A control method and system for a current-source quasi-Z-source photovoltaic grid-connected inverter

Abstract

This invention discloses a current-source quasi-Z-source photovoltaic grid-connected inverter control method and system, comprising: inputting the inductor current error to the outer loop PI controller to obtain the given value of the grid-connected current, and calculating the grid-connected current error; after the grid-connected current error is adjusted by the middle loop passive sliding mode controller, outputting the given value of the capacitor voltage, and calculating the capacitor voltage error; after the capacitor voltage error is adjusted by the inner loop passive controller, outputting the reference value of the inverter output current; acquiring the ambient temperature and illuminance and inputting them to a pre-trained GRU neural network to obtain the reference value of the photovoltaic array's maximum power point output current, and then calculating the open-circuit duty cycle; based on the reference value of the inverter output current and the open-circuit duty cycle, generating PWM switching signals for each bridge arm through SVPWM modulation technology to drive the switching transistors of each bridge arm to turn on and off; this invention can improve the dynamic performance and steady-state accuracy of the system, and enhance its robustness against sudden changes in illuminance and grid disturbances.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of photovoltaic grid-connected inverters, and particularly relates to a current-mode quasi-Z-source photovoltaic grid-connected inverter control method and system. BACKGROUND

[0002] In a photovoltaic grid-connected power generation system, an inverter as a core unit of energy conversion, its topological structure and control strategy directly affect the power generation efficiency, grid-connected stability and anti-interference ability of the system. The current-mode quasi-Z-source inverter has the advantages of strong current limiting ability, fast dynamic response, high system reliability and single-stage grid-connected, and has a wide application prospect in the field of photovoltaic grid-connected power generation.

[0003] The current-mode quasi-Z-source inverter usually includes a quasi-Z-source impedance network and a CL filter on the AC side, both of which exhibit nonlinear characteristics. Among them, the CL filter is prone to cause resonance peaks due to its low damping coefficient, resulting in an increase in the total harmonic distortion (THD) of the grid-connected current. Although the existing passive damping method can suppress resonance, it will cause problems such as increased system loss and increased filter size, and in high-power application scenarios, the damping resistor will generate a lot of heat. The traditional active damping method (such as inductance current differential feedback) is relatively complex to implement, and it is difficult to balance damping effect and control stability.

[0004] In addition, the output characteristics of a photovoltaic array are easily affected by changes in light intensity and temperature. When the existing MPPT algorithm (such as the perturbation and observation method and the conductance increment method) is applied to a current-mode photovoltaic power generation system, due to the sharp decrease in current on the right side of the power-voltage curve at the maximum power point, it is easy to cause current out of control and power instantaneous drop, thereby making it difficult to achieve efficient and stable maximum power point tracking.

[0005] The above-mentioned existing technical defects seriously restrict the performance of the current-mode quasi-Z-source photovoltaic grid-connected inverter, resulting in problems such as poor grid-connected power quality, slow dynamic response and low MPPT efficiency, and new control strategies and algorithms are urgently needed to break through the technical bottleneck. SUMMARY

[0006] The technical problem to be solved by the present application is how to achieve efficient maximum power point tracking while ensuring that the grid-connected current is synchronized with the grid, the total harmonic distortion is low, the dynamic response is fast, and the robustness is strong.

[0007] To solve the above problems, the technical solution adopted by the present application is:

[0008] The present application provides a current-mode quasi-Z-source photovoltaic grid-connected inverter control method, comprising:

[0009] The mathematical model of the DC side of the inverter and the AC side in the dq coordinate system is constructed, the mathematical model of the current-mode quasi-Z-source grid-connected inverter is established by combining the mathematical model of the DC side of the inverter and the AC side in the dq coordinate system, and the inner loop passive controller and the middle loop passive sliding mode controller are constructed based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter;

[0010] The real-time inductor current in the impedance network of the DC side of the inverter and the inductor current reference value are obtained, and the inductor current error is calculated; the inductor current error is input to the outer loop PI controller to obtain the given value of the grid-connected current;

[0011] The grid-connected current error is calculated from the given value of the grid-connected current and the measured value of the grid-connected current, and the given value of the capacitor voltage is output after the grid-connected current error is adjusted by the middle loop passive sliding mode controller;

[0012] The capacitor voltage error is calculated from the given value of the capacitor voltage and the measured value of the capacitor voltage, and the reference value of the inverter output current is output after the capacitor voltage error is adjusted by the inner loop passive controller;

[0013] The ambient temperature and the light intensity are obtained and input to the pre-trained GRU neural network to obtain the maximum power point output current reference value of the photovoltaic array, and the photovoltaic array output current error is calculated according to the maximum power point output current reference value of the photovoltaic array and the measured value of the photovoltaic array output current, and the open-circuit duty ratio is obtained by inputting the photovoltaic array output current error to the PI controller;

[0014] According to the reference value of the inverter output current and the open-circuit duty ratio, the SVPWM modulation technology is used to generate the PWM switching signals of each bridge arm to drive the on-off of the switching tubes of each bridge arm, thereby realizing the grid-connected control of the inverter.

[0015] Further, the mathematical model of the DC side of the inverter is constructed, specifically including:

[0016] The state equation of the impedance network in the non-open-circuit state is expressed as:

[0017]

[0018] The state equation of the impedance network in the open-circuit state is expressed as:

[0019]

[0020] Let the open-circuit duty ratio be , then the non-open-circuit duty ratio is , and the state average of the state equations of the impedance network in the non-open-circuit state and the open-circuit state is taken to obtain the mathematical model of the DC side of the inverter as:

[0021]

[0022] In the formula, t is time, is the current flowing through the inductance and in the impedance network; is the voltage across the capacitance and in the impedance network; is the output current of the photovoltaic array; L is the inductance value of the inductance and in the impedance network; C is the capacitance value of the capacitance and in the impedance network; is the output voltage of the DC link of the inverter; is the output voltage of the photovoltaic array; is the inductance value of the input inductance on the DC side.

[0023] Further, the mathematical model of the AC side of the inverter in the dq coordinate system is constructed, which specifically includes:

[0024] The mathematical model of the AC side of the inverter in the three-phase stationary abc coordinate system is:

[0025]

[0026]

[0027] In the formula, is the inductance in the CL filter, is the kth phase grid-connected current, is the internal resistance of the filter inductance, is the voltage across the kth phase filter capacitance, is the kth phase grid voltage, is the capacitance in the CL filter, is the output current of the DC link, is the switching state of the kth phase bridge arm;

[0028] The mathematical model of the AC side of the inverter in the three-phase stationary abc coordinate system is converted into the mathematical model of the AC side of the inverter in the αβ coordinate system by using the transformation matrix , and the expression formula is:

[0029]

[0030]

[0031] The mathematical model of the AC side of the inverter in the αβ coordinate system is converted into the mathematical model of the AC side of the inverter in the dq coordinate system by using the transformation matrix , and the expression formula is:

[0032]

[0033]

[0034] wherein, is the angle between d-axis and a-axis, ; ω is the dq-axis rotation angular frequency; t is time; is the d-axis component of grid-connected current; is the q-axis component of grid-connected current; is the d-axis component of filter capacitor voltage across; is the q-axis component of filter capacitor voltage across; is the d-axis component of grid voltage; is the q-axis component of grid voltage; is the d-axis component of bridge arm switching state; is the q-axis component of bridge arm switching state.

[0035] Further, the mathematical model of the current-mode quasi-Z-source grid-connected inverter is established by combining the DC side of the inverter and the AC side model in the dq coordinate system, and specifically includes:

[0036]

[0037]

[0038]

[0039] wherein, is the current flowing through the inductance and in the impedance network; is the open-circuit duty ratio; is the DC link output current; is the inverter DC link output voltage; is the d-axis component of bridge arm switching state; is the q-axis component of bridge arm switching state; is the d-axis component of filter capacitor voltage across; is the q-axis component of filter capacitor voltage across.

[0040] Further, the middle-loop passive sliding mode controller and the inner-loop passive controller are constructed based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter, and specifically include:

[0041] According to the mathematical model of the current-mode quasi-Z-source grid-connected inverter, a port-controlled Hamiltonian (PCHD) model is constructed; based on the port-controlled Hamiltonian model, the interconnection and damping assignment passive control (IDA-PBC) method is adopted to derive the control law of the middle-loop passive controller and the inner-loop passive controller;

[0042] The sliding mode surface is set based on the desired equilibrium point, and a sliding mode control law is derived according to a mathematical model of an inverter AC side in a dq coordinate system; the sliding mode control law is fused with a passive control law corresponding to the middle-ring passive controller to obtain a control law of the middle-ring passive sliding mode controller.

[0043] Further, a given value of a capacitor voltage output by the middle-ring passive sliding mode controller after regulating a grid-connected current error, specifically includes:

[0044]

[0045] In the formula, is a d-axis component of a voltage reference value on both sides of a filter capacitor; is a d-axis component of a grid voltage; and is an injected damping coefficient; is an inductance in a CL filter; is an internal resistance of a filter inductor; , , and is a reaching law coefficient; is a d-axis component of a grid-connected current reference value; is a d-axis component of a grid-connected current measurement value; is a dq-axis rotation frequency; is a q-axis component of a voltage reference value on both sides of a filter capacitor; is a q-axis component of a grid voltage; is a q-axis component of a grid-connected current reference value; is a q-axis component of a grid-connected current measurement value.

[0046] Further, a reference value of an inverter output current output by the capacitor voltage error after being regulated by the inner-ring passive controller, specifically includes:

[0047]

[0048] In the formula, is a d-axis component of an inverter output current reference value; is a d-axis component of a bridge arm switch state; is a q-axis component of a bridge arm switch state; is a current reference value flowing through an inductance and in an impedance network; is an open-circuit duty ratio; is a d-axis component of a grid-connected current reference value; and is an injected damping coefficient; d-axis component of the voltage reference across the filter capacitor; d-axis component of the voltage measurement across the filter capacitor; frequency of dq-axis rotation; capacitance in the CL filter; q-axis component of the voltage reference across the filter capacitor; q-axis component of the voltage measurement across the filter capacitor; q-axis component of the inverter output current reference; q-axis component of the grid-connected current reference.

[0049] Further, the GRU neural network training process comprises:

[0050] obtaining the light intensity and the ambient temperature of the photovoltaic array as input features of the training sample, and obtaining the corresponding actual maximum power point current as the true label of the training sample;

[0051] randomly initializing the hyperparameter combination of the GRU neural network; training the GRU neural network by using the training sample to obtain a maximum power point predicted current; calculating the root mean square error between the maximum power point predicted current and the actual current; based on the root mean square error, updating the hyperparameter combination of the GRU neural network by using a Bayesian optimization method;

[0052] repeating the above training process iteratively until the termination condition of the Bayesian optimization is met, and finally outputting a GRU neural network model with optimal root mean square error.

[0053] The second aspect of the application provides a current-mode quasi-Z-source photovoltaic grid-connected inverter control system, comprising:

[0054] a model building module, configured to build mathematical models of the inverter DC side and the AC side in the dq coordinate system, combine the mathematical models of the inverter DC side and the AC side in the dq coordinate system, and establish a mathematical model of the current-mode quasi-Z-source grid-connected inverter; and based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter, build a middle-ring passive sliding mode controller and an inner-ring passive controller;

[0055] an inductor current tracking control module, configured to obtain the real-time inductor current in the inductor impedance network of the inverter DC side and the inductor current reference value, and calculate an inductor current error; and input the inductor current error into an outer-ring PI controller to obtain a given value of the grid-connected current;

[0056] a grid-connected current tracking control module, configured to calculate a grid-connected current error from the given value of the grid-connected current and the measurement value of the grid-connected current, and output a given value of the capacitor voltage after the grid-connected current error is adjusted by the middle-ring passive sliding mode controller;

[0057] The capacitor voltage tracking control module is used to calculate the capacitor voltage error from the given value and the measured value of the capacitor voltage. After the capacitor voltage error is adjusted by the inner loop passive controller, the reference value of the inverter output current is output.

[0058] The prediction module acquires the ambient temperature and light intensity and inputs them into a pre-trained GRU neural network to obtain the reference value of the maximum power point output current of the photovoltaic array. Based on the reference value of the maximum power point output current of the photovoltaic array and the measured value of the output current of the photovoltaic array, the module calculates the output current error of the photovoltaic array and inputs the output current error of the photovoltaic array into the PI controller to obtain the open circuit duty cycle.

[0059] The bridge arm switch control module generates PWM switching signals for each bridge arm using SVPWM modulation technology based on the reference value of the inverter output current and the open-circuit duty cycle, thereby driving the switching transistors of each bridge arm to turn on and off, and realizing the grid-connected control of the inverter.

[0060] A third aspect of the present invention provides an electronic terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the current-mode quasi-Z-source photovoltaic grid-connected inverter control method of the first aspect.

[0061] Compared with existing technologies, the beneficial effects of this invention are as follows: the three-loop composite control strategy can significantly improve the system damping ratio, effectively suppress the resonance spike of the AC side CL filter, reduce the total harmonic distortion (THD) of the grid-connected current, improve the dynamic response speed and steady-state accuracy of the system, and enhance its robustness under sudden changes in illumination and grid disturbances; this invention can ensure that the system can quickly and accurately track the global maximum power point, effectively improve the overall conversion efficiency of the photovoltaic array, and fundamentally solve the problems of current runaway and instantaneous power drop that exist when the traditional disturbance observation method is applied to current-source photovoltaic power generation systems. Attached Figure Description

[0062] Figure 1 This is a structural diagram of the current-source quasi-Z-source photovoltaic grid-connected inverter provided in Embodiment 1 of the present invention;

[0063] Figure 2 This is the equivalent circuit of the quasi-Z source inverter in the non-open circuit state provided in Embodiment 1 of the present invention;

[0064] Figure 3 This is the equivalent circuit of the quasi-Z source inverter in the open-circuit state provided in Embodiment 1 of the present invention;

[0065] Figure 4 This is a structural block diagram of the current-source quasi-Z-source photovoltaic grid-connected inverter control method provided in Embodiment 1 of the present invention;

[0066] Figure 5is an internal structure diagram of the GRU neural network provided in Embodiment 1 of the present application;

[0067] Figure 6 is a grid-connected current waveform diagram under different control strategies provided in Embodiment 1 of the present application;

[0068] Figure 7 is a grid-connected active power and reactive power comparison diagram under different control strategies provided in Embodiment 1 of the present application;

[0069] Figure 8 is a photovoltaic array output power comparison diagram under different control strategies provided in Embodiment 1 of the present application;

[0070] Figure 9 is a photovoltaic array output power curve diagram under the three-closed-loop composite control strategy in the case of light mutation provided in Embodiment 1 of the present application;

[0071] Figure 10 is a grid-connected current output waveform diagram under the three-closed-loop composite control strategy in the case of light mutation provided in Embodiment 1 of the present application;

[0072] Figure 11 is a grid-connected active power and reactive power output waveform diagram under the three-closed-loop composite control strategy in the case of light mutation provided in Embodiment 1 of the present application. DETAILED DESCRIPTION

[0073] The present application will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, and cannot be used to limit the protection scope of the present application.

[0074] Embodiment 1

[0075] As shown in Figure 4 , the present embodiment provides a current-mode quasi-Z-source photovoltaic grid-connected inverter control method, which comprises:

[0076] The topology structure of the three-phase current-mode quasi-Z-source photovoltaic grid-connected inverter is as shown in Figure 1 . In the figure, , , . u pv is the photovoltaic array output voltage, i pv is the photovoltaic array output current, u out is the inverter DC link output voltage, i pn is the DC link output current. i sa , i sb , i sc is the three-phase inverter output current, i ga , i gb , i gc is the three-phase grid-connected current, u sau sb u sc The voltage across the filter capacitor, u ga u gb u gc For the three-phase power grid voltage, C f and L f For the capacitor and inductor in the CL filter, R f This is the internal resistance of the filter inductor.

[0077] Constructing a mathematical model for the DC side of the inverter, specifically including:

[0078] like Figure 2 As shown, the state equation of the impedance network in the non-open-circuit state is expressed by the following formula:

[0079]

[0080] like Figure 3 As shown, the state equation of the impedance network in the open-circuit state is expressed by the following formula:

[0081]

[0082] Let the open circuit duty cycle be... Then the non-open circuit duty cycle is Taking the state average of the state equations of the impedance network under both open-circuit and closed-circuit conditions, the mathematical model of the DC side of the inverter is obtained as follows:

[0083]

[0084] In the formula, t represents time. For the inductance flowing through the impedance network and The current; Capacitor in an impedance network and Voltage at both ends; L is the output current of the photovoltaic array; L is the inductance in the impedance network. and The inductance value; C is the capacitance in the impedance network. and The capacitance value; This refers to the DC-DC output voltage of the inverter. This refers to the output voltage of the photovoltaic array. DC-side input inductor The inductance value.

[0085] Constructing a mathematical model of the inverter's AC side in the dq coordinate system, specifically including:

[0086] The mathematical model of the inverter's AC side in the three-phase stationary abc coordinate system is as follows:

[0087]

[0088]

[0089] In the formula, For the inductor in the CL filter, Let k be the grid-connected current of phase k. The internal resistance of the filter inductor is... The voltage across the filter capacitor of phase k is... Let k be the voltage of the power grid phase. For the capacitor in the CL filter, For DC link output current, This represents the switching state of the k-th phase bridge arm;

[0090] Using the transformation matrix The mathematical model of the inverter's AC side in the three-phase stationary abc coordinate system is transformed into a mathematical model of the inverter's AC side in the αβ coordinate system, expressed by the following formula:

[0091]

[0092]

[0093] Using the transformation matrix The mathematical model of the inverter's AC side in the αβ coordinate system is transformed into the mathematical model of the inverter's AC side in the dq coordinate system, expressed by the following formula:

[0094]

[0095]

[0096] In the formula, Let be the angle between the d-axis and the α-axis. ω is the angular frequency of rotation along the dq axis, and t is time. This represents the d-axis component of the grid-connected current. This represents the q-axis component of the grid-connected current. This represents the d-axis component of the voltage across the filter capacitor. This represents the q-axis component of the voltage across the filter capacitor. This represents the d-axis component of the grid voltage; This represents the q-axis component of the grid voltage. The d-axis component represents the state of the bridge arm switch. This represents the q-axis component of the bridge arm switching state.

[0097] From the different switching states of the current-source inverter, we can obtain:

[0098]

[0099] When the inverter is in a non-open circuit state, we have:

[0100]

[0101] In the formula, For the inductance flowing through the impedance network and The current; Duty cycle for open circuit; This is the DC link output current; This refers to the DC-DC output voltage of the inverter. The d-axis component represents the state of the bridge arm switch. The q-axis component represents the state of the bridge arm switch. This represents the d-axis component of the voltage across the filter capacitor. This represents the q-axis component of the voltage across the filter capacitor.

[0102] Based on the mathematical models of the inverter's DC side and AC side in the dq coordinate system, and combining equations (10) and (11), the mathematical model of the current-source quasi-Z-source grid-connected inverter can be obtained, specifically including:

[0103]

[0104] Based on the mathematical model of a current-source quasi-Z-source grid-connected inverter, a passive controller and a passive sliding mode controller are constructed, specifically including:

[0105] A PCHD model of a three-phase current-source quasi-Z-source photovoltaic grid-connected inverter is established, expressed by the following formula:

[0106]

[0107] In the formula, x is the state variable of the quasi-Z-source photovoltaic grid-connected inverter; u is the control variable of the quasi-Z-source photovoltaic grid-connected inverter; y is the output variable of the quasi-Z-source photovoltaic grid-connected inverter; J(x) is the interconnection matrix of the quasi-Z-source photovoltaic grid-connected inverter, which has antisymmetric characteristics; R(x) is the dissipation matrix of the quasi-Z-source photovoltaic grid-connected inverter, which is a positive definite symmetric matrix; g(x) is the structure matrix of the direct effect of the control variables on the state variables; and H(x) is the energy storage function.

[0108] Based on the PCHD model parameters J(x), R(x), H(x), g(x) and the desired equilibrium point x*, a passive controller is constructed using the Interconnected and Damped Distribution Passive Control (IDA-PBC) method. It is assumed that a function J exists. a (x), R a (x), u, and the vector function K(x) satisfy:

[0109]

[0110] And the following constraints are met:

[0111] (1) Structural conservation, that is:

[0112]

[0113] In the formula, J d (x) and R d (x) represents the new interconnection matrix and dissipation matrix; J a (x) and R a (x) represents the interconnection and damping injection matrix.

[0114] (2) Integrability, i.e.:

[0115]

[0116] (3) H d (x) at the desired equilibrium point x * If there exists an extremum at a certain point, then the vector function K(x) satisfies:

[0117]

[0118] (4) At the expected equilibrium point x * The Jacobian matrix at point K(x) satisfies:

[0119]

[0120] Then the expected equilibrium point This is a stable equilibrium point for a quasi-Z-source photovoltaic grid-connected inverter, and the quasi-Z-source photovoltaic grid-connected inverter can be expressed as:

[0121]

[0122] In the formula, Let be the desired energy storage function. This indicates that the energy injected into the quasi-Z-source photovoltaic grid-connected inverter is controlled, and there is... .

[0123] According to equation (14), the passive controller based on the PCHD model can be obtained as follows:

[0124]

[0125] The energy storage functions are as follows:

[0126]

[0127] In the formula, D is a positive definite diagonal matrix, and we have:

[0128]

[0129] Substituting equation (22) into equation (20), we get:

[0130]

[0131] Based on the mathematical model of the current-source quasi-Z-source grid-connected inverter, a port-controlled dissipation Hamiltonian (PCHD) model is constructed, specifically as follows:

[0132]

[0133]

[0134]

[0135]

[0136]

[0137] Let the desired equilibrium point be set. ;

[0138] Pick , Combining equation (23), we can obtain:

[0139]

[0140] Expanding on:

[0141]

[0142] The control laws for the middle-loop passive controller and the inner-loop passive controller can be derived, expressed as follows:

[0143]

[0144]

[0145] In the formula, The d-axis component of the inverter output current reference value; The d-axis component represents the state of the bridge arm switch. The q-axis component represents the state of the bridge arm switch. For the inductance flowing through the impedance network and The current reference value; Duty cycle for open circuit; The d-axis component of the grid-connected current reference value; The d-axis component of the grid-connected current measurement value; , , and The damping coefficient is the injection damping coefficient. The d-axis component represents the reference voltage value across the filter capacitor. The d-axis component of the voltage measurement value across the filter capacitor; The angular frequency of rotation along the dq axis; For the capacitor in the CL filter; The inductor in the CL filter; This is the internal resistance of the filter inductor; The q-axis component of the reference voltage value across the filter capacitor; This represents the q-axis component of the voltage measurement across the filter capacitor. The q-axis component of the inverter output current reference value; The q-axis component of the grid-connected current reference value; The q-axis component of the grid-connected current measurement value; This represents the d-axis component of the grid voltage; This represents the q-axis component of the grid voltage.

[0146] The sliding mode surface is set based on the desired equilibrium point, and the sliding mode control law is derived according to the mathematical model of the inverter AC side in the dq coordinate system. The sliding mode control law is then integrated with the passive control law corresponding to the middle-loop passive controller to obtain the control law of the middle-loop passive sliding mode controller, specifically including:

[0147] Based on the requirements of grid-connected current loop control, a sliding mode surface is selected. Sliding surface for:

[0148]

[0149] According to sliding mode control based on the exponential reaching law, let:

[0150]

[0151] In the formula, , , , These are the convergence law coefficients, and all are greater than 0.

[0152] To avoid high-frequency chattering, the hyperbolic tangent function tanh( ) replaces the symbolic function sgn( From this, we can obtain:

[0153]

[0154] Combining equations (9), (33), and (35), the reference value for the capacitor voltage can be obtained as follows:

[0155]

[0156] Combining equations (31) and (36), we obtain:

[0157]

[0158] Substituting equation (37) into equation (31), we obtain the control law of the passive sliding mode controller, expressed as:

[0159]

[0160] In the formula, The d-axis component represents the reference voltage value across the filter capacitor. This represents the d-axis component of the grid voltage; and The damping coefficient is the injection damping coefficient. The inductor in the CL filter; This is the internal resistance of the filter inductor; , , and Here are the coefficients for the approach law; The d-axis component of the grid-connected current reference value; The d-axis component of the grid-connected current measurement value; The angular frequency of rotation along the dq axis; The q-axis component of the reference voltage value across the filter capacitor; This represents the q-axis component of the grid voltage. The q-axis component of the grid-connected current reference value; This represents the q-axis component of the grid-connected current measurement.

[0161] like Figure 5 The diagram shows the internal structure of the GRU neural network. The training process for the GRU neural network includes:

[0162] The irradiance and ambient temperature of the photovoltaic array are used as input features for the training samples, and the corresponding actual maximum power point current is used as the true label of the training samples.

[0163] The hyperparameter combination of the GRU neural network is randomly initialized; the GRU neural network is trained using training samples to obtain the maximum power point predicted current; the root mean square error between the maximum power point predicted current and the actual current is calculated; based on the root mean square error, the hyperparameter combination of the GRU neural network is updated using a Bayesian optimization method.

[0164] Repeat the above training process iteratively until the termination condition of Bayesian optimization is met, and finally output the GRU neural network model with the optimal root mean square error.

[0165] Obtain the real-time inductor current and inductor current reference value in the DC-side impedance network of the inverter and calculate the inductor current error; input the inductor current error to the outer loop PI controller to obtain the given value of the grid-connected current;

[0166] The grid-connected current error is calculated from the given value and the measured value of the grid-connected current. After the grid-connected current error is adjusted by the passive sliding mode controller in the middle loop, the given value of the output capacitor voltage is obtained.

[0167] The capacitor voltage error is calculated from the given value and the measured value of the capacitor voltage. After the capacitor voltage error is adjusted by the inner loop passive controller, the reference value of the inverter output current is output.

[0168] The ambient temperature and light intensity are acquired and input into a pre-trained GRU neural network to obtain the reference value of the maximum power point output current of the photovoltaic array. The output current error of the photovoltaic array is calculated based on the reference value of the maximum power point output current of the photovoltaic array and the measured value of the output current of the photovoltaic array. The output current error of the photovoltaic array is input into the PI controller to obtain the open circuit duty cycle.

[0169] Based on the reference value of the inverter output current and the open-circuit duty cycle, SVPWM modulation technology (space vector pulse width modulation technology) is used to generate PWM switching signals for each bridge arm, driving the switching transistors of each bridge arm to turn on and off, thereby realizing the grid-connected control of the inverter.

[0170] This study focuses on a photovoltaic array consisting of two modules connected in series under uniform light intensity and temperature conditions. The required environmental data is obtained from measured data on an experimental platform, sampled from 5:30 AM to 5:30 PM with a 1-minute interval. A simulation model of the photovoltaic array is established, and given the measured light intensity and temperature parameters, the maximum power point current obtained from the simulation is used as the sample output, resulting in 720 sets of sample data. 80% of these are used as the training set, and the remaining 20% ​​are used as the test set.

[0171] The root mean square error (RMSE) and mean absolute error (MAE) are used as evaluation metrics to assess the prediction accuracy of the MPPT algorithm. The formulas for calculating RMSE and MAE are as follows:

[0172]

[0173]

[0174] In the formula, N0 is the sample size. and These are the actual current value and the predicted current value at time n, respectively.

[0175] To verify the superiority of the BO-GRU algorithm (i.e., the gated recurrent unit neural network algorithm based on Bayesian optimization in this embodiment) in the application of maximum power point tracking in photovoltaic power generation systems, the prediction accuracy of the BO-GRU algorithm is compared with that of BP neural network, GRU neural network and PSO-GRU algorithm (gated recurrent unit neural network algorithm based on particle swarm optimization). The parameter settings of different MPPT algorithms are shown in Table 1.

[0176] Table 1. Parameter settings for different MPPT algorithms

[0177]

[0178] Table 2 shows the statistical data of the root mean square error (RMSE), mean absolute error (MAE), and training time for each MPPT algorithm, which were run 10 times. The optimal values ​​for the same evaluation metric are shown in bold. Table 3 shows the optimal prediction parameters for the BO-GRU and PSO-GRU algorithms.

[0179] Table 2. Comparison of Model Prediction Results and Training Time

[0180]

[0181] Table 3. Parameter optimization results of different algorithms

[0182]

[0183] As shown in Table 2: (1) The average values ​​of the evaluation metrics RMSE and MAE of the BO-GRU algorithm decreased by 88.38% and 88.75% respectively compared to the GRU neural network. This indicates that obtaining the optimal neural network hyperparameters through the Bayesian optimization algorithm can effectively improve the performance of the neural network, verifying the superiority of the proposed strategy. (2) Although the PSO-GRU algorithm can obtain the optimal prediction accuracy, it is prone to premature convergence, and the training time of the PSO-GRU algorithm is nearly 20 times that of the BO-GRU algorithm. (3) The difference between the optimal prediction accuracy of BO-GRU and PSO-GRU is almost negligible, but the training time is greatly reduced, and the prediction results are similar each time, indicating good algorithm stability and better overall performance.

[0184] To verify the superiority of the proposed three-loop composite control strategy (PI-passive sliding mode-passive control of inductor current outer loop, passive sliding mode control of grid-connected current middle loop, and passive control of capacitor voltage inner loop), a simulation model of a current-mode quasi-Z-source photovoltaic grid-connected power generation system was built using MATLAB / Simulink. The simulation results were compared with the following methods: PI control of both inductor current outer loop and grid-connected current middle loop, and P control of capacitor voltage inner loop (hereinafter referred to as PI-PI-P); PI control of inductor current outer loop, and passive control of both grid-connected current middle loop and capacitor voltage inner loop (hereinafter referred to as PI-passive-passive control); and PI control of inductor current outer loop and PR control of grid-connected current inner loop (hereinafter referred to as PI-PR).

[0185] The simulation uses a 15×2 photovoltaic array. The photovoltaic module parameters are shown in Table 4. Under standard conditions, the output power of this photovoltaic array is approximately 10kW. The DC-side inductor L3 is selected as 10 mH; the quasi-Z source impedance network inductor L is selected as 2 mH; the capacitor C is selected as 60 μF; and the AC-side filter capacitor C... f 47μF; AC side filter inductor L f The internal resistance R of the filter inductor is 3mH. f The resistance is 0.2 Ω; the peak grid voltage is 311V; and the carrier frequency is 15kHz.

[0186] Table 4. Characteristic parameters of photovoltaic modules

[0187]

[0188] Figure 6From top to bottom, the waveforms represent the grid-connected current output under ideal grid voltage and standard photovoltaic array operation: PI-passive sliding mode-passive control, PI-PI-P control, PI-passive-passive control, and PI-PR control strategies. The output waveforms of active and reactive power under these four different control strategies are shown below. Figure 7 As shown, the output power waveform of the photovoltaic array is as follows: Figure 8 As shown.

[0189] Depend on Figure 7 and Figure 8 It can be seen that under the four different control strategies, the grid-connected reactive power is 0W and the photovoltaic array output power is about 10kW after the system reaches a steady state. This indicates that all four different control strategies can achieve maximum power point tracking control of the photovoltaic array while achieving grid connection with unity power factor. Among them, the PI-passive sliding mode-passive control strategy has the best maximum power point tracking speed.

[0190] Under the PI-passive sliding mode-passive control strategy, the THD (Total Harmonic Distortion) values ​​of the three-phase grid-connected current are 0.44%, 0.44%, and 0.47%, respectively; under the PI-PI-P control strategy, the THD values ​​are 0.83%, 0.82%, and 0.85%, respectively; under the PI-passive-passive control strategy, the THD values ​​are 0.69%, 0.70%, and 0.70%, respectively; and under the PI-PR control strategy, the THD values ​​are 0.88%, 0.91%, and 0.89%, respectively. It can be seen that the PI-PR control strategy has the highest THD value, while the PI-passive sliding mode-passive control strategy has the lowest THD value. Compared to the dual closed-loop control strategy of inductor current outer loop and grid-connected current inner loop, the three closed-loop control strategy of inductor current outer loop, grid-connected current middle loop, and capacitor voltage inner loop has better performance. Moreover, compared to PI-passive-passive control and PI-PI-P control, the PI-passive sliding mode-passive control strategy has certain advantages in harmonic suppression and better steady-state accuracy.

[0191] In actual operation, the light intensity is not constant. For example, if the light intensity changes from 1000 W / m² in 0.2 seconds... 2 Reduced to 700W / m 2 At this point, the maximum power point current of the photovoltaic array is approximately 12.4694A, and the maximum output power is approximately 7080W. Under normal grid operation, the output power curve of the photovoltaic array under the PI-passive sliding mode-passive control strategy is as follows: Figure 9 As shown, the grid-connected current output waveform is as follows: Figure 10 As shown, the output waveforms of grid-connected active and reactive power are as follows: Figure 11 As shown.

[0192] Depend on Figure 9 ,Figure 10 and Figure 11 It is known that when the light intensity decreases, the output power of the photovoltaic array decreases, and the grid-connected current also decreases accordingly. The PI-passive sliding mode-passive control strategy enables the system to quickly reach a new steady state, achieving unity power factor grid connection while ensuring that the photovoltaic array remains stable near the maximum power point, exhibiting good dynamic performance.

[0193] Example 2

[0194] This embodiment provides a current-source quasi-Z-source photovoltaic grid-connected inverter control system. The control system is used to execute the control method described in Embodiment 1. The control system includes:

[0195] The model building module is used to construct mathematical models of the inverter's DC side and AC side in the dq coordinate system. It combines the mathematical models of the inverter's DC side and AC side in the dq coordinate system to establish a mathematical model of a current-mode quasi-Z-source grid-connected inverter. Based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter, it constructs a middle-loop passive sliding mode controller and an inner-loop passive controller.

[0196] The inductor current tracking control module is used to obtain the real-time inductor current and inductor current reference value in the DC-side impedance network of the inverter and calculate the inductor current error; the inductor current error is input to the outer loop PI controller to obtain the given value of the grid-connected current;

[0197] The grid-connected current tracking control module is used to calculate the grid-connected current error from the given value and the measured value of the grid-connected current. After the grid-connected current error is adjusted by the passive sliding mode controller, the given value of the capacitor voltage is output.

[0198] The capacitor voltage tracking control module is used to calculate the capacitor voltage error from the given value and the measured value of the capacitor voltage. After the capacitor voltage error is adjusted by the inner loop passive controller, the reference value of the inverter output current is output.

[0199] The prediction module acquires the ambient temperature and light intensity and inputs them into a pre-trained GRU neural network to obtain the reference value of the maximum power point output current of the photovoltaic array. Based on the reference value of the maximum power point output current of the photovoltaic array and the measured value of the output current of the photovoltaic array, the module calculates the output current error of the photovoltaic array and inputs the output current error of the photovoltaic array into the PI controller to obtain the open circuit duty cycle.

[0200] The bridge arm switch control module generates PWM switching signals for each bridge arm using SVPWM modulation technology based on the reference value of the inverter output current and the open-circuit duty cycle, thereby driving the switching transistors of each bridge arm to turn on and off, and realizing the grid-connected control of the inverter.

[0201] Example 3

[0202] This embodiment provides an electronic terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the current-type quasi-Z-source photovoltaic grid-connected inverter control method described in Embodiment 1.

[0203] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A control method of a current-mode quasi-Z-source photovoltaic grid-connected inverter, characterized in that, include: Mathematical models of the inverter's DC side and AC side in the dq coordinate system are constructed. By combining the mathematical models of the inverter's DC side and AC side in the dq coordinate system, a mathematical model of a current-mode quasi-Z-source grid-connected inverter is established. Based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter, a middle-loop passive sliding mode controller and an inner-loop passive controller are constructed. Obtain the real-time inductor current and inductor current reference value in the DC-side impedance network of the inverter, and calculate the inductor current error. The inductor current error is input to the outer loop PI controller to obtain the given value of the grid-connected current; The grid-connected current error is calculated from the given value and the measured value of the grid-connected current. After the grid-connected current error is adjusted by the passive sliding mode controller in the middle loop, the given value of the output capacitor voltage is obtained. The capacitor voltage error is calculated from the given value and the measured value of the capacitor voltage. After the capacitor voltage error is adjusted by the inner loop passive controller, the reference value of the inverter output current is output. The ambient temperature and light intensity are acquired and input into a pre-trained GRU neural network to obtain the reference value of the maximum power point output current of the photovoltaic array. The output current error of the photovoltaic array is calculated based on the reference value of the maximum power point output current of the photovoltaic array and the measured value of the output current of the photovoltaic array. The output current error of the photovoltaic array is input into the PI controller to obtain the open circuit duty cycle. Based on the reference value of the inverter output current and the open-circuit duty cycle, SVPWM modulation technology is used to generate PWM switching signals for each bridge arm, driving the switching transistors of each bridge arm to turn on and off, thereby realizing the grid-connected control of the inverter.

2. The control method according to claim 1, characterized by, Constructing a mathematical model for the DC side of the inverter, specifically including: The state equation of the impedance network in the non-open-circuit state is expressed by the following formula: ； The state equation for an impedance network in the open-circuit state is expressed as follows: ； Let the open-circuit duty ratio be The non-open-circuit duty ratio is Taking the state average of the state equations of the impedance network in the non-open-circuit state and the open-circuit state, the mathematical model of the DC side of the inverter is obtained as ； In the formula, t is time, is the current flowing through the inductance and in the impedance network; is the voltage across the capacitance and in the impedance network; is the output current of the photovoltaic array; L is the inductance value of the inductance and in the impedance network; C is the capacitance value of the capacitance and in the impedance network; is the output voltage of the inverter DC link; is the output voltage of the photovoltaic array; is the inductance value of the input inductance on the DC side.

3. The control method according to claim 2, characterized by, Constructing a mathematical model of the inverter's AC side in the dq coordinate system, specifically including: The mathematical model of the inverter's AC side in the three-phase stationary abc coordinate system is as follows: ； ； in the formula, Lk is the inductance in the CL filter, Ik is the kth phase grid current, Rk is the internal resistance of the filter inductance, Vk is the voltage across the kth phase filter capacitor, Vk is the kth phase grid voltage, Ck is the capacitance in the CL filter, Ik is the DC link output current, Sk is the switching state of the kth phase bridge leg; Utilizing transformation matrices The mathematical model of the inverter AC side in the three-phase stationary abc coordinate system is converted into the mathematical model of the inverter AC side in the αβ coordinate system, and the expression formula is: ； ； Utilizing transformation matrices The mathematical model of the inverter AC side in the αβ coordinate system is converted into the mathematical model of the inverter AC side in the dq coordinate system, and the expression formula is: ； ； in the formula, is the included angle between the d-axis and the a-axis, ; ω is the dq-axis rotation angular frequency; t is time; is the d-axis component of the grid-connected current; is the q-axis component of the grid-connected current; is the d-axis component of the voltage across the filter capacitor; is the q-axis component of the voltage across the filter capacitor; is the d-axis component of the grid voltage; is the q-axis component of the grid voltage; is the d-axis component of the bridge arm switching state; is the q-axis component of the bridge arm switching state.

4. The control method according to claim 3, characterized by By combining the DC side model and the AC side model in the dq coordinate system of the inverter, a mathematical model of a current-mode quasi-Z-source grid-connected inverter is established, specifically including: ； ； ； In the formulas, is the current flowing through the inductance and of the impedance network; is the open circuit duty cycle; is the DC link output current; is the inverter DC link output voltage; is the d-axis component of the bridge leg switching state; is the q-axis component of the bridge leg switching state; is the d-axis component of the voltage across the filter capacitor; is the q-axis component of the voltage across the filter capacitor.

5. The control method according to claim 1, characterized by, Based on the mathematical model of a current-source quasi-Z-source grid-connected inverter, a middle-loop passive sliding mode controller and an inner-loop passive controller are constructed, specifically including: A port-controlled dissipation Hamiltonian model is constructed based on the mathematical model of a current-source quasi-Z-source grid-connected inverter. Based on the port-controlled dissipation Hamiltonian model, the control laws of the middle-loop passive controller and the inner-loop passive controller are derived using interconnection and damping distribution passive control methods. The sliding mode surface is set based on the desired equilibrium point, and the sliding mode control law is derived based on the mathematical model of the inverter AC side in the dq coordinate system. The sliding mode control law is then integrated with the passive control law corresponding to the middle loop passive controller to obtain the control law of the middle loop passive sliding mode controller.

6. The control method according to claim 5, characterized by The given value of the output capacitor voltage after the grid-connected current error is adjusted by the passive sliding mode controller of the central loop includes: ； is the d-axis component of the voltage reference across the filter capacitor; is the d-axis component of the grid voltage; is the d-axis component of the grid voltage; and is the damping injection coefficient; is the inductance in the CL filter; is the internal resistance of the filter inductance; , , and is the approaching law coefficient; is the d-axis component of the grid current reference; is the d-axis component of the grid current measurement; is the dq-axis rotation frequency; is the q-axis component of the voltage reference across the filter capacitor; is the q-axis component of the grid voltage; is the q-axis component of the grid current reference; is the q-axis component of the grid current measurement.

7. The control method according to claim 5, characterized by, The capacitor voltage error, after being adjusted by the inner-loop passive controller, outputs a reference value for the inverter's output current, specifically including: ； in the formula, is the d-axis component of the inverter output current reference value; is the d-axis component of the bridge leg switching state; is the q-axis component of the bridge leg switching state; is the current reference value flowing through the inductance and in the impedance network; is the open-loop duty ratio; is the d-axis component of the grid-connected current reference value; and is the injected damping coefficient; is the d-axis component of the voltage reference value across the filter capacitor; is the d-axis component of the voltage measurement across the filter capacitor; is the dq-axis rotation angular frequency; is the capacitance in the CL filter; is the q-axis component of the voltage reference value across the filter capacitor; is the q-axis component of the voltage measurement across the filter capacitor; is the q-axis component of the inverter output current reference value; is the q-axis component of the grid-connected current reference value.

8. The control method according to claim 1, characterized by, The training process for a GRU neural network includes: The irradiance and ambient temperature of the photovoltaic array are used as input features for the training samples, and the corresponding actual maximum power point current is used as the true label of the training samples. Randomly initialize a hyperparameter combination of a GRU neural network; train the GRU neural network using training samples to obtain a maximum power point predicted current; calculate a root mean square error between the maximum power point predicted current and an actual current; based on the root mean square error, update the hyperparameter combination of the GRU neural network using a Bayesian optimization method; Repeat the above training process until a termination condition of the Bayesian optimization is met, and finally output a GRU neural network model with an optimal root mean square error.

9. A current-fed quasi-Z-source photovoltaic grid-connected inverter control system, characterized in that, Comprise: A model building module is configured to build mathematical models of a DC side of an inverter and an AC side in a dq coordinate system, combine the mathematical models of the DC side of the inverter and the AC side in the dq coordinate system, and establish a mathematical model of a current-mode quasi-Z-source grid-connected inverter; a middle-loop passive sliding mode controller and an inner-loop passive controller are built based on the mathematical model of the current-mode quasi-Z-source grid-connected inverter; An inductor current tracking control module is configured to obtain real-time inductor current in an impedance network of the DC side of the inverter and an inductor current reference value, and calculate an inductor current error; The inductor current error is input to an outer-loop PI controller to obtain a given value of a grid-connected current; A grid-connected current tracking control module is configured to calculate a grid-connected current error from the given value of the grid-connected current and a measured value of the grid-connected current, and output a given value of a capacitor voltage after the grid-connected current error is adjusted by the middle-loop passive sliding mode controller; A capacitor voltage tracking control module is configured to calculate a capacitor voltage error from the given value of the capacitor voltage and a measured value of the capacitor voltage, and output a reference value of an inverter output current after the capacitor voltage error is adjusted by the inner-loop passive controller; A prediction module is configured to obtain an ambient temperature and an illumination intensity, input them to a pre-trained GRU neural network to obtain a maximum power point output current reference value of a photovoltaic array, calculate a photovoltaic array output current error from the maximum power point output current reference value of the photovoltaic array and a measured value of a photovoltaic array output current, and input the photovoltaic array output current error to a PI controller to obtain an open-circuit duty ratio; A bridge arm switch control module is configured to generate PWM switch signals of each bridge arm by SVPWM modulation technology according to the reference value of the inverter output current and the open-circuit duty ratio, drive each bridge arm switch tube to turn on and off, and realize grid-connected control of the inverter.

10. An electronic terminal comprising a processor and a storage medium; the storage medium is configured to store instructions; characterized in that, The processor is configured to operate according to the instructions to perform the steps of the current-mode quasi-Z-source photovoltaic grid-connected inverter control method of any one of claims 1 to 8.

Images