A method and system for regulating photovoltaic array power generation efficiency

Abstract

This invention relates to the field of photovoltaic power generation operation monitoring and electrical control technology, and particularly to a method and system for improving the power generation efficiency of photovoltaic arrays. The method includes: simultaneously acquiring non-stationary wind speed sequences, photothermal parameters, high-frequency DC current, and DC voltage sequences; constructing a three-dimensional time-frequency energy spectrum by performing empirical mode decomposition and Hilbert transform on the high-frequency DC current; targeting and eliminating flutter eigenmodes in the energy spectrum to reconstruct a pure electrical sequence by combining the wind speed calculation with the wind load disturbance frequency band; substituting the pure electrical sequence and voltage sequence into a fractional-order impedance differential equation containing constant-phase elements to establish an anomalous dynamic mapping relationship; solving this mapping to obtain the fractional order, and outputting hidden crack distortion parameters when the value falls below a preset benchmark; calculating the maximum Lyapunov exponent of the thermodynamic system based on this, and using the negative value range as a constraint to solve for the safety clamping voltage and issue control commands. This invention achieves precise decoupling of wind load disturbances and can actively avoid the risk of module thermal runaway.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation operation monitoring and electrical control technology, and in particular to a regulation method and system for improving the power generation efficiency of photovoltaic arrays. Background Technology

[0002] As the photovoltaic industry expands into high-latitude and complex climate regions, the refined operation and maintenance of large-scale demonstration photovoltaic bases faces severe challenges. Traditional photovoltaic array fault monitoring and efficiency control schemes are mostly based on the assumption of a stable environment. They typically use time-domain averaging filtering, static integer-order equivalent circuit models for power attenuation assessment, and rely on fixed physical temperature thresholds to implement passive power outage protection.

[0003] However, traditional photovoltaic regulation mostly adopts steady-state filtering and static protection. Due to the difficulty in decoupling mechanical interference and the lack of thermal runaway prediction, problems such as misjudgment of microcracks and thermal runaway burnout of modules are caused. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a regulation method and system for improving the power generation efficiency of photovoltaic arrays. It aims to improve the problems caused by the fact that traditional photovoltaic regulation mostly adopts steady-state filtering and static protection, which is difficult to decouple mechanical interference and lacks thermal runaway prediction, thus causing misjudgment of microcracks and thermal runaway burnout of modules.

[0005] In a first aspect, the present invention provides the following technical solution: a method for regulating and improving the power generation efficiency of a photovoltaic array, comprising the following steps: S1. Synchronously collect the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array; S2. Perform empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals, and perform Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum; S3. Calculate the non-stationary wind speed sequence to obtain the wind load disturbance frequency band, remove the flutter eigenmodes that are isomorphic to the wind load disturbance frequency band from the three-dimensional time-frequency energy spectrum, and superimpose the remaining eigenmode functions with the monotonic residual to reconstruct a pure electrical sequence. S4. Substitute the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing constant phase elements to establish an anomalous dynamic mapping relationship; S5. Solve the anomalous dynamic mapping relationship to obtain the fractional order. When the fractional order is less than the preset benchmark value, output the hidden crack distortion parameter. S6. Based on the hidden crack distortion parameters and the photothermal parameters, calculate the maximum Lyapunov exponent of the thermodynamic system, solve for the safety clamping voltage with the maximum Lyapunov exponent being in the negative range as a constraint, and issue a control command to the inverter to operate at the safety clamping voltage.

[0006] By adopting the above technical solution, wind flutter is eliminated in the frequency domain and the parameters of hidden cracks are extracted by solving the anomalous mapping. Then, the clamping voltage is issued based on the chaotic boundary, thereby improving the problem that traditional photovoltaic regulation mostly adopts steady-state filtering and static protection, which is difficult to decouple mechanical interference and lacks thermal runaway prediction, thus causing misjudgment of hidden cracks and thermal runaway burnout of modules.

[0007] Optionally, in S1, the synchronous acquisition of the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array includes: Using three-dimensional ultrasonic anemometers and high-precision solar intensity meters deployed at the characteristic spatial nodes of the target photovoltaic array, non-stationary wind speed sequences and photothermal parameters are collected under the triggering of a unified clock synchronization signal. Using a high-frequency Hall current sensor connected in series at the output of the target photovoltaic array combiner box and a voltage sensor connected in parallel at the output of the combiner box, a high-frequency DC current sequence and a DC voltage sequence are acquired under the trigger of the clock synchronization signal. The non-stationary wind speed sequence, the solar thermal parameters, the high-frequency DC current sequence, and the DC voltage sequence are subjected to timestamp alignment and outlier removal to generate a multimodal synchronization dataset.

[0008] Optionally, in S2, performing empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals includes: Extract all local maxima and local minima of the high-frequency DC current sequence, and use cubic spline interpolation function to fit the local maxima and local minima respectively to construct upper and lower envelopes; Calculate the local mean sequence of the upper envelope and the lower envelope, and subtract the local mean sequence from the high-frequency DC current sequence to obtain the first intermediate sequence; Verify whether the first intermediate sequence satisfies the envelope condition that the number of extreme points is equal to the number of zero-crossing points and the symmetry condition that the local mean of the upper and lower envelopes is zero. If the envelope condition and the symmetry condition are satisfied, the first intermediate sequence is extracted as an intrinsic mode function, and the intrinsic mode function is subtracted from the high-frequency DC current sequence to obtain the remaining sequence. The remaining sequence is used as a new original sequence to repeat the decomposition operation until the remaining sequence becomes a monotonic residual without extreme points.

[0009] Optionally, in S2, performing a Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum includes: Perform a quadrature phase shift operation on each of the intrinsic mode functions to construct the analytic signal corresponding to each of the intrinsic mode functions; Extract the instantaneous amplitude and instantaneous phase of the analytical signal, and obtain the instantaneous frequency by performing a derivative operation on the instantaneous phase with respect to the time variable; Using the time series as the horizontal axis, the instantaneous frequency as the vertical axis, and the square of the instantaneous amplitude as the vertical axis, the instantaneous frequencies and instantaneous amplitudes of all the intrinsic mode functions are mapped to a three-dimensional coordinate system to construct a three-dimensional time-frequency energy spectrum.

[0010] Optionally, in S3, calculating the non-stationary wind speed sequence to obtain the wind load disturbance frequency band includes: Calculate the long-term trend component of the non-stationary wind speed sequence, and subtract the long-term trend component from the non-stationary wind speed sequence to obtain the time-domain fluctuating wind speed component. The autocorrelation function is calculated on the time-domain fluctuating wind speed component, and the fast Fourier transform is performed on the obtained autocorrelation function to generate the turbulent power spectral density distribution matrix. Extract the center frequency of the main energy peak in the turbulent power spectral density distribution matrix, and define the frequency range based on the center frequency of the main energy peak and a preset frequency band broadening coefficient. Set the frequency range as the wind load disturbance frequency band.

[0011] Optionally, in S3, removing flutter eigenmodes isomorphic to the wind-borne disturbance frequency band from the three-dimensional time-frequency energy spectrum, and superimposing the remaining eigenmode functions with the monotonic residuals to reconstruct a pure electrical sequence includes: Calculate the energy centroid frequency of each of the intrinsic mode functions in the three-dimensional time-frequency energy spectrum; The energy centroid frequency is matched and verified with the upper and lower limits of the wind load disturbance frequency band; If the energy centroid frequency falls within the wind load disturbance frequency band, the corresponding intrinsic mode function is determined to be the flutter intrinsic mode caused by the aerodynamic flutter of the support frame, and the instantaneous amplitude sequence of the flutter intrinsic mode is forcibly set to zero and removed. The intrinsic mode function set after amplitude zeroing is linearly summed and reconstructed with the monotonic residual to generate a clean electrical sequence that eliminates aerodynamic flutter interference.

[0012] Optionally, in S4, substituting the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing a constant-phase element to establish an anomalous dynamic mapping relationship includes: Construct a fractional-order equivalent circuit topology for a photovoltaic module that includes a photocurrent source, diodes, and parallel constant-phase elements; Based on the fractional equivalent circuit topology, Kirchhoff's current law is used to construct nodal equations, and a fractional differential operator is introduced to perform derivative operations on the voltage variables to characterize the charging and discharging characteristics of the constant phase element. The pure electrical sequence is input into the node equation as the total output current variable, and the DC voltage sequence is input into the node equation as the terminal voltage variable, thus establishing an anomalous dynamic mapping relationship containing unknown fractional orders.

[0013] Optionally, in S5, the fractional order is obtained by solving the anomalous dynamic mapping relationship. When the fractional order is less than a preset benchmark value, the output hidden crack distortion parameters include: The nonlinear damped least squares iterative algorithm is used to identify the parameters of the anomalous dynamic mapping relationship and extract the converged fractional order. Obtain a preset reference value to characterize the standard Gaussian diffusion of charge carriers in a lossless silicon wafer; Compare the extracted fractional order with the preset benchmark value. If the fractional order drops and becomes less than the preset benchmark value, it is determined that there is a carrier sub-diffusion hindrance phenomenon inside the silicon wafer. Based on the difference between the preset benchmark value and the fractional order, a hidden crack distortion parameter is generated to quantify the severity of the hidden crack, and the hidden crack distortion parameter is output.

[0014] Optionally, in S6, the maximum Lyapunov exponent of the thermodynamic system is calculated based on the microcrack distortion parameters and the photothermal parameters. The safety clamping voltage is solved using the constraint that the maximum Lyapunov exponent is in the negative range. The control command to be issued to the inverter to operate at the safety clamping voltage includes: Using the microcrack distortion parameters and the photothermal parameters as state variables, the nonlinear phase space of the thermodynamic system of the target photovoltaic array is reconstructed. A virtual thermal perturbation is applied in the nonlinear phase space, and the maximum logarithmic rate of change of the divergence distance of the adjacent state trajectories over time is tracked and calculated to obtain the maximum Lyapunov exponent. The constraint condition for the thermodynamic system to maintain a convergent and stable state is that the maximum Lyapunov exponent is always in the negative range; When the maximum Lyapunov exponent is detected to be increasing and approaching the zero bifurcation boundary, the maximum Lyapunov exponent is kept negative as the objective function, and the reverse search algorithm is used to optimize and calculate the safety clamping voltage that limits thermal runaway. A voltage limit command containing the safety clamp voltage is generated, and the voltage limit command is sent to the inverter as a control command to perform maximum power point tracking deviation action.

[0015] Secondly, the present invention provides the following technical solution: a control system for improving the power generation efficiency of a photovoltaic array, comprising the following modules: A multimodal non-stationary parameter acquisition module is used to simultaneously acquire non-stationary wind speed sequences, photothermal parameters, high-frequency DC current sequences, and DC voltage sequences of the target photovoltaic array; The time-frequency energy spectrum construction module is used to perform empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals, and to perform Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum. The pure electrical sequence reconstruction module is used to calculate the wind load disturbance frequency band of the non-stationary wind speed sequence, remove flutter eigenmodes that are isomorphic to the wind load disturbance frequency band from the three-dimensional time-frequency energy spectrum, and superimpose the remaining eigenmode functions with the monotonic residual to reconstruct the pure electrical sequence. An anomalous dynamic mapping establishment module is used to substitute the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing constant phase elements to establish an anomalous dynamic mapping relationship. The hidden crack distortion parameter identification module is used to solve the anomalous dynamic mapping relationship to obtain the fractional order. When the fractional order is less than a preset benchmark value, the hidden crack distortion parameter is output. The anti-chaotic bifurcation collaborative control module is used to calculate the maximum Lyapunov exponent of the thermodynamic system based on the hidden crack distortion parameters and the photothermal parameters, solve the safety clamping voltage with the maximum Lyapunov exponent being in the negative range as a constraint, and issue a control command to the inverter to operate at the safety clamping voltage.

[0016] The present invention has the following beneficial effects: 1. In this invention, by eliminating wind-borne flutter in the frequency domain and solving the anomalous mapping to extract the parameters of hidden cracks, and then issuing clamping voltage based on the chaotic boundary, the problem of traditional photovoltaic regulation, which mostly adopts steady-state filtering and static protection, is improved due to the difficulty in decoupling mechanical interference and the lack of thermal runaway prediction, resulting in misjudgment of hidden cracks and thermal runaway burnout of modules.

[0017] 2. In this invention, by constructing a fractional-order impedance differential equation containing constant-phase elements and extracting the fractional order, the sub-diffusion retardation phenomenon of charge carriers can be accurately quantified. This improves the problem that traditional photovoltaic modeling mostly uses ideal integer-order capacitors, which cannot reflect the non-local variation of microcracks, resulting in underreporting of micro-damage and inaccurate attenuation quantification.

[0018] 3. In this invention, flutter eigenmodes are eliminated by performing empirical mode decomposition and turbulent power spectrum calculation, thereby decoupling mechanical vibration from electrical signals. This improves the problem that traditional signal denoising, which mostly uses global static Fourier transform, is prone to local abrupt feature smoothing dissipation, resulting in incomplete high-frequency ripple cleaning and loss of effective signals. Attached Figure Description

[0019] Figure 1 This is a flowchart of a control method for improving the power generation efficiency of a photovoltaic array proposed in this invention; Figure 2 This is an architecture diagram of a control system for improving the power generation efficiency of a photovoltaic array, as proposed in this invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: In a first embodiment of the present invention, the present invention provides a method for regulating and improving the power generation efficiency of a photovoltaic array, such as... Figure 1 As shown, it includes the following steps: S1. Synchronously collect the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array; Furthermore, in S1, the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array are simultaneously acquired, including: Using three-dimensional ultrasonic anemometers and high-precision solar intensity meters deployed at the characteristic spatial nodes of the target photovoltaic array, non-stationary wind speed sequences and photothermal parameters are collected under the triggering of a unified clock synchronization signal. Using a high-frequency Hall current sensor connected in series at the output of the target photovoltaic array combiner box and a voltage sensor connected in parallel at the output of the combiner box, high-frequency DC current sequence and DC voltage sequence are acquired under the triggering of a clock synchronization signal. Timestamp alignment and outlier removal are performed on non-stationary wind speed sequences, solar thermal parameters, high-frequency DC current sequences, and DC voltage sequences to generate a multimodal synchronization dataset.

[0022] Specifically, in step S1, the core objective is to establish a rigorous time-domain mapping relationship between underlying meteorological physical quantities and electrical parameters, preventing blind source separation failure due to phase misalignment in the subsequent frequency domain decoupling algorithm. The input data for this step consists of the physical wind field, radiation field, and raw output power of the photovoltaic module ports in the natural environment of the target photovoltaic array. The output result is a multimodal synchronization dataset matrix that has been cleaned of outliers and aligned with the time axis.

[0023] Conventional meteorological monitoring often employs mechanical anemometers. The inherent rotational inertia of these mechanical structures prevents them from capturing the high-frequency turbulent shear that causes minute aerodynamic flutter in photovoltaic (PV) supports. This technical solution deploys three-dimensional ultrasonic anemometers and high-precision solar intensity meters at characteristic spatial nodes such as the vertices or diagonal intersections of the target PV array. The three-dimensional ultrasonic anemometer directly calculates the wind speed vector using the time difference between the propagation of ultrasonic pulses with and against the wind, eliminating mechanical hysteresis, and is used to collect non-stationary wind speed sequences and solar thermal parameters.

[0024] On the electrical monitoring side, conventional photovoltaic centralized monitoring systems typically have data sampling rates on the order of seconds or even minutes. This low-frequency sampling directly and smoothly filters out the high-frequency ripple of transient currents caused by the mechanical vibration of the support structure transmitted into the electrical circuit. Therefore, a high-frequency Hall current sensor must be connected in series at the combiner box output, and a voltage sensor in parallel, to capture high-frequency DC current and DC voltage sequences at a millisecond-level sampling rate.

[0025] The meteorological and electrical sensors are directly connected to the same trigger module with hard-wired timing functionality. At the instant a unified clock synchronization signal is issued, the ultrasonic anemometer latches the three-dimensional spatial components of the wind speed, and the Hall current sensor synchronously latches the DC bus current amplitude. The moment of transient gust impact in the physical space and the moment of current fluctuation in the electrical circuit are forcibly bound to the same timestamp reference.

[0026] After the underlying data acquisition is completed, the cleaning process begins. High-frequency electrical sequences are prone to pulse spikes caused by electromagnetic interference during transmission. Therefore, the Raida criterion is used to remove outliers. Let any one-dimensional discrete original sampling sequence be... , containing elements subscript This represents the sequence number of the sampling time point. Calculate the sequence. arithmetic mean and standard deviation The mathematical expression is: ; ; in This represents the total number of sampling points for the discrete sequence within the current calculation window. The absolute deviation of each sampling point element is calculated; if it exists... In this case, determine the current sampled element. Abnormal spikes caused by high-frequency electromagnetic noise are removed from the original sequence, and the arithmetic mean of adjacent normal sampling points is extracted for interpolation to fill them.

[0027] After outlier cleaning, the meteorological and electrical data sequences are aligned and stitched together based on the timestamps assigned by the clock synchronization signal to construct a multimodal synchronization dataset matrix. : ; in represent The multimodal synchronization dataset matrix at each time step. represent The three-dimensional non-stationary wind speed values ​​at any given time. represent The value of light and heat radiation at any given time. represent The value of high-frequency DC current at any given time. represent The DC voltage value at time t. This matrix set serves as the system's sole underlying standard data source, directly passed to the subsequent Empirical Mode Decomposition (EMD) module as the computational basis.

[0028] S2. Perform empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals, and perform Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum; Furthermore, in S2, empirical mode decomposition is performed on the high-frequency DC current sequence to obtain the eigenmode functions and monotonic residuals, including: Extract all local maxima and local minima of the high-frequency DC current sequence, and use cubic spline interpolation function to fit the local maxima and local minima respectively to construct the upper envelope and lower envelope; Calculate the local mean sequences of the upper and lower envelopes, and subtract the local mean sequences from the high-frequency DC current sequence to obtain the first intermediate sequence; Verify whether the first intermediate sequence satisfies the envelope condition that the number of extreme points is equal to the number of zero-crossing points, and the symmetry condition that the local mean of the upper and lower envelopes is zero. If the envelope condition and the symmetry condition are satisfied, the first intermediate sequence is extracted as the intrinsic mode function, and the high-frequency DC current sequence is subtracted from the intrinsic mode function to obtain the residual sequence. The residual sequence is used as the new original sequence to repeat the decomposition operation until the residual sequence becomes a monotonic residual without extreme points.

[0029] In S2, performing a Hilbert transform on the eigenmode functions to construct the three-dimensional time-frequency energy spectrum includes: Perform a quadrature phase shift operation on each eigenmode function to construct the analytic signal corresponding to each eigenmode function; Extract the instantaneous amplitude and instantaneous phase of the analytical signal, and obtain the instantaneous frequency by taking the derivative of the instantaneous phase with respect to the time variable; Using time series as the horizontal axis, instantaneous frequency as the vertical axis, and the square of instantaneous amplitude as the vertical axis, the instantaneous frequency and instantaneous amplitude of all intrinsic mode functions are mapped to a three-dimensional coordinate system to construct a three-dimensional time-frequency energy spectrum.

[0030] Specifically, in step S2, the input data stream originates from the high-frequency DC current sequence acquired and cleaned in step S1. This current sequence is unidirectionally input to the empirical mode decomposition module, which outputs a set of intrinsic mode functions and monotonic residuals. The set of intrinsic mode functions is then passed to the Hilbert transform operator for phase differentiation, ultimately outputting a three-dimensional time-frequency energy spectrum data matrix for use by the next-level wind-borne frequency band elimination module.

[0031] The mechanical and aerodynamic flutter of photovoltaic (PV) supports caused by strong winds is a non-stationary random process. The slight deflection of the PV panel's tilt angle caused by this flutter couples into the electrical circuit, forming a high-frequency alternating ripple. This ripple deeply overlaps with the irreversible intrinsic decay signal caused by microcracks within the silicon wafer in the time domain. Traditional Fourier transforms are only applicable to stationary signals; forced application leads to the smooth dissipation of true frequency characteristics over time, failing to identify sudden wind-induced disturbances. Introducing empirical mode decomposition (EMD) can eliminate the need for preset basis functions, adaptively separating the current signal into multiple broadband components with single physical meaning based on its local time scale. Furthermore, using Hilbert transform to extract instantaneous frequencies allows the originally one-dimensional overlapping current time-series signal to be distributed into a three-dimensional geometric space composed of time, frequency, and energy, completely decoupling the mechanical vibration parameters from the electrical decay parameters mathematically.

[0032] The input high-frequency DC current sequence is set as follows: The time-independent variables are Traverse the sequence to find The local maxima and local minima are identified. A cubic spline interpolation function is used to fit the extreme points, generating an upper envelope covering the top of the signal. And the lower envelope of the support signal. Calculate the local mean sequence of the two envelopes. : Subtract this local mean sequence from the original high-frequency DC current sequence to obtain the first intermediate sequence. : Then, perform intrinsic modal condition verification. Requirements: Within the data length, the total number of extreme points is equal to or differs from the total number of zero-crossing points by at most one, and the mean of the envelope defined by the local maxima and local minima at any time point must be strictly equal to zero. If the condition is not met, [the data will be processed]. This is treated as a repeated envelope fitting and averaging operation on the original sequence to be processed. If the condition is met, then... Extracted as first-order intrinsic mode functions .

[0033] Subtract the first-order intrinsic mode function from the high-frequency DC current sequence to calculate the remaining sequence. : ; Will Enter the cyclic decomposition operation. When the... The remaining sequence generated by the next decomposition The iterative calculation terminates when the function exhibits monotonic characteristics without extreme points. The high-frequency DC current sequence is reconstructed as follows: ; in This represents the total order of the intrinsic mode functions obtained through iteration. Represents the natural index of the modal order. Representing the eigenmode functions of order 1 This represents the monotonic residual quantity that reflects the DC noise floor and long-term temperature drift trend.

[0034] For any isolated first-order eigenmode function Perform the Hilbert transform and calculate the corresponding orthogonal phase shift function. : ;in This represents the dimensionless integration variable during the time-domain integration process. A complex-domain analytic signal is constructed using the antiderivative and the orthogonal phase-shift function. : ;in Represents the imaginary unit. Substitute this into the formula for the modulus and argument of a complex number to calculate the instantaneous amplitude. and instantaneous phase : ; ; For instantaneous phase Regarding time variables Perform first derivative calculation to obtain instantaneous frequency. : ; The instantaneous amplitude and frequency variations of all intrinsic mode functions are projected onto a three-dimensional coordinate system. A time frame is set. The horizontal axis represents the instantaneous frequency. The vertical axis is defined by the square of the instantaneous amplitude, and the alternating current energy is represented by the square of the instantaneous amplitude. A three-dimensional time-frequency energy spectrum is defined. : ; in This represents the real part parameter of the extracted complex function. This equation maps a one-dimensional time-series signal into a three-dimensional data matrix with frequency evolution and time-tracking properties.

[0035] S3. Calculate the non-stationary wind speed sequence to obtain the wind load disturbance frequency band, remove the flutter eigenmodes that are isomorphic to the wind load disturbance frequency band in the three-dimensional time-frequency energy spectrum, and superimpose the remaining eigenmode functions with the monotonic residuals to reconstruct the pure electrical sequence. Furthermore, in S3, the calculation of non-stationary wind speed sequences to obtain wind load disturbance frequency bands includes: Calculate the long-term trend component of the non-stationary wind speed sequence, and subtract the long-term trend component from the non-stationary wind speed sequence to obtain the time-domain fluctuating wind speed component. Autocorrelation function calculation is performed on the time-domain fluctuating wind speed component, and fast Fourier transform is performed on the obtained autocorrelation function to generate the turbulent power spectral density distribution matrix; Extract the center frequency of the main energy peak in the turbulent power spectral density distribution matrix, and define the frequency range based on the center frequency of the main energy peak and the preset frequency band broadening coefficient. Set the frequency range as the wind load disturbance frequency band.

[0036] In S3, flutter eigenmodes isomorphic to the wind-borne disturbance frequency band are removed from the three-dimensional time-frequency energy spectrum. The remaining eigenmode functions are superimposed with monotonic residuals to reconstruct a pure electrical sequence, including: Calculate the energy centroid frequency of each intrinsic mode function in the three-dimensional time-frequency energy spectrum; The energy centroid frequency is matched and verified with the upper and lower limits of the wind load disturbance frequency band; If the energy center frequency falls within the wind load disturbance frequency band, the corresponding intrinsic mode function is determined to be the flutter intrinsic mode caused by the aerodynamic flutter of the support frame, and the instantaneous amplitude sequence of the flutter intrinsic mode is forcibly set to zero and eliminated. The intrinsic mode function set after amplitude zeroing is reconstructed by linear summation with the monotonic residual to generate a clean electrical sequence that eliminates aerodynamic flutter interference.

[0037] Specifically, the data input stream in step S3 includes the non-stationary wind speed sequence extracted in the previous step, as well as the three-dimensional time-frequency energy spectrum, intrinsic mode functions, and monotonic residuals constructed in the previous step. The data output stream is the clean electrical sequence generated by reconstruction and passed to the next-level module. This step aims to establish a cross-frequency domain mapping determination mechanism between meteorological physical fields and electrical micro-signals, and to achieve bottom-level signal cleaning by targeted removal of aeromechanical disturbances.

[0038] Extract the long-term trend component of the non-stationary wind speed sequence. Subtract the long-term trend component from the non-stationary wind speed sequence to extract the time-domain fluctuating wind speed component representing sudden gusts. Let the original non-stationary wind speed sequence be... The long-term trend component is The time-domain pulsating wind speed component is ,parameter Representing the time variable. The wind speed components satisfy the absolute difference equation: Perform autocorrelation function calculation on the time-domain fluctuating wind speed components. Set the time delay variable as... The integration time window length is Calculate the autocorrelation function : Perform a Fast Fourier Transform on the autocorrelation function to project the time-domain parameters into the frequency domain, generating the turbulent power spectral density distribution matrix. , in the formula Represents frequency variables. Represents the imaginary unit: ; Extract the energy peak with the highest amplitude from the turbulent power spectral density distribution matrix. Identify the frequency point corresponding to the energy peak as its center frequency. Introducing a preset bandwidth broadening factor. Calculate the bandwidth expansion. : Based on the center frequency of the main energy peak, a lower limit is defined. and upper limit The frequency range is defined as the wind disturbance frequency band.

[0039] In the three-dimensional time-frequency energy spectrum, the energy centroid frequency is calculated for each eigenmode function. : ; in The order index of the intrinsic mode function. This represents the instantaneous energy density distribution term of the corresponding order eigenmode function in the frequency domain.

[0040] The energy centroid frequency of each intrinsic mode function is compared numerically with the upper and lower limits of the wind load disturbance frequency band. If the energy centroid frequency falls within the wind load disturbance frequency band, the corresponding intrinsic mode function is determined to be a flutter intrinsic mode caused by the coupling of the support aerodynamic flutter to the electrical circuit. For the determined flutter intrinsic mode, the instantaneous amplitude sequence of the flutter intrinsic mode is forcibly set to zero in the data matrix.

[0041] Extract the intrinsic mode function set after amplitude zeroing, and perform a linear summation operation on it along with the monotonic residuals to reconstruct a clean electrical sequence that eliminates aerodynamic flutter interference. : ; in Represents the total modal order. This represents the first [unit / item] after the amplitude is set to zero and filtered. eigenmode functions of order 1 It represents the monotonic residual quantity.

[0042] By calculating the wind load disturbance frequency band of the non-stationary wind speed sequence and targeting the zero-isomorphic flutter eigenmode in the three-dimensional time-frequency energy spectrum, the traditional technical defect of the signal processing method cannot accurately separate broadband mechanical vibration interference from the real micro-electrical degradation signal of photovoltaic modules. This avoids the problem of misjudging the false fluctuations of physical flutter of the support under extreme wind conditions as irreversible microcrack attenuation inside the photovoltaic array.

[0043] S4. Substitute the pure electrical sequence and DC voltage sequence into the fractional impedance differential equation containing constant phase elements to establish an anomalous dynamic mapping relationship. Furthermore, in S4, the pure electrical sequence and DC voltage sequence are substituted into the fractional impedance differential equation containing a constant-phase element to establish an anomalous dynamic mapping relationship, including: Construct a fractional-order equivalent circuit topology for a photovoltaic module that includes a photocurrent source, diodes, and parallel constant-phase elements; Based on the fractional equivalent circuit topology, Kirchhoff's current law is used to construct the node equations, and a fractional differential operator is introduced to perform derivative operations on the voltage variables to characterize the charging and discharging characteristics of the constant phase element. By inputting the pure electrical sequence as the total output current variable into the nodal equations and the DC voltage sequence as the terminal voltage variable into the nodal equations, an anomalous dynamic mapping relationship containing unknown fractional orders is established.

[0044] Specifically, the data input stream of step S4 consists of the clean electrical sequence reconstructed from the previous steps and the DC voltage sequence acquired in real time by the underlying hardware. The data output stream is the constructed anomalous dynamic mapping equation containing an unknown fractional order, which is passed to the microcrack distortion parameter identification module as the model to be solved. This step aims to completely remove the unconventional electrical distortion characteristics caused by microcracks from the underlying physical model level.

[0045] Traditional photovoltaic equivalent circuit models often use ideal integer-order capacitance to simulate the depletion layer capacitance and diffusion capacitance inside the PN junction. However, in extreme high-latitude climates, silicon wafers experience alternating freeze-thaw stress, leading to physical microcracks that impede the transport paths of internal charge carriers and trap them. The carrier transport mechanism deviates significantly from the classical Gaussian diffusion law, exhibiting sub-diffusion physical characteristics with spatial nonlocality and time-memory dependence. Conventional integer-order ordinary differential equations are completely incapable of describing this anomalous relaxation process.

[0046] To address this physical phenomenon, the proposed solution abandons the ideal capacitor model and instead uses constant-phase elements to construct a fractional-order equivalent circuit topology for photovoltaic modules. This topology consists of a photocurrent source, diode branches, and constant-phase element branches connected in parallel, with each parallel branch connected in series with an equivalent series resistance.

[0047] Based on the fractional-order equivalent circuit topology, Kirchhoff's current law is applied to establish nodal equilibrium equations for the internal parallel nodes. The total output current equals the current generated by the photogenerated current source minus the current flowing through the diode branch and the charging / discharging current flowing through the constant-phase element branch. The reconstructed, clean electrical sequence after removing mechanical chatter is substituted as the total output current variable, and the measured DC voltage sequence is substituted as the terminal voltage variable. A fractional-order differential operator describing anomalous diffusion is introduced to characterize the dynamic characteristics of the constant-phase element, establishing the following anomalous dynamic mapping relationship: ; in Represents a variable over time A changing pure electrical sequence. This represents the photocurrent generated by a photovoltaic module when it is excited by sunlight. This represents the reverse saturation current of the diode. It represents the fundamental electron charge constant. Represents a variable over time A changing sequence of DC voltages. This represents the equivalent series resistance inside a photovoltaic module. This represents the ideal factor of a diode. This represents the Boltzmann constant. This represents the current absolute physical temperature of the photovoltaic module. The pseudocapacitance proportional coefficient representing a constant-phase element. The order is represented as Fractional differential operators. This is the unknown fractional order that characterizes the degree of subdiffusion of charge carriers, and its value is strictly limited to the range of real numbers greater than zero and less than or equal to one.

[0048] This step maps the purified electrical signal after aerodynamic decoupling to a fractional-order differential space, thus visualizing the microstructural variations caused by microcracks as fractional-order variables to be solved in mathematical equations. This overcomes the technical deficiency of existing steady-state diagnostic models, which are prone to failure under non-stationary photothermal alternating conditions, and provides a theoretical calculation foundation with a unique solution for accurately identifying the true physical damage state inside the photovoltaic array.

[0049] S5. Solve the anomalous dynamic mapping relationship to obtain the fractional order. When the fractional order is less than the preset benchmark value, output the hidden crack distortion parameters. Furthermore, in S5, the fractional order is obtained by solving the anomalous dynamic mapping relationship. When the fractional order is less than a preset benchmark value, the output hidden crack distortion parameters include: A nonlinear damped least squares iterative algorithm is used to identify parameters of the anomalous dynamic mapping relationship and extract the fractional order after convergence. Obtain a preset reference value to characterize the standard Gaussian diffusion of charge carriers in a lossless silicon wafer; Compare the extracted fractional order with the preset benchmark value. If the fractional order drops and is less than the preset benchmark value, it is determined that there is a carrier sub-diffusion hindrance phenomenon inside the silicon wafer. Based on the difference between the preset baseline value and the fractional order, a microcrack distortion parameter is generated to quantify the severity of the microcrack, and the microcrack distortion parameter is output.

[0050] Specifically, the data input stream for step S5 consists of the anomalous dynamic mapping relationship containing unknown parameters and the related discrete sampling sequence passed from the previous steps. The data output stream consists of the microcrack distortion parameters that quantify the degree of physical damage to the silicon wafer. This step transforms the previous mathematical differential model into specific equipment health status criteria, and extracts the quantitative characteristics reflecting microscopic lattice defects through nonlinear solutions, providing triggering conditions for subsequent top-level control.

[0051] A parameter identification mechanism is established to address anomalous dynamic mapping relationships. Since the model contains derivative and exponential terms of non-integer order, a nonlinear damped least squares iterative algorithm is employed to fit and solve the equations. An objective function is constructed with the sum of squared residuals as its core. : ; in This represents a vector of unknown parameters containing the fractional order to be solved. This represents the total number of sampling points set. Representing the Each discrete sampling time. Representative at The measured values ​​of the pure electrical sequence are collected and reconstructed at all times. This represents the theoretical current value calculated under the current parameter vector based on the anomalous dynamic mapping relationship.

[0052] Introducing a damping factor to control the iterative convergence direction and step size, and establishing the parameter iterative update matrix equation: ; in Representing the Add a parameter vector for each iteration. Representing the The parameter vector for the next iteration. This represents the Jacobian matrix of the theoretical model with respect to the parameter vector. This represents the transpose of the Jacobian matrix. The damping factor represents the control algorithm's switching between gradient descent and Gauss-Newton's method. This represents a diagonal matrix formed by multiplying the transpose of the Jacobian matrix by itself and its main diagonal elements. Representing the The residual vector generated by the next iteration. The superscript negative one represents the inverse operation of the matrix.

[0053] The algorithm executes a loop iteration until the objective function is reached. The value is less than the set minimum tolerance or the maximum number of iterations is reached. From the finally converged parameter vector... Extract the actual fractional order of the current target photovoltaic array .

[0054] The carrier transport within a intact photovoltaic silicon wafer follows the standard Gaussian diffusion law, and theoretically, the order of the equivalent constant-phase element should approach one. A preset benchmark value is set to characterize the standard Gaussian diffusion of carriers in a lossless silicon wafer. The value is usually taken as one. The actual fractional order extracted will be... Compared with the preset benchmark value Perform numerical comparisons.

[0055] When photovoltaic modules are subjected to alternating freeze-thaw stress in frigid environments, resulting in physical microcracks, the continuity of the internal crystal lattice is disrupted. The numerous newly created defect energy levels trap and delay free charge carriers, causing the macroscopic trajectory of the charge carriers to degenerate from normal Brownian motion into a subdiffusion-impeded process. This physical structural variation directly manifests as a drop in the order parameter in fractional-order differential models. If the actual fractional order is determined... Less than the preset benchmark value If this is confirmed, then the sub-diffusion retardation of charge carriers inside the target silicon wafer is confirmed, that is, the presence of physical microcracks.

[0056] Generate hidden crack distortion parameters based on the difference between the two. : ;in This parameter represents the microcrack distortion and is a dimensionless real number. The greater the deviation of the actual fractional order from the preset benchmark value, the larger the value of this microcrack distortion parameter, indicating a wider distribution of the microcrack network inside the silicon wafer and a more severe obstruction of carrier sub-diffusion. This parameter is output to the collaborative control module as a state trigger quantity, avoiding the technical blind spot of conventional monitoring systems that cannot qualitatively determine the root cause of microcracks by relying solely on macroscopic power reduction.

[0057] S6. Based on the microcrack distortion parameters and photothermal parameters, calculate the maximum Lyapunov exponent of the thermodynamic system. Solve for the safety clamping voltage with the maximum Lyapunov exponent in the negative range as a constraint, and issue a control command to the inverter to operate at the safety clamping voltage. Furthermore, in S6, based on the microcrack distortion parameters and photothermal parameters, the maximum Lyapunov exponent of the thermodynamic system is calculated. The safety clamping voltage is solved using the constraint that the maximum Lyapunov exponent is in the negative range. The control commands issued to the inverter to operate at the safety clamping voltage include: Using the microcrack distortion parameters and photothermal parameters as state variables, the nonlinear phase space of the thermodynamic system of the target photovoltaic array is reconstructed. By applying virtual thermal perturbation in the nonlinear phase space and tracking and calculating the maximum logarithmic rate of change of the divergence distance between adjacent state trajectories over time, the maximum Lyapunov exponent is obtained. The constraint condition for the thermodynamic system to maintain a convergent and stable state is that the maximum Lyapunov exponent is always in the negative range; When the maximum Lyapunov exponent is detected to be rising and approaching the zero bifurcation boundary, the maximum Lyapunov exponent is kept negative as the objective function, and the reverse search algorithm is used to optimize and calculate the safety clamping voltage that limits thermal runaway. Generate a voltage limit command containing a safety clamping voltage, and send the voltage limit command as a control command to the inverter to perform maximum power point tracking deviation action.

[0058] Specifically, the data input stream in step S6 consists of the microcrack distortion parameters output by the preceding analysis module and the real-time photothermal parameters collected by the underlying meteorological nodes. The data output stream consists of control commands containing safety clamping voltage values ​​sent to the underlying inverter equipment. This module receives the front-end micro-damage identification results and is responsible for performing physical isolation control to prevent thermal runaway on the macro-equipment side.

[0059] The microcrack distortion parameters and photothermal parameters representing environmental radiation and temperature are extracted as core state variables to reconstruct the multidimensional phase space of the nonlinear thermodynamic system of the target photovoltaic array. Within this phase space, a small virtual thermal perturbation is applied to the system's current initial state. With time parameter Continuously advance, tracking the dynamic evolution path of two adjacent state trajectories in phase space. Calculate the maximum divergence distance between the two trajectories. The logarithmic rate of change over time is used to obtain the maximum Lyapunov exponent for evaluating the degree of chaos in the system. The specific mathematical formula is as follows: ; in The maximum Lyapunov exponent represents the thermodynamic system. The time parameter representing the evolution of the phase space state trajectory. This represents the distance of the virtual thermal perturbation vector artificially applied in phase space at the initial moment. Represents the passage of time The maximum divergence vector distance between two adjacent state trajectories after evolution. The double vertical bar notation represents taking the norm of the spatial vector.

[0060] Nonlinear dynamics theory establishes that the sign property of the Lyapunov exponent directly determines the macroscopic stability boundary of the system. Here, a hard constraint is set for the thermodynamic system to maintain a convergent stable state: the maximum Lyapunov exponent must be constantly within a negative value range less than zero. The system is continuously monitored and calculated... Numerical trend. When the value shows a clear upward trend and approaches the zero-point bifurcation boundary, it indicates that the photovoltaic silicon wafer with hidden crack defects is about to cross the thermal equilibrium critical point and enter a chain thermal runaway state under the current high-radiation alternating environment.

[0061] After triggering the boundary warning, the system sets the maximum Lyapunov exponent as negative as the objective function for optimization, uses a back-search algorithm to deduce the electrical operating boundary that satisfies the stability conditions of the system, and calculates the safety clamping voltage that can limit abnormal heating of components. Based on the calculated safety clamping voltage A standard voltage limit command is generated. This command is then sent to the underlying inverter as the final electrical control signal. Upon receiving the control signal, the inverter is forced to deviate from its normal maximum power point tracking trajectory, and the DC-side operating voltage is constantly clamped at a safe clamping voltage level.

[0062] This control procedure overcomes the passive limitations of relying on static temperature threshold power-off protection for photovoltaic (PV) operation and maintenance. Addressing the unique engineering scenarios in harsh high-latitude environments where muddy roads prevent maintenance personnel from reaching the site promptly for troubleshooting, this solution utilizes chaotic dynamics theory to construct a feedforward prediction model. Before irreversible hot spot burns occur in faulty PV modules, the system proactively implements electrical derating and voltage clamping intervention. This closed-loop defense mechanism, transcending environmental characteristics and micro-electrical considerations, eliminates the physical risk of severe microcracks in modules suffering extreme damage due to localized overheating during prolonged maintenance gaps, thus safeguarding the bottom line of safe operation for large-scale PV array assets.

[0063] Example 2: In large-scale empirical photovoltaic (PV) base applications in high-latitude, frigid regions, the target PV array is subjected to long-term strong turbulent wind disturbances and severe freeze-thaw cycles. Furthermore, the main roads in the area become extremely muddy due to seasonal rainfall, creating a physical defect resolution delay window of several hours. The core technical challenge in this complex scenario is the deep time-frequency aliasing between the nonlinear aerodynamic flutter signal of the PV support under non-stationary turbulent wind excitation and the fractional-order anomalous impedance effect caused by microcracks within the silicon wafer. This renders conventional stationary signal decoupling methods completely ineffective, failing to accurately identify the true microscopic physical damage. Moreover, during this physical defect resolution delay window, existing control systems lack precise quantification of the chaotic evolution boundary of the nonlinear thermodynamic system. This makes defective PV modules highly susceptible to crossing thermal equilibrium bifurcation boundaries under continuous agitation in an alternating light and heat environment, potentially leading to chain-like thermal runaway or even physical burnout. To address these issues, this invention provides a control system for improving the power generation efficiency of PV arrays, the structure of which is as follows... Figure 2 As shown. The specific implementation process of this system is as follows: The multimodal acquisition module simultaneously acquires non-stationary meteorological and high-frequency electrical reference data. Conventional monitoring is prone to misjudging mechanical swaying of the support structure caused by extreme wind conditions as component failure. The time-frequency energy spectrum construction and pure electrical sequence reconstruction modules work together to accurately isolate flutter signals isomorphic to wind loads in the three-dimensional time-frequency domain and filter out aerodynamic interference noise.

[0064] To address microcracks in silicon wafers caused by freeze-thaw stress, an anomalous dynamic mapping module breaks through the limitations of conventional integer-order equivalent circuits. The microcrack distortion parameter identification module extracts the actual fractional order by solving differential equations containing constant-phase elements, quantifying the carrier sub-diffusion hindrance phenomenon caused by microscopic lattice defects, thus enabling the identification of irreversible physical damage.

[0065] Faced with maintenance delays caused by muddy conditions in the harsh environment, the anti-chaotic bifurcation collaborative control module takes into account the underlying hidden crack parameters and calculates the maximum Lyapunov exponent of the nonlinear thermodynamic system. Before a faulty component crosses the thermal equilibrium bifurcation boundary, the system proactively calculates and issues a safety clamping voltage command to the inverter to perform electrical derating, cutting off the physical path of chain thermal runaway induced by local abnormal heating and ensuring the safe operation of photovoltaic assets.

[0066] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for regulating and controlling the power generation efficiency of a photovoltaic array, characterized in that, Includes the following steps: S1. Synchronously collect the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array; S2. Perform empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals, and perform Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum; S3. Calculate the non-stationary wind speed sequence to obtain the wind load disturbance frequency band, remove the flutter eigenmodes that are isomorphic to the wind load disturbance frequency band from the three-dimensional time-frequency energy spectrum, and superimpose the remaining eigenmode functions with the monotonic residual to reconstruct a pure electrical sequence. S4. Substitute the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing constant phase elements to establish an anomalous dynamic mapping relationship; S5. Solve the anomalous dynamic mapping relationship to obtain the fractional order. When the fractional order is less than the preset benchmark value, output the hidden crack distortion parameter. S6. Based on the hidden crack distortion parameters and the photothermal parameters, calculate the maximum Lyapunov exponent of the thermodynamic system, solve for the safety clamping voltage with the maximum Lyapunov exponent being in the negative range as a constraint, and issue a control command to the inverter to operate at the safety clamping voltage.

2. The method for controlling and improving the power generation efficiency of a photovoltaic array according to claim 1, characterized in that, In S1, the synchronous acquisition of the non-stationary wind speed sequence, photothermal parameters, high-frequency DC current sequence, and DC voltage sequence of the target photovoltaic array includes: Using three-dimensional ultrasonic anemometers and high-precision solar intensity meters deployed at the characteristic spatial nodes of the target photovoltaic array, non-stationary wind speed sequences and photothermal parameters are collected under the triggering of a unified clock synchronization signal. Using a high-frequency Hall current sensor connected in series at the output of the target photovoltaic array combiner box and a voltage sensor connected in parallel at the output of the combiner box, a high-frequency DC current sequence and a DC voltage sequence are acquired under the trigger of the clock synchronization signal. The non-stationary wind speed sequence, the solar thermal parameters, the high-frequency DC current sequence, and the DC voltage sequence are subjected to timestamp alignment and outlier removal to generate a multimodal synchronization dataset.

3. The method for controlling and improving the power generation efficiency of a photovoltaic array according to claim 1, characterized in that, In S2, performing empirical mode decomposition on the high-frequency DC current sequence to obtain the eigenmode functions and monotonic residuals includes: Extract all local maxima and local minima of the high-frequency DC current sequence, and use cubic spline interpolation function to fit the local maxima and local minima respectively to construct upper and lower envelopes; Calculate the local mean sequence of the upper envelope and the lower envelope, and subtract the local mean sequence from the high-frequency DC current sequence to obtain the first intermediate sequence; Verify whether the first intermediate sequence satisfies the envelope condition that the number of extreme points is equal to the number of zero-crossing points and the symmetry condition that the local mean of the upper and lower envelopes is zero. If the envelope condition and the symmetry condition are satisfied, the first intermediate sequence is extracted as an intrinsic mode function, and the intrinsic mode function is subtracted from the high-frequency DC current sequence to obtain the remaining sequence. The remaining sequence is used as a new original sequence to repeat the decomposition operation until the remaining sequence becomes a monotonic residual without extreme points.

4. The method for controlling and improving the power generation efficiency of a photovoltaic array according to claim 3, characterized in that, In S2, performing a Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum includes: Perform a quadrature phase shift operation on each of the intrinsic mode functions to construct the analytic signal corresponding to each of the intrinsic mode functions; Extract the instantaneous amplitude and instantaneous phase of the analytical signal, and obtain the instantaneous frequency by performing a derivative operation on the instantaneous phase with respect to the time variable; Using the time series as the horizontal axis, the instantaneous frequency as the vertical axis, and the square of the instantaneous amplitude as the vertical axis, the instantaneous frequencies and instantaneous amplitudes of all the intrinsic mode functions are mapped to a three-dimensional coordinate system to construct a three-dimensional time-frequency energy spectrum.

5. The method for regulating and controlling the power generation efficiency of a photovoltaic array according to claim 1, characterized in that, In S3, calculating the wind load disturbance frequency band by obtaining the non-stationary wind speed sequence includes: Calculate the long-term trend component of the non-stationary wind speed sequence, and subtract the long-term trend component from the non-stationary wind speed sequence to obtain the time-domain fluctuating wind speed component. The autocorrelation function is calculated on the time-domain fluctuating wind speed component, and the fast Fourier transform is performed on the obtained autocorrelation function to generate the turbulent power spectral density distribution matrix. Extract the center frequency of the main energy peak in the turbulent power spectral density distribution matrix, and define the frequency range based on the center frequency of the main energy peak and a preset frequency band broadening coefficient. Set the frequency range as the wind load disturbance frequency band.

6. The method for regulating and controlling the power generation efficiency of a photovoltaic array according to claim 5, characterized in that, In S3, flutter eigenmodes isomorphic to the wind-borne disturbance frequency band are removed from the three-dimensional time-frequency energy spectrum. The remaining eigenmode functions are then superimposed with the monotonic residuals to reconstruct a pure electrical sequence, including: Calculate the energy centroid frequency of each of the intrinsic mode functions in the three-dimensional time-frequency energy spectrum; The energy centroid frequency is matched and verified with the upper and lower limits of the wind load disturbance frequency band; If the energy centroid frequency falls within the wind load disturbance frequency band, the corresponding intrinsic mode function is determined to be the flutter intrinsic mode caused by the aerodynamic flutter of the support frame, and the instantaneous amplitude sequence of the flutter intrinsic mode is forcibly set to zero and removed. The intrinsic mode function set after amplitude zeroing is linearly summed and reconstructed with the monotonic residual to generate a clean electrical sequence that eliminates aerodynamic flutter interference.

7. The method for regulating and controlling the power generation efficiency of a photovoltaic array according to claim 1, characterized in that, In S4, substituting the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing a constant-phase element to establish the anomalous dynamic mapping relationship includes: Construct a fractional-order equivalent circuit topology for a photovoltaic module that includes a photocurrent source, diodes, and parallel constant-phase elements; Based on the fractional equivalent circuit topology, Kirchhoff's current law is used to construct nodal equations, and a fractional differential operator is introduced to perform derivative operations on the voltage variables to characterize the charging and discharging characteristics of the constant phase element. The pure electrical sequence is input into the node equation as the total output current variable, and the DC voltage sequence is input into the node equation as the terminal voltage variable, thus establishing an anomalous dynamic mapping relationship containing unknown fractional orders.

8. The method for regulating and controlling the power generation efficiency of a photovoltaic array according to claim 7, characterized in that, In S5, the fractional order is obtained by solving the anomalous dynamic mapping relationship. When the fractional order is less than a preset benchmark value, the output hidden crack distortion parameters include: The nonlinear damped least squares iterative algorithm is used to identify the parameters of the anomalous dynamic mapping relationship and extract the converged fractional order. Obtain a preset reference value to characterize the standard Gaussian diffusion of charge carriers in a lossless silicon wafer; Compare the extracted fractional order with the preset benchmark value. If the fractional order drops and becomes less than the preset benchmark value, it is determined that there is a carrier sub-diffusion hindrance phenomenon inside the silicon wafer. Based on the difference between the preset benchmark value and the fractional order, a hidden crack distortion parameter is generated to quantify the severity of the hidden crack, and the hidden crack distortion parameter is output.

9. The method for controlling and improving the power generation efficiency of a photovoltaic array according to claim 1, characterized in that, In S6, based on the hidden crack distortion parameters and the photothermal parameters, the maximum Lyapunov exponent of the thermodynamic system is calculated. The safety clamping voltage is solved using the constraint that the maximum Lyapunov exponent is in the negative range. The control command issued to the inverter to operate at the safety clamping voltage includes: Using the microcrack distortion parameters and the photothermal parameters as state variables, the nonlinear phase space of the thermodynamic system of the target photovoltaic array is reconstructed. A virtual thermal perturbation is applied in the nonlinear phase space, and the maximum logarithmic rate of change of the divergence distance of the adjacent state trajectories over time is tracked and calculated to obtain the maximum Lyapunov exponent. The constraint condition for the thermodynamic system to maintain a convergent and stable state is that the maximum Lyapunov exponent is always in the negative range; When the maximum Lyapunov exponent is detected to be increasing and approaching the zero bifurcation boundary, the maximum Lyapunov exponent is kept negative as the objective function, and the reverse search algorithm is used to optimize and calculate the safety clamping voltage that limits thermal runaway. A voltage limit command containing the safety clamp voltage is generated, and the voltage limit command is sent to the inverter as a control command to perform maximum power point tracking deviation action.

10. A control system for improving the power generation efficiency of a photovoltaic array, characterized in that, A method for improving the power generation efficiency of a photovoltaic array according to any one of claims 1-9, comprising the following modules: A multimodal non-stationary parameter acquisition module is used to simultaneously acquire non-stationary wind speed sequences, photothermal parameters, high-frequency DC current sequences, and DC voltage sequences of the target photovoltaic array; The time-frequency energy spectrum construction module is used to perform empirical mode decomposition on the high-frequency DC current sequence to obtain the intrinsic mode functions and monotonic residuals, and to perform Hilbert transform on the intrinsic mode functions to construct a three-dimensional time-frequency energy spectrum. The pure electrical sequence reconstruction module is used to calculate the wind load disturbance frequency band of the non-stationary wind speed sequence, remove flutter eigenmodes that are isomorphic to the wind load disturbance frequency band from the three-dimensional time-frequency energy spectrum, and superimpose the remaining eigenmode functions with the monotonic residual to reconstruct the pure electrical sequence. An anomalous dynamic mapping establishment module is used to substitute the pure electrical sequence and the DC voltage sequence into the fractional impedance differential equation containing constant phase elements to establish an anomalous dynamic mapping relationship. The hidden crack distortion parameter identification module is used to solve the anomalous dynamic mapping relationship to obtain the fractional order. When the fractional order is less than a preset benchmark value, the hidden crack distortion parameter is output. The anti-chaotic bifurcation collaborative control module is used to calculate the maximum Lyapunov exponent of the thermodynamic system based on the hidden crack distortion parameters and the photothermal parameters, solve the safety clamping voltage with the maximum Lyapunov exponent being in the negative range as a constraint, and issue a control command to the inverter to operate at the safety clamping voltage.

Images