, then it is determined that the photovoltaic output is weak, actively disconnect from the grid, and disconnect the relay; the dummy load consumes the photovoltaic power;

[0052] When Vpv < a3, gradually unload the dummy load until all loads are completely cut off.

[0053] Among them, a1 is the minimum operating voltage of the inverter boost, a2 is the grid-connected voltage under the minimum power consumption of the inverter, a3 is the starting voltage of the switching power supply, and a1 > a2 > a3.

[0054] The beneficial effects of the present invention are mainly reflected in:

[0055] 1. It can achieve principal component analysis to extract the main influencing factors of power fluctuations, thereby outputting power references to meet the needs of the subsequent control strategy adjustment, and significantly improving the power generation efficiency.

[0056] 2. It meets the soft start requirements during the output strengthening period and the shutdown protection control requirements during the output weakening period, basically eliminating frequent starts and stops, reducing the damage risk of switching components, and making the inverter operation control more reliable and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives, and advantages of the present application will become more apparent:

[0058] Figure 1 is a schematic flow chart of the inverter control method based on the feature recognition model of the present invention.

[0059] Figure 2 is a schematic structural diagram of the control module in the inverter control method based on the feature recognition model of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the inventions. It should also be noted that, for ease of description, only the parts related to the relevant inventions are shown in the accompanying drawings. It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0062] The present invention provides an inverter control method based on a feature recognition model, such as Figure 1 As shown, the following steps are included:

[0063] Data acquisition, real-time acquisition of input voltage V PV 、Input current I PV 、Ambient temperature T temp , bus voltage V bus , real-time power P ac .

[0064] According to the input voltage V PV Change trend, slope CIK, input voltage V PV The inverter time-sharing control strategy is implemented based on the fluctuation difference CIvar.

[0065] Output capability becomes stronger, input voltage V PV When the trend is upward and CIK>0;

[0066] Output plateau period, input voltage V PV Stable and CIvar tends to 0;

[0067] Output capacity weakening period, input voltage V PV When the trend is down and CIK is less than 0;

[0068] A power prediction model is constructed to perform soft start during the period when the output capacity becomes stronger.

[0069] The principal component analysis is used to extract the main influencing factors on power fluctuation; when the main influencing factor is the input voltage V PV 、Input current I PV When the inverter efficiency equivalent model is used, the output power P ref1 ; When the main influencing factor is the input voltage V PV、Ambient temperature T temp When the photovoltaic cell equivalent model is used, the output power P ref2 ; When the main influencing factor is the input voltage V PV , bus voltage V bus When the boost circuit mathematical model is used, the output power P ref3 .

[0070] Active shutdown is performed during the period of output capacity weakening.

[0071] Specific implementation process and principle description:

[0072] Inverter data collection is performed in real time, and an inverter time-sharing control strategy is adopted according to the PV voltage change trend. The time-sharing control strategy generally belongs to the existing technology, and only the solutions that meet the time-sharing control strategy are within the protection scope of this case.

[0073] Generally speaking, the inverter time-sharing strategy can distinguish between the morning when the photovoltaic capacity changes from weak to strong, the daytime when the photovoltaic output is stable, and the evening when the photovoltaic capacity changes from strong to weak. As long as the time-sharing control of the photovoltaic output can be achieved, it will be fine.

[0074] In this case, the slope CIK and input voltage V PV The fluctuation difference CIvar is used to achieve the distinction, wherein CIvar tending to 0 is the prior art in this field. Generally, a standard deviation threshold is set, and within the standard deviation threshold range, it tends to 0, which will not be elaborated here.

[0075] During the daytime photovoltaic power generation process, photovoltaics will be affected by many factors such as cloudy weather, ambient temperature, and load, resulting in unstable power generation efficiency. In response to this situation, this case can realize the principal component analysis that affects power changes and extract the principal components that have a greater impact on power fluctuations. The corresponding model output power is then carried out according to the main influencing factors, and the back-end control pulse reference is realized, thereby maximizing the power generation efficiency.

[0076] More specifically, when Vpv and Vbus are the main influencing factors, the response rate can be dynamically adjusted by adjusting the PI parameters of the boost loop. When Vpv and IPV are the main influencing factors, it means that the maximum power point fluctuates due to cloudy weather. The MPPT step size and disturbance direction can be dynamically adjusted to quickly track the maximum power point. When Vpv and T are the main influencing factors, it means that the ambient temperature is too high, the power is limited, and the internal fan is working. At this time, the power is limited by the inverter fixed power and supports a certain degree of over-matching. At this time, the MPPT side must balance the current of each channel to ensure the uniformity of the module temperature.

[0077] In a specific embodiment, when performing principal component analysis, a data set X is generated, X={V PV , I PV , Vbus , T temp}, find the mean CIave and standard deviation CIvar of each data vector, the sampling average

[0078] Construct the covariance matrix V: n is the number of sampling points.

[0079] Solve for the eigenvalue λ of V i and the eigenvector ω i .

[0080] Variance Contribution and cumulative variance contribution As an evaluation criterion, the eigenvector corresponding to the cumulative variance contribution rate of more than 80% is selected as the principal component.

[0081] The eigenvalue is Λ=[λ 1 ,λ 2 ,λ 3 , ..., λ m ] corresponds to the eigenvector W m .

[0082] W m =[ω 1 ,ω 2 ,ω 3 ,......,ω m ] T , m<4, principal component

[0083] Principal component reconstruction dataset

[0084] The inverter efficiency equivalent model, photovoltaic cell equivalent model, and boost circuit mathematical model are described in detail:

[0085] When the inverter efficiency equivalent model is output;

[0086] P ref1 =E AC =E DC ·η=V DC I DC η;

[0087] Among them, E AC is the inverter output power; E DC is the DC input power; η is the comprehensive conversion efficiency.

[0088] When the photovoltaic cell equivalent model is output;

[0089]

[0090] Where, I is the output current; V is the output voltage; I pv is the photocurrent; I o Saturation current; R s Series equivalent resistance; R p Shunt resistance; n ideality factor; k Boltzmann constant; T ambient temperature; T n Rated temperature; q charge element; Eg 0 Semiconductor band gap energy; N s Number of modules connected in series;

[0091] P ref2 =V·I.

[0092] Preferably, in step S4, when the boost circuit mathematical model is output;

[0093]

[0094] V busref -10V≤V bus ≤V busref +10V;

[0095] Where D is the duty cycle of the boost circuit, Va, Vb, and Vc are the three-phase voltage peaks of the three-phase grid;

[0096] P ref3 =max(V PV I PV ).

[0097] That is, the MPPT algorithm model is adjusted accordingly according to the main influencing factors to improve the power generation efficiency.

[0098] In a specific embodiment, at power P ref1 , Power P ref2、 Power ref3 When outputting respectively, through the real-time power P ac After difference processing, power control correction is performed to output control pulses.

[0099] Specifically, in general, in terms of control strategy, according to the power P ref1 , Power P ref2、 Power ref3 The outputs are output separately and the corresponding control pulses are adjusted. Correction can be performed through difference calculation, waveform correction, etc.

[0100] In this embodiment, the original data vector is processed by principal component analysis to reduce the dimension, and the variance contribution rate (η i)>80% of the variables are transmitted to the power control correction module as the main component factors. In the calculation module of the controller, the weights of the main component factors are adjusted in real time, and the final output control pulses are calculated by difference and other methods: boost module EPWM output (Boost_duty), inverter module EPWM output (Inv_duty), relay control module EPWM (Relay_control).

[0101] This embodiment is only one way to achieve power control correction. Sample other power P using the three models of this case. ref1 , power P ref2、 Power ref3 The schemes for respectively outputting corresponding corrective adjustment control are all within the scope of protection of this case.

[0102] In a specific embodiment, the time-sharing control strategy is specifically described:

[0103] In general, a feature recognition model is used to propose a series of data feature values, generally including: slope CIK, maximum value CImax; minimum value CImin, average value CIave; standard deviation CIvar; first-order derivative CI′; second-order derivative CI″.

[0104] Generate a data set X, X = {VPV, IPV, Vbus, Ttemp};

[0105] Slope

[0106] Maximum value data set CImax = max{X 1 ,X 2 ,......,X n );

[0107] Minimum data set CImin=min{X 1 ,X 2 ,......,X n};

[0108] average value

[0109] Standard Deviation

[0110] First Derivative

[0111] Second Derivative

[0112] is the sampling average value, n is the number of sampling points. To achieve real-time performance and accuracy, the sampling interval is set to 20 ms and the number of sampling points is 300. Of course, the sampling interval and the number of sampling points can be flexibly adjusted.

[0113] PV voltage rising trend: Perform moving average processing on the sampling data: k is the sampling window. Solve the slope of the moving average sequence to reflect the overall trend of voltage change. When the voltage is in the rising trend, CIk(MA)>0;

[0114] PV voltage stable trend: When the voltage is stable, the standard deviation CIvar(MA) approaches 0;

[0115] PV voltage falling trend: When the voltage is in the falling trend, CIk(MA)<0;

[0116] This is the basis for time-sharing control.

[0117] In a specific embodiment, predict the energy that the PV battery can provide by the capacity of the bus capacitor:

[0118]

[0119] E C is the capacity that the bus capacitor can store, C is the capacitance value, and U is the bus voltage;

[0120] Set the grid connection threshold Pb. When E C > Pb, allow the inverter to perform self-check and grid connection. That is, for starting up in the morning, adopt the power prediction method to make the inverter start up softly, avoiding damage to the switching devices caused by frequent start and stop of the boost circuit.

[0121] In a specific embodiment, calculate the PV battery power Ppv according to Vpv and Ipv. When Ppv < P1, enter the active shutdown program;

[0122] P1 = PR + Paux + Pfan + Prly + Pboost, where PR is the module loss, Paux is the sps power, Pfan is the fan power, Prly is the relay power, and Pboost is the switching loss of the boost circuit;

[0123] When Vpv < a1, increase the fake load boost_duty to simulate the increase of the fake load, and continuously monitor Vpv and Ipv;

[0124] When Vpv < a2, increasing the fake load boost_duty can no longer maintain the stability of the BUS voltage, then it is determined that the PV output is weak, actively disconnect from the grid, and disconnect the relay; the fake load consumes the PV power;

[0125] When Vpv < a3, gradually unload the dummy load until all loads are completely cut off.

[0126] Among them, a1 is the minimum operating voltage of the inverter boost, a2 is the grid-connected voltage at the lowest power consumption of the inverter, and a3 is the starting voltage of the switching power supply, where a1 > a2 > a3.

[0127] That is, for the condition of insufficient photovoltaic energy in the evening, respond to the active shutdown program of the inverter, determine the shutdown moment with the help of the dummy load, and at the same time consume the battery power under low irradiance, avoiding the repeated start and stop of DCSPS without intervention and damaging the switching devices.

[0128] In addition, it has an optimized MPPT algorithm for large PV voltage changes in the morning and evening:

[0129] Taking the data in the morning and evening as an example, the most obvious change is the PV voltage. The slope reflects the overall change trend, which is used to judge the characteristics of the morning and evening. CIK > 0 is in the morning, CIK < 0 is in the evening, and the absolute value of the derivative is used to judge the speed of voltage change. The variance Civar reflects the fluctuation of the PV voltage and reflects the control accuracy of the MPPT algorithm. When the variance is large, it is necessary to dynamically adjust the step size and perturbation direction of the MPPT algorithm.

[0130] Secondly, the characteristics of MPPT control reflected by the BUS voltage also need to be considered. Generally, when the energy is sufficient, the BUS voltage is maintained above the minimum rectification point. Taking the pressure difference between the actual bus and the reference bus as the characteristic criterion: ΔV bus =V busref -V bus ; Similarly, the slope is used to judge the time. Since the analyzed data is the BUS error, CIK < 0 is in the morning and CIK > 0 is in the evening. The characteristics of the bus are more obvious than those of the PV. Because when the PV output is insufficient, the IV characteristic is soft and it is easy to出现 the phenomenon that the boost output voltage cannot reach the target value. 0000

[0131] The change of temperature has hysteresis, but to a certain extent, it reflects the overall trend of the photovoltaic cell output, excluding the randomness introduced by other variables due to control method differences or sampling errors.

[0132] Through the above description, it can be found that it can realize principal component analysis to extract the main influencing factors of power fluctuation, thereby outputting power reference, meeting the adjustment requirements of the backend control strategy, and significantly improving the power generation efficiency. It meets the soft start requirements during the output strengthening period and the shutdown protection control requirements during the output weakening period, basically eliminating frequent start and stop, reducing the damage risk of switching components, and making the inverter operation control more reliable and stable.

[0133] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus / device.

[0134] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. The inverter control method based on the feature recognition model is characterized by It includes the following steps: S1 data acquisition, real-time acquisition of input voltage V PV 、Input current I PV 、Ambient temperature T temp , bus voltage V bus , real-time power P ac ; S2 according to the input voltage V PV Change trend, slope CIK, input voltage V PV The fluctuation difference CIvar is used to implement the inverter time-sharing control strategy; Output capability becomes stronger, input voltage V PV When the trend is upward and CIK>0; Output steady period, input voltage V PV Stable and CIvar tends to 0; Output capacity weakening period, input voltage V PV When the trend is down and CIK is less than 0; During the period when the output ability of S3 becomes stronger, a power prediction model is constructed for soft start; S4 uses principal component analysis to extract the main influencing factors on power fluctuation; When the main influencing factor is the input voltage V PV 、Input current I PV When the inverter efficiency equivalent model is used, the output power P ref1 ; When the main influencing factor is the input voltage V PV 、Ambient temperature T temp When the photovoltaic cell equivalent model is used, the output power P ref2 ; When the main influencing factor is the input voltage V PV , bus voltage V bus When the boost circuit mathematical model is used, the output power P ref3 ; During the period when the output ability of S5 weakens, active shutdown is performed.

2. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S4, a data set X is generated, where X={V PV , I PV , V bus , T temp }, find the mean CIave and standard deviation CIvar of each data vector, the sampling average Construct the covariance matrix V: n is the number of sampling points; Solve for the eigenvalue λ of V i and the eigenvector ω i ; Variance Contribution and cumulative variance contribution As the evaluation criteria, the eigenvector corresponding to the cumulative variance contribution rate of more than 80% was selected as the principal component; The eigenvalue is A = [λ1, λ2, λ3, ..., λ m ] corresponds to the eigenvector W m , W m =[ω1,ω2,ω3,......,ω m ] T ,m<4,main component Principal component reconstruction dataset 3. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S4, when the inverter efficiency equivalent model outputs; P ref1 =E AC =E DC ·η=V DC ·I DC ·or; Among them, E AC is the inverter output power; E DC is the DC input power; η is the comprehensive conversion efficiency.

4. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S4, when the photovoltaic cell equivalent model outputs; Where, I is the output current; V is the output voltage; I pv is the photocurrent; I o Saturation current; R s Series equivalent resistance; R p Shunt resistance; n ideality factor; k Boltzmann constant; T ambient temperature; T n Rated temperature; q charge element; Eg0 semiconductor band gap energy; N s Number of modules connected in series; P ref2 =V·I。 5. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S4, when the boost circuit mathematical model outputs; V busref -10V≤V bus ≤V busref +10V; wherein, D is the duty ratio of the boost circuit, and Va, Vb, and Vc are the three-phase voltage peaks of the three-phase power grid; P ref3 =max(V PV ·I PV )。 6. The inverter control method based on the feature recognition model according to any one of claims 1 to 5, wherein: In step S2, at power P ref1 , Power P ref2、 Power ref3 When outputting respectively, through the real-time power P ac After difference processing, power control correction is performed to output control pulses.

7. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S2, a data set X is generated, X = {VPV, IPV, Vbus, Ttemp}; Slope Maximum value data set CImax = max{X1,X2,.....,X n }; Minimum data set CImin=min{X1,X2,...,X n }; average value Standard Deviation First Derivative Second Derivative is the sampling mean, and n is the number of sampling points.

8. The inverter control method based on the feature recognition model according to claim 1, wherein: In step S3, the energy that the PV cell can provide is predicted through the capacity of the bus capacitor: E C is the capacity that the bus capacitor can store, C is the capacitance value, and U is the bus voltage; Set the grid connection threshold Pb, when E C >Pb, the inverter is allowed to self-check and connect to the grid.

9. The inverter control method based on the feature recognition model according to claim 1 is characterized in that Step S5 includes: The photovoltaic cell power Ppv is calculated according to Vpv and Ipv. When Ppv < P1, enter the active shutdown program; P1 = PR + Paux + Pfan + Prly + Pboost, where PR is the module loss, Paux is the sps power, Pfan is the fan power, Prly is the relay power, and Pboost is the boost circuit switching loss; When Vpv < a1, increase the fake load boost_duty to simulate the increase of the fake load, and continuously monitor Vpv and Ipv; When Vpv < a2, increasing the fake load boost_duty can no longer maintain the stability of the BUS voltage, then it is determined that the photovoltaic output is weak, actively disconnect from the grid, and disconnect the relay; the fake load consumes the photovoltaic power; When Vpv < a3, gradually unload the fake load until all loads are cut off. wherein, a1 is the minimum operating voltage of the inverter boost, a2 is the grid-connected voltage under the minimum power consumption of the inverter, and a3 is the starting voltage of the switching power supply, and a1 > a2 > a3.

Images