A voltage adaptive solar charging control method and system

By collecting the real-time terminal voltage and voltage drop slope of the photovoltaic array, transient disturbances are identified, maximum power point tracking is paused, and the output voltage is adjusted using the photovoltaic array mechanism model and power law function. This solves the problem of response lag in traditional algorithms and achieves efficient solar charging control.

CN123074075APending Publication Date: 2026-11-10FOSHAN SHENGYANG SOLAR TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611154802.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-31
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Traditional solar charging control algorithms are slow to respond to sudden environmental changes and are prone to misjudgments, leading to reverse tracking or oscillation of the operating point and resulting in energy loss. Most systems rely on high-precision current sensors for power calculation, which not only increases hardware costs and system complexity but also introduces measurement noise interference, affecting control stability. If maximum power point tracking continues to be performed during transient disturbances, it is very easy to cause control divergence, resulting in a sharp drop in system efficiency or even abnormal operation.

Method used

The system collects the real-time terminal voltage of the photovoltaic array, calculates the voltage drop slope and trend, identifies transient disturbance events, pauses the maximum power point tracking operation, calls the photovoltaic array mechanism model for mathematical inversion calculation, performs multi-point weighted trend verification based on historical data, uses the power law function to calculate the adaptive dynamic step size adjustment of the output voltage, and corrects the series and parallel resistance parameters in the photovoltaic array physical model.

Benefits of technology

By verifying historical data trends and adjusting the adaptive step size, control oscillations are suppressed, ensuring the accuracy of the adjustment direction, improving the dynamic response speed, reducing power loss and oscillation amplitude in steady state, and enhancing the control stability and efficiency of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123074075A_ABST
    Figure CN123074075A_ABST
Patent Text Reader

Abstract

The application discloses a voltage self-adaptive solar charging control method and system, relates to the photovoltaic power generation technical field, and comprises the following steps: collecting real-time terminal voltage of a photovoltaic array, and calculating voltage drop slope and change trend of the real-time terminal voltage; determining whether a transient disturbance event is triggered according to the voltage drop slope and the change trend; when the transient disturbance event is triggered, suspending maximum power tracking operation; calling a preset photovoltaic array mechanism model to perform mathematical inversion calculation on the real-time terminal voltage to obtain virtual current, and combining the real-time terminal voltage to calculate real-time power of a current period; extracting historical voltage data and historical power data of a previous period, combining the real-time power, performing multi-point weighted trend checking to determine a voltage regulation direction, and updating the regulation direction in a non-transient disturbance event period; and the dynamic response speed is greatly improved, and meanwhile, power loss and oscillation amplitude in a steady state are greatly reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and in particular to a voltage adaptive solar charging control method and system. Background Technology

[0002] With the accelerated global green energy transition, solar photovoltaic power generation technology has gradually evolved from an auxiliary energy source to a core component of the energy structure. As the core link in the energy conversion and management of photovoltaic systems, solar charging control technology has undergone iterative evolution from early ordinary switching control and pulse width modulation technology to maximum power point tracking technology. Currently, the third-generation control technology represented by MPPT has become the industry mainstream. By real-time detection of the voltage and current of the photovoltaic array, it calculates and tracks the maximum power point, theoretically achieving a tracking efficiency of 99% and a system power generation efficiency of over 97%, thus improving energy utilization efficiency.

[0003] Traditional algorithms are slow to respond to sudden environmental changes and are prone to misjudgment, leading to reverse tracking or oscillation of the operating point and resulting in energy loss. Most systems rely on high-precision current sensors for power calculation, which not only increases hardware costs and system complexity but also introduces measurement noise interference, affecting control stability. If maximum power point tracking continues to be performed during transient disturbances, it is very easy to cause control divergence, resulting in a sharp drop in system efficiency or even abnormal operation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a voltage-adaptive solar charging control method to solve the problems of traditional algorithms' lagging response to sudden environmental changes, which easily leads to misjudgments, causing reverse tracking or oscillation of the operating point and resulting in energy loss. Most systems rely on high-precision current sensors for power calculation, which not only increases hardware costs and system complexity but also introduces measurement noise interference, affecting control stability. During transient disturbances, if maximum power point tracking calculations continue to be performed, it is very easy to cause control divergence, leading to a sharp drop in system efficiency or even abnormal operation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a voltage adaptive solar charging control method, which includes acquiring the real-time terminal voltage of a photovoltaic array and calculating the voltage drop slope and change trend of the real-time terminal voltage. The voltage drop slope and trend are used to determine whether a transient disturbance event has been triggered. When a transient disturbance event is triggered, the maximum power point tracking operation is paused. The virtual current is obtained by mathematically inverting the real-time terminal voltage using a pre-set photovoltaic array mechanism model, and the real-time power of the current cycle is calculated in combination with the real-time terminal voltage. Extract historical voltage and power data from previous cycles, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage regulation direction, and update the regulation direction within non-transient disturbance event cycles; Based on the voltage error between the real-time terminal voltage and the target reference voltage, an adaptive dynamic step size is calculated using a preset power law function, and the output voltage of the photovoltaic array is adjusted based on the voltage adjustment direction and the adaptive dynamic step size.

[0007] As a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the step of performing transient disturbance event judgment includes: The real-time voltage drop amplitude is compared with the preset voltage drop threshold, and the output power change trend is compared with the unperturbed power trend predicted based on the current photovoltaic array physical model. When the drop exceeds the threshold and the power trend deviation exceeds the preset range, a transient disturbance event is determined to have occurred, and a disturbance activation flag is generated.

[0008] In a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the disturbance activation flag includes: In response to the disturbance activation flag, the update of the maximum power tracking duty cycle is stopped, the current duty cycle is maintained, the maximum power tracking state is switched to the tracking freeze state, and the data buffer for transient voltage and transient power is activated.

[0009] As a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the data buffer includes: Using the trigger time of the disturbance activation flag as a synchronization reference, steady-state voltage and power data for a first preset duration before the disturbance occurs, and transient voltage and power data for a second preset duration including voltage drop and recovery edge after the disturbance occurs are extracted from the buffer and spliced ​​together to form a transient identification dataset.

[0010] As a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the transient identification dataset includes: Using a single diode model, the partial derivatives of each characteristic parameter of the photovoltaic array with respect to the output power are calculated analytically, and the variation of the partial derivatives within the time period covered by the transient identification dataset is compared. The parameter with the largest change range is selected as the dominant parameter to be identified, and the dominant parameter to be identified includes at least series resistance and parallel resistance.

[0011] As a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the dominant parameters to be identified include: When a transient disturbance event is detected, the timing of the disturbance activation flag is used as the synchronization reference. The terminal voltage sequence and the corresponding power sequence of the photovoltaic array are collected only when the maximum power point tracking control is in a frozen state. Based on the disturbance activation flag, a transient identification dataset is constructed from the collected data. The transient identification dataset includes: steady-state voltage and power data within a first preset time before the disturbance occurs, and transient voltage and power data within a second preset time after the disturbance occurs, including voltage drop edge and recovery edge. The terminal voltage sequence in the transient identification dataset is used as the input of the recursive least squares algorithm, and the corresponding power sequence is used as the output reference. A recursive fitting model is established using the current-voltage implicit equation of a single diode model, wherein the parameters to be estimated in the recursive fitting model are limited to the dominant parameters to be identified, including series resistance and parallel resistance. During the recursive process, sampled data from the transient identification dataset are introduced point by point. The residual between the model output power and the output reference is calculated based on the current parameter estimate. The gain vector is updated based on the residual to correct the parameter estimate, thereby minimizing the cumulative sum of squared residuals within the range of the transient identification dataset. The transient identification dataset is formed by splicing the steady-state data before the disturbance and the transient data after the disturbance, so that the input data simultaneously includes the parameter-sensitive change range and the steady-state constraint range, thereby improving the convergence speed and identification accuracy of the dominant parameter to be identified. After traversing the entire transient identification dataset, the converged series resistance and parallel resistance estimates are output, and these estimates are written into the photovoltaic array physical model to replace the original parameters and obtain the corrected photovoltaic array physical model.

[0012] As a preferred embodiment of the voltage adaptive solar charging control method of the present invention, the modified photovoltaic array physical model includes: During the detection of transient disturbance events, the maximum power point tracking control algorithm is kept in a frozen state to keep the output duty cycle of the DC-DC converter constant, thereby acquiring the terminal voltage sequence and power sequence of the photovoltaic array under fixed operating point conditions; Based on the data collected during the freeze period, a steady-state pause dataset for verifying the photovoltaic array physical model is formed, and the updated series resistance and parallel resistance parameters are read from the corrected photovoltaic array physical model. The real-time terminal voltage, the updated series resistance and parallel resistance parameters, and the preset photocurrent parameters, reverse saturation current parameters and ideal factor parameters in the photovoltaic array physical model are all substituted into the current-voltage implicit equation of the single diode model. To address the nonlinear coupling characteristic where current simultaneously appears in the exponential and resistance voltage drop terms in the implicit equation, the Newton-Raphson iterative method is employed for numerical solution. In each iteration, the residual and derivative information of the equation are calculated, and the current estimate is updated. During the iteration process, the current change between two adjacent iterations is lower than the preset convergence threshold as the stopping condition, and the converged current solution is used as the virtual current value. The calculation of the virtual current is based on the photovoltaic array physical model that has been corrected by transient perturbation data, so that the virtual current can reflect the actual parameter changes after the perturbation, thereby improving the accuracy of power calculation under dynamic illumination conditions. The virtual current is multiplied by the real-time terminal voltage to obtain the current cycle power, and the power is input into the maximum power point tracking control algorithm for subsequent duty cycle adjustment.

[0013] In a second aspect, the present invention provides a voltage adaptive solar charging control system, including a voltage acquisition module, a disturbance freezing module, an inversion calculation module, a trend verification module, and an adaptive adjustment module. The voltage acquisition module is used to acquire the real-time terminal voltage of the photovoltaic array and calculate the voltage drop slope and trend of the real-time terminal voltage. The disturbance freezing module is used to determine whether a transient disturbance event is triggered based on the voltage drop slope and change trend. When a transient disturbance event is triggered, the maximum power point tracking operation is paused. The inversion calculation module is used to call a preset photovoltaic array mechanism model to perform mathematical inversion calculation on the real-time terminal voltage to obtain the virtual current, and combine the real-time terminal voltage to calculate the real-time power of the current cycle. The trend verification module is used to extract historical voltage data and historical power data from the previous period, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage adjustment direction, and update the adjustment direction within the non-transient disturbance event period. The adaptive adjustment module is used to calculate the adaptive dynamic step size based on the voltage error between the real-time terminal voltage and the target reference voltage, combined with a preset power law function, and adjust the output voltage of the photovoltaic array based on the voltage adjustment direction and the adaptive dynamic step size.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the voltage adaptive solar charging control method as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the voltage adaptive solar charging control method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By extracting historical data from previous cycles and combining it with the current real-time power for multi-point weighted trend verification, the voltage adjustment direction is determined. The historical trend in the time series smooths out instantaneous noise interference, fundamentally suppressing control oscillations caused by measurement noise, and ensuring the absolute accuracy of the adjustment direction. Based on the voltage error and using the nonlinear characteristics of the preset power law function, the adaptive dynamic step size is calculated to adjust the output voltage. This allows the system to automatically adopt a large step size for fast response when far from the target point, and automatically reduce the step size for fine adjustment when approaching the target point. This completely solves the inherent contradiction in the traditional fixed step size algorithm that "fast speed leads to large oscillations, and high precision leads to slow response." While significantly improving the dynamic response speed, it greatly reduces the power loss and oscillation amplitude in steady state. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of a voltage-adaptive solar charging control method; Figure 2 This is a schematic diagram of a voltage-adaptive solar charging control system. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Reference Figures 1-2 As one embodiment of the present invention, this embodiment provides a voltage adaptive solar charging control method, comprising the following steps: S1. Collect the real-time terminal voltage of the photovoltaic array and calculate the voltage drop slope and trend of the real-time terminal voltage.

[0023] Furthermore, the real-time voltage drop is compared with a preset voltage drop threshold, and the output power trend is compared with the unperturbed power trend predicted based on the current photovoltaic array physical model.

[0024] When the drop exceeds the threshold and the power trend deviation exceeds the preset range, a transient disturbance event is determined to have occurred, and a disturbance activation flag is generated.

[0025] It should be noted that in transient disturbance data, voltage and power changes have different statistical characteristics: voltage shows abrupt deviations from the local mean, while power shows nonlinear deviations from the model prediction. Not only are their dimensions different, but they also exhibit asynchronous characteristics in the early stages of the disturbance. Under these data characteristics, if a single physical quantity is used for judgment, it is impossible to cover both types of responses simultaneously. If linear fusion is performed directly, it is difficult to distinguish between slowly varying fluctuations and abrupt disturbances. It is necessary to construct a joint criterion that can simultaneously characterize "local abruptness" and "model deviation," so that the disturbance judgment is based on the data distribution characteristics, thereby ensuring the stability and consistency of the judgment mechanism under different operating conditions. Under these data characteristics, if voltage deviation and power model deviation are not constrained simultaneously, and judgment is made based on only a single dimension, it is impossible to form a stable distinction between slowly varying illumination and local abruptness, thus causing the disturbance triggering condition to fail under different operating conditions.

[0026] The premise for the construction to be valid is that the photovoltaic array exhibits different data response characteristics under local shading disturbances and normal slow-changing illumination conditions: during slow-changing illumination, the terminal voltage and output power usually show slow and continuous changes, and the trend of change is consistent over time; while during sudden changes in local shading, due to the local mismatch effect of the components, the terminal voltage will drop rapidly, and the output power will deviate significantly from the predicted value of the physical model based on the assumption of uniform illumination. Moreover, the two exhibit synchronous sudden change characteristics in a short period of time. Based on the data characteristics of "rapid voltage drop + synchronous occurrence of power model deviation", it is feasible to jointly determine the two types of physical quantities.

[0027] Because different photovoltaic arrays have different terminal voltage levels, directly using the absolute value of voltage drop as the criterion would result in different disturbance levels represented by the same drop in high-voltage and low-voltage arrays, making it impossible to form a unified judgment standard. Furthermore, comparing only the voltage drop slope is easily affected by sampling noise and short-term control disturbances. Therefore, instead of directly using the absolute voltage drop, a preset voltage drop threshold is used as a normalization benchmark to convert the voltage drop into a dimensionless relative deviation. A fusion method is adopted, where the squares of two factors are taken separately and then weighted and summed. By normalizing the voltage deviation amplitude and power model deviation separately to obtain dimensionless factors, and then using squaring as a nonlinear amplification method, a disturbance intensity factor is constructed. Considering the relative relationship between the voltage drop amplitude and the threshold, the absolute drop is converted into a comparable value through normalization. The normalized voltage deviation factor is defined as follows: ; in, This represents the drop in real-time terminal voltage relative to the historical average within the sliding window. To preset a voltage drop threshold, the degree of voltage deviation is referenced to the threshold, thus eliminating the dimensional differences between different system voltage levels.

[0028] Because output power varies significantly with light intensity and operating point, directly using the absolute difference between measured power and model power would naturally lead to larger deviations in the high-power range, while effective perturbations in the low-power range might be weakened, resulting in inconsistent judgment scales across different operating ranges. Therefore, instead of directly using the absolute power deviation, we use the current model-predicted power as the normalization benchmark for the corresponding operating point, converting the power deviation into a relative deviation. We consider the degree of deviation between the measured power and the power of the unperturbed model. For the power dimension, the measured power... Compared to the model-predicted power calculated using the current photovoltaic array physical model under the assumption of no disturbances and by substituting the real-time terminal voltage, Compare the two and take the absolute value of their difference. As the deviation to be normalized, As the reference power for the corresponding operating point, the influence of the absolute power order is eliminated by phase division, and a power deviation factor is defined. The expression is: ; The absolute power deviation is converted into a proportional value relative to the current operating point, thus avoiding the problem of naturally large absolute deviations in the high power range.

[0029] After obtaining the two dimensionless factors, they need to be merged into a single disturbance intensity index. The core requirement of the fusion is that an over-threshold alarm should only be output when voltage dips and power model deviations occur simultaneously, while suppressing isolated fluctuations of individual factors.

[0030] Relying solely on the voltage deviation factor can miss early shadowing scenarios where power is clearly abnormal but voltage has not yet fully dropped. Relying solely on the power deviation factor may also misjudge load disturbances or measurement noise as shadowing disturbances. If linear addition is used, moderate two-factor fluctuations may be amplified. If a product is used, either factor approaching zero will suppress the abnormal contribution of the other. If the maximum value is used, it degenerates into a single-factor judgment. A squared weighted fusion method is used to compress small fluctuations and amplify deviations, while preserving the joint confirmation relationship of the two factors. To achieve this nonlinear joint confirmation characteristic, [further details are needed]. and Perform squaring operations separately—when the factor values ​​are small, the squaring operation compresses the data to the near-zero region, suppressing background fluctuations under slowly varying conditions. When both factors increase to a certain extent simultaneously, the squaring term increases sharply at a superlinear rate, forming an equivalent two-dimensional nonlinear "sum gate," into which a weighting coefficient is introduced. We perform a weighted summation of the two squared terms to adjust the relative influence of the voltage and power dimensions in the final criterion, constructing the perturbation strength factor, expressed as: ; Where α is the range of values. The weighting coefficients are used to adjust the relative influence of the voltage and power dimensions.

[0031] To eliminate the influence of different dimensions, voltage and power are normalized separately. To avoid misjudgment caused by single-factor fluctuations, a criterion that is triggered only when both factors are present is needed. By comparing various fusion methods such as linear superposition, product form, and maximum value selection, it was found that linear superposition is too sensitive to moderate fluctuations, product form is too strong in suppressing single-factor anomalies, and maximum value form degenerates into single-factor judgment, none of which meet the requirements. Therefore, square-weighted fusion is chosen to suppress small fluctuations and amplify disturbances nonlinearly, thereby forming a judgment function that meets the requirements of disturbance identification.

[0032] when Exceeding the preset disturbance intensity threshold At that time, a perturbation activation flag is generated, and only then is the entire processing chain of parameter identification and model correction uniquely activated, ensuring that the identification algorithm runs only on transient datasets with high information density and avoiding parameter oscillations caused by invalid identification.

[0033] Unlike existing technologies that judge disturbances based on a single voltage change rate or a single power change, this method does not simply set a fixed threshold for triggering. Instead, it introduces both a voltage deviation factor and a power model deviation factor, and constructs a disturbance intensity factor by using a square-weighted fusion method after normalization.

[0034] The problem of asynchronous voltage and power responses in the early stages of local shading was considered under different voltage levels, different power levels, and local shading. The disturbance judgment was changed from a single empirical threshold judgment to a quantitative judgment under dual physical quantity constraints. Only when the abnormal change in terminal voltage and the deviation of the model predicted power both point to the external disturbance in the same period of time will a disturbance effective flag be generated. This reduces the probability of false triggering caused by normal gradual changes in illumination, sampling noise or control perturbations, and provides a reliable triggering entry for subsequent parameter identification.

[0035] S2. Determine whether a transient disturbance event is triggered based on the voltage drop slope and trend. If a transient disturbance event is triggered, pause the maximum power tracking operation.

[0036] Furthermore, in response to the disturbance activation flag, the update of the maximum power tracking duty cycle is stopped, the current duty cycle is maintained, the maximum power tracking state is switched to the tracking freeze state, and the data buffer for transient voltage and transient power is activated.

[0037] It should be noted that during transient disturbances, the statistical characteristics of the acquired data depend on the singularity of the disturbance source. When the maximum power point tracking control is running continuously, voltage and power changes are simultaneously affected by the superposition of external shadow disturbances and internal control adjustments, resulting in data exhibiting multi-source mixed characteristics. Under these data characteristics, if the control variables are not constrained, it is impossible to separate the response caused by changes in physical parameters from the observed data. It is necessary to construct data acquisition conditions dominated only by external disturbances. Considering that the duty cycle is the only variable in the control system that directly affects the operating point, freezing the duty cycle can achieve operating point locking without introducing additional control variables. This ensures that the acquired data only reflects the response to external disturbances, guaranteeing that the data relied upon for subsequent calculations has the characteristics of a single disturbance source, thus providing the feasible conditions required for parameter identification. The duty cycle is the only control variable that directly affects the operating point. Therefore, apart from freezing, there is no other control path that can achieve disturbance isolation without introducing additional variables.

[0038] In a DC-DC converter, the duty cycle is the only control variable that directly determines the operating point of the photovoltaic array. When the load conditions remain unchanged, a fixed duty cycle means that the equivalent load of the array is fixed, thus stabilizing the operating point. By freezing the duty cycle, a system environment with "constant control input" can be constructed at the circuit level, so that the system output is driven only by external input. Based on this control structure characteristic, it can be ensured that the acquired data has a single disturbance source attribute, thereby providing a separable data basis for subsequent parameter identification.

[0039] If MPPT adjustment continues after a disturbance occurs, the duty cycle change will actively alter the photovoltaic array's operating point, causing the acquired voltage and power changes to simultaneously include external shading and internal control effects, making it impossible for subsequent parameter identification to distinguish the source of the change. Introducing additional compensation bias or damping adjustment would add new control variables and disrupt the single-source disturbance characteristic of the data. Freezing the duty cycle prevents the control side from injecting new operating point disturbances into the system. This completely cuts off the path of maximum power point tracking adjustment's influence on the array's operating point from the control source as soon as a disturbance is detected. The duty cycle is the only variable connecting the control algorithm and the circuit's operating state; by forcibly setting the duty cycle value for the current control cycle... Compared to the value of the previous period By maintaining equality, the operating point can be instantly locked at the value at the moment of disturbance triggering without introducing any additional bias or damping function. This can be achieved using the duty cycle freeze command, expressed as: ; This means that the duty cycle of the DC-DC converter is forced to remain at the value at the trigger moment and is no longer updated. By using the duty cycle freezing mechanism, a quasi-steady-state operating point is constructed in a short time, so that the collected data only reflects changes in the external environment, thereby providing a data basis for subsequent parameter identification from a single disturbance source.

[0040] The freezing method is not simply stopping control, but rather based on the structural characteristic of "uniqueness of control variables". It achieves maximum data purity through minimal intervention of variables. Compared with the introduction of compensation control or filtering, it does not add new degrees of control freedom and avoids introducing new uncertainties. It reflects the technical path of replacing complex control compensation with system structural constraints.

[0041] timestamp of the disturbance taking effect As a synchronization reference, a data buffer for transient voltage and transient power is initiated, and forward backtracking is used to capture the transient data. Short-term steady-state data of the interval, continuously recorded backward. The transient fall edge and recovery edge data of the interval ensure that the transient identification dataset formed subsequently has extremely high purity and signal-to-noise ratio from the data source, so that the change of each sampling point faithfully reflects the dynamics of physical parameters caused by shadow.

[0042] Unlike existing technologies that continue to perform maximum power point tracking during disturbances, instead of changing the duty cycle to track the power peak, the duty cycle is frozen immediately after the disturbance activation flag is generated, so that the photovoltaic array maintains a fixed equivalent load state for a short period of time.

[0043] The superimposed effects of external shading changes and internal MPPT regulation on voltage and power sequences were considered. By fixing control variables, internal regulation factors were removed from the sampled data, so that the data collected during the disturbance period mainly reflected the physical response caused by external shading. The transient identification dataset constructed subsequently no longer contained the operating point shift caused by the maximum power point tracking algorithm itself, which improved the interpretability, purity and reliability of parameter identification of the data.

[0044] S3. Call the preset photovoltaic array mechanism model to perform mathematical inversion calculation on the real-time terminal voltage to obtain the virtual current, and combine it with the real-time terminal voltage to calculate the real-time power of the current cycle.

[0045] Furthermore, using the trigger time of the disturbance activation flag as the synchronization benchmark, the steady-state voltage and power data for a first preset duration before the disturbance occurs, and the transient voltage and power data for a second preset duration including voltage drop and recovery edge after the disturbance occur are extracted from the buffer and spliced ​​together to form a transient identification dataset.

[0046] Using a single diode model, the partial derivatives of each characteristic parameter of the photovoltaic array with respect to the output power are calculated analytically, and the variation of the partial derivatives within the time period covered by the transient identification dataset is compared.

[0047] The parameter with the largest change range is selected as the dominant parameter to be identified. The dominant parameter to be identified includes at least the series resistance and the parallel resistance.

[0048] It should be noted that in transient disturbance data, the impact of different physical parameters on output power varies. Statistically, some parameters cause small power response amplitudes throughout the entire disturbance process, while others generate responses at specific stages. Given these data characteristics, directly identifying all parameters would result in the limited data being scattered across a high-dimensional parameter space, leading to an ill-posed identification problem. It is necessary to select dominant parameters based on the data's response characteristics to match the identification problem's dimension with the data's information content. If full parameter identification is performed without parameter selection, the parameter space will be underdetermined under the condition of limited transient data length, making it impossible to obtain a stable solution. By quantifying the dynamic range of parameter sensitivity during the transient process, the parameters that contribute most to the current dataset can be identified, ensuring that the selected parameters are effectively excited by the data, thus enabling the subsequent identification process to possess convergence and stability.

[0049] In the single diode model, the influence of each physical parameter on the output power is separable in different operating ranges. That is, within a certain voltage variation range, the sensitivity of different parameters to power varies. Especially during the shadowing disturbance, due to the change in the current path and leakage channel, the sensitivity functions corresponding to the series resistance and parallel resistance show obvious fluctuations over time, while the changes of other parameters are relatively smooth. By distinguishing the parameters through the magnitude of sensitivity changes, a feasible basis for screening the dominant parameters can be found.

[0050] While empirical models are computationally simple, their parameters lack clear physical meaning, making it difficult to support subsequent targeted corrections to series and parallel resistances. Complex multi-diode models, though more accurate, have a large number of parameters and high computational cost, making them unsuitable for embedded real-time identification. Therefore, a single-diode model, striking a balance between physical interpretability and computational complexity, is adopted as the unified basis for sensitivity analysis and parameter identification. This model defines the mechanistic model describing the physical characteristics of the photovoltaic array. In engineering applications, the single-diode model achieves a good balance between computational cost and accuracy, and its parameters have clear physical meaning, making it suitable for embedded real-time identification. The current-voltage equation is used as the analytical basis for sensitivity analysis, expressed as: ; in, For output current, For photocurrent, It is the reverse saturation current. As an ideal factor, Thermoelectric voltage, For series resistance, For parallel resistors, Terminal voltage, output power .

[0051] To quantify the impact of changes in each physical parameter on the output power, for each physical parameter in the model... Derive the partial derivatives of power with respect to parameters The partial derivative characterizes the response strength of the output power when the parameters change by a unit, i.e., the parameter sensitivity, within the time range covered by the transient identification dataset. Internally, the terminal voltage of each sampling point is... Substitute into the sensitivity function ,in The nominal values ​​of the current model parameters are used to obtain the sequence of changes in the sensitivity of each parameter over time.

[0052] Since the impact of each parameter on power differs during disturbances, directly identifying all parameters would lead to the finite transient data being dispersed into a high-dimensional parameter space, increasing parameter coupling and underdetermining risks. Relying solely on experience to select parameters is insufficient to adapt to changes in parameter sensitivity under different disturbance events. Therefore, we first calculate the time-varying sequence of each parameter's sensitivity and select the peak-to-valley difference, which reflects the complete dynamic response range, as the screening criterion, rather than using the mean, which is easily overwhelmed by the steady-state phase, or the variance, which is difficult to distinguish peak shapes. To screen out parameters that truly undergo significant changes during this disturbance from the changing sequence, a quantitative index that fully reflects the dynamic range of sensitivity is needed. If the mean is used, the consistently low sensitivity values ​​in the steady-state phase will overwhelm the high sensitivity contribution in the transient phase. If the variance is used, although it reflects the degree of dispersion, it cannot distinguish between two situations with different identification values: a single extreme peak and a sustained high plateau. Therefore, we directly take the difference between the maximum and minimum values ​​of the sensitivity sequence over the entire time period, defining this difference as the sensitivity fluctuation amplitude. ; The use of peak-valley difference as a screening criterion is based on the consideration that transient data exhibits the characteristic of "short-term strong excitation": during the disturbance process, the sensitivity of key parameters often manifests as an instantaneous peak. If the mean or variance index is used, the peak information may be diluted by long-term steady-state data. By directly extracting the difference between the maximum and minimum values, transient excitation information can be retained, thereby more accurately reflecting the contribution of the parameters to the current dataset.

[0053] The fluctuation amplitude directly quantifies the complete dynamic range of the power response that the parameter can cause from the lowest to the highest during the entire transient process: the larger the amplitude, the stronger the power response that the parameter can produce at a certain stage of the transient process. The change of the true value is characterized in the collected data, so the greater the contribution of the parameter to the information of the current dataset, the higher the data-driven accuracy that can be obtained when identifying the parameter.

[0054] Shadowing primarily alters carrier transport paths and interface leakage paths, and series resistance and parallel resistors It then changes drastically, and the photocurrent... Reverse saturation current and ideal factor Primarily determined by the semiconductor material properties and junction temperature, it remains essentially constant during a single blocking event. and The corresponding sensitivity fluctuation will be much greater than that of other parameters.

[0055] By setting a fluctuation amplitude threshold It automatically selects parameters exceeding a threshold as the dominant parameters to be identified, thereby shrinking the five-dimensional parameter space to two dimensions, so that recursive identification only requires estimation. and This fundamentally ensures the well-positionedness and convergence speed of the identification.

[0056] Unlike existing technologies that directly identify all model parameters uniformly, this approach does not blindly increase the number of parameters to be identified. Instead, it first performs analytical sensitivity calculations on each parameter based on a single diode model, and then selects the dominant parameters to be identified based on the sensitivity fluctuation amplitude.

[0057] Considering the different physical changes of various parameters during shadowing perturbations and the difficulty of supporting simultaneous estimation of high-dimensional parameters with limited transient data, this paper quantifies the contribution of each parameter to the current dataset by sensitivity amplitude, and limits the identification objects to series and parallel resistances in response to perturbations. Without adding additional sensors and complex hardware, the identification dimension is reduced, parameter coupling and underdetermined risk are reduced, and the transient data of limited length can be used to correct key parameters.

[0058] S4. Extract historical voltage and power data from the preceding cycle, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage regulation direction, and update the regulation direction within the non-transient disturbance event cycle.

[0059] Furthermore, when a transient disturbance event is detected, the timing of the disturbance activation flag is used as the synchronization reference, and the terminal voltage sequence and corresponding power sequence of the photovoltaic array are collected only during the period when the maximum power point tracking control is in a frozen state.

[0060] Based on the disturbance activation flag, a transient identification dataset is constructed from the collected data. The transient identification dataset includes: steady-state voltage and power data within a first preset time period before the disturbance occurs, and transient voltage and power data within a second preset time period after the disturbance occurs, including voltage drop edges and recovery edges.

[0061] The terminal voltage sequence in the transient identification dataset is used as the input to the recursive least squares algorithm, and the corresponding power sequence is used as the output reference.

[0062] A recursive fitting model is established based on the current-voltage implicit equation of a single diode model. The parameters to be estimated in the recursive fitting model are limited to the dominant parameters to be identified, including series resistance and parallel resistance.

[0063] During the recursive process, sampled data from the transient identification dataset are introduced point by point. The residual between the model output power and the output reference is calculated based on the current parameter estimate. The gain vector is updated based on the residual to correct the parameter estimate, thereby minimizing the cumulative sum of squared residuals within the transient identification dataset.

[0064] The transient identification dataset is formed by splicing the steady-state data before the disturbance and the transient data after the disturbance, so that the input data simultaneously includes the parameter-sensitive change range and the steady-state constraint range, thereby improving the convergence speed and identification accuracy of the dominant parameter to be identified.

[0065] After traversing the entire transient identification dataset, the converged series and parallel resistance estimates are output and written into the photovoltaic array physical model to replace the original parameters and obtain the corrected photovoltaic array physical model.

[0066] It should be noted that the parameter identification process is not an improvement on the algorithm itself, but is determined by the structural characteristics of the transient data. In the constructed dataset, the first-stage steady-state data provides constraint information, while the second-stage transient data provides strong excitation information. The data as a whole exhibits a non-uniform distribution characteristic from low excitation to high excitation.

[0067] Given the characteristics of the data, if a batch computation method is used, the temporal structure of the data cannot be fully utilized; if continuous online updates are used, low-information-density data interference will be introduced. It is necessary to construct a computation method that can gradually introduce information according to the data sequence and dynamically adjust the parameter estimation, so that the algorithm process matches the data structure, thereby ensuring that the parameter convergence process is stable and efficient. If a non-recursive form is used, it is impossible to dynamically adjust the parameter estimation during the gradual introduction of data, thus making it difficult to utilize the high excitation characteristics of the transient segment to achieve fast convergence.

[0068] Since the system lacks a physical current sensor, current error cannot be directly used as the identification target. Forcibly introducing a current estimate as an observation reference would make the identification target dependent on another unverified intermediate variable, increasing error propagation. Therefore, it is preferable to directly construct the identification target using the system's measurable terminal voltage and power, allowing parameter corrections to directly serve the power calculations required for MPPT control. This is achieved by considering the measured power from all N sampling points in the transient identification dataset. Power calculated by the model Using the sum of squares of the differences as the optimization criterion, we construct the objective function within the least squares framework: ; The objective function is based on measured power. For reference, using model power This method avoids the problem of not being able to directly compare currents due to the lack of a current sensor.

[0069] Without a current sensor, the system can directly acquire the terminal voltage and output power. Since power is essentially the product of voltage and current, it already implicitly contains current information. By comparing the model output power with the measured power, the error in current calculation can be indirectly constrained, thereby avoiding direct observation of the current. Based on the observability of the data, using power as the identification target is sufficient in terms of information, thus ensuring that the parameter identification problem is mathematically solvable.

[0070] The application method is modified based on the structural characteristics of transient data: the recursive process is only started after a disturbance is triggered and the data has high excitation characteristics, so that the algorithm runs in the information-sufficient range, thereby avoiding the parameter drift problem that occurs in traditional online identification under low excitation conditions, which reflects the design idea of ​​the algorithm based on data quality triggering.

[0071] Since the current appears in both the exponential and resistance voltage drop terms in the single diode model, the model power cannot be directly written as an explicit function of Rs and Rsh.

[0072] Ignoring implicit coupling and using a linear approximation weakens the true impact of parameter changes on power under shadowing perturbations. At each recursive sampling point, the model current is first calculated based on the current parameter estimate, then the model power is calculated. Recursive least squares updates the parameters online by introducing data point by point. The core of the recursion lies in correcting the parameters at each step using new sampling points. Step 1: Input sampling points. Since the current-voltage equation of the single diode model is implicit, it is not possible to directly write the Pmodel with respect to the sampling points. The explicit expression for the model current is needed, therefore the implicit equations must first be solved to obtain the model current, by estimating the current parameters. and inherent parameters , , , Substituting the values, we use an embedded iterative solution to handle the nonlinear coupling of the current, i.e., solving the equations: ; The solution to the implicit equation is the model current. Thus, the model power is obtained. .

[0073] Since the accuracy of parameter estimation is ultimately reflected in the deviation between model power and measured power, and the direction of the deviation reflects the direction of parameter correction, while the magnitude of the deviation reflects the strength of the correction, instead of constructing additional indirect evaluation quantities, we directly use the power residual as the feedback quantity for recursive updates. Based on the difference between model power and measured power, we drive parameter updates by defining the power residual. Considering that the sign and magnitude of the residual directly reflect the direction and degree of parameter deviation, we use the power residual as the feedback quantity. ; To convert the residuals into parameter corrections, it is necessary to know the direction in which the model output is sensitive to parameter changes. This can be achieved by analytical differentiation to obtain the gradient vector of the model power with respect to the parameters. The gradient is derived from the implicit equation of a single diode based on the implicit function differentiation rule, and reflects the slope of the local influence of each parameter on the output power at the current operating point.

[0074] Since the uncertainty varies across different parameter directions, a fixed step size update will result in slow convergence in directions with high uncertainty and oscillations in directions with high certainty. Furthermore, transient data exhibits time-varying characteristics, meaning older and newer data cannot always maintain equal weight. The covariance matrix is ​​coupled with the gradient direction, and a forgetting factor is introduced to adjust the weights of old and new data. When generating parameter corrections using gradient information, the uncertainty distribution of parameter estimation and the weight allocation of old and new data must be considered comprehensively. A covariance matrix Pk⁻¹ records the uncertainty of parameter estimation in each direction and is coupled with the gradient vector, allowing the correction to have a larger step size in directions with high uncertainty and a smaller step size in directions with high certainty. A forgetting factor is also introduced. To assign higher weights to new data and gradually decay old data, the smaller λ is, the stronger the tracking ability. In the denominator of the gain vector, With the statistical weights of the gradient in the space of current parameter uncertainty To ensure numerical stability and achieve adaptive step size allocation, the gain vector is constructed by adding the values: ; Wherein, the forgetting factor λ satisfies In addition to λ, the denominator also contains the quadratic form hkᵀPk⁻¹hk, reflecting the statistical weight of the current gradient in the parameter uncertainty space. Placing λ in the addend position in the denominator instead of subtracting it ensures numerical stability, and at the same time, the choice of λ achieves a balance between preserving old information and utilizing new information. The smaller the value, the faster the old data decays, and the stronger the algorithm's tracking ability. When the value is 1, it degenerates into standard recursive least squares, with no forgetting effect.

[0075] Since the residuals themselves only represent the difference between the model output and the measured output, and cannot directly explain how each parameter should change, it is necessary to combine them with the gain vector to distribute the residuals to different parameter directions. The power residuals are then weighted and mapped using the gain vector, causing the parameters to be recursively corrected along the direction that reduces the residuals. Based on the gain vector and the current residuals, parameter corrections are generated by linearly distributing the residuals according to the gain vector. The parameter estimates are then updated using a recursive update formula. ; residual Through the gain vector The weighted driving parameters shift in the direction of reducing residuals, and the magnitude of the correction is controlled by both the residual amplitude and the gain.

[0076] Since each parameter correction introduces new constraints on certain parameter directions at the current sampling point, subsequent recursions will incorrectly allocate correction step sizes based on the old uncertainty distribution if the covariance matrix is ​​not updated. Therefore, this invention synchronously subtracts the information already explained by the current gradient direction after parameter updates and incorporates a forgetting factor to maintain tracking of new data. After parameter updates, the covariance matrix needs to be updated synchronously to reflect changes in parameter estimation uncertainty. This is achieved by subtracting the information constrained by the current gradient direction from the old covariance matrix and using a forgetting factor for weighted compensation. The new covariance matrix is ​​obtained using the recursive update formula: ; in, It is the identity matrix. This represents the proportion of the update in the gradient direction that the gain occupies. The information retention factor remaining after deduction is divided by... This involves normalizing and compensating for the forgetting effect, ensuring that the size of the covariance matrix adapts to the decay weights of the data.

[0077] The transient identification dataset, formed by concatenating steady-state data before and after perturbation with transient data after perturbation, creates an information supply structure that gradually steepens. During the recursive process, the first half of the steady-state data exhibits small residuals and gentle gradient information, allowing parameter estimation to gradually establish the correct covariance scale and parameter baseline under steady-state constraints. In the latter half of the transient data, significant voltage changes lead to highly sensitive gradients, increasing the residuals. The gain vector drives the parameters to converge rapidly to their true values ​​with a larger step size. Combined with the old data decay effect of the forgetting factor, the recursive least squares algorithm can output high-precision results after traversing the complete dataset. and The estimated values ​​are written into the photovoltaic array physical model for correction. The entire process is performed only when triggered by transient disturbance events, making full use of the high information density dataset.

[0078] Unlike existing technologies that perform online identification based on long-term historical data or continuous operating data, this method does not continuously perform recursive least squares updates in every control cycle. Instead, it performs recursive identification only on the dataset formed by splicing the steady-state segment before the disturbance and the transient segment after the disturbance during the period when a transient disturbance event is triggered and MPPT is frozen.

[0079] Considering the issues of strong constraints but insufficient excitation in steady-state data and strong excitation but easy divergence in transient data, the two types of data are organized according to the perturbation time sequence. This allows the recursive algorithm to first obtain a stable parameter baseline, and then use the high-sensitivity transient segment to quickly correct the parameters. The recursive least squares algorithm no longer operates on ordinary sampled data with low information density, but instead operates on high information density datasets that have been triggered, filtered, and structured. This improves the convergence speed and stability of series resistance and parallel resistance estimation under limited computing resources.

[0080] S5. Based on the voltage error between the real-time terminal voltage and the target reference voltage, calculate the adaptive dynamic step size using a preset power law function, and adjust the output voltage of the photovoltaic array based on the voltage adjustment direction and the adaptive dynamic step size.

[0081] Furthermore, during the detection of transient disturbance events, the maximum power point tracking control algorithm is kept in a frozen state, so that the output duty cycle of the DC-DC converter remains unchanged, thereby acquiring the terminal voltage sequence and power sequence of the photovoltaic array under fixed operating point conditions.

[0082] Based on the data collected during the freeze period, a steady-state transient dataset is formed for verifying the physical model of the photovoltaic array, and the updated series resistance and parallel resistance parameters are read from the corrected physical model of the photovoltaic array.

[0083] The real-time terminal voltage, the updated series and parallel resistance parameters, and the preset photocurrent parameters, reverse saturation current parameters, and ideality factor parameters in the photovoltaic array physical model are all substituted into the current-voltage implicit equation of the single diode model.

[0084] To address the nonlinear coupling characteristic where current appears simultaneously in both the exponential and voltage drop terms in the implicit equations, the Newton-Raphson iterative method is employed for numerical solution. In each iteration, the residuals and derivatives of the equations are calculated, and the current estimates are updated.

[0085] During the iteration process, the current change between two adjacent iterations is lower than the preset convergence threshold as the stopping condition, and the converged current solution is used as the virtual current value.

[0086] The calculation of virtual current is based on the photovoltaic array physical model that has been corrected by transient perturbation data, which enables the virtual current to reflect the actual parameter changes after the perturbation, thereby improving the accuracy of power calculation under dynamic illumination conditions.

[0087] The current cycle power is obtained by multiplying the virtual current by the real-time terminal voltage, and the power is input into the maximum power point tracking control algorithm for subsequent duty cycle adjustment.

[0088] It should be noted that after the model parameters are corrected by perturbation data, the current-voltage relationship of the photovoltaic array has changed. The calculation process is no longer applicable to the fixed model or offline lookup table method. The model output current must be recalculated based on the current parameters and real-time voltage. Essentially, this is an online solution problem of implicit nonlinear equations. Since the current exhibits self-coupling characteristics in the equations, it cannot be obtained directly through analytical methods. Therefore, it is necessary to construct a stable and fast-converging numerical solution method so that the virtual current can be obtained in real time in each control cycle. This ensures that the model calculation results are consistent with the current physical state, thereby providing reliable input for subsequent control. If a fixed model or offline lookup table method is used, the model and lookup table data will be inconsistent after the parameters are updated, causing the calculation results to deviate from the actual physical state.

[0089] During the operation of a photovoltaic array, the terminal voltage is a directly measurable quantity. In a single diode model, given the parameters, there is a one-to-one correspondence between current and voltage. That is, for a given voltage, there is a unique physically feasible solution to the equation. After the model parameters are identified and corrected, the corresponding current value can be deduced by solving the implicit equation, thus realizing the process of inverting current from voltage.

[0090] Since the corrected Rs and Rsh have altered the current-voltage relationship in the single-diode model, continuing to use a fixed-parameter model or offline lookup table method will fail to reflect the actual array state after shadowing perturbation. Furthermore, the current exists simultaneously in the exponential and resistance-voltage drop terms in the model and cannot be directly obtained through simple algebraic transformations. Therefore, the single-diode equations are rearranged into a zero-point solution form to enable online inversion of the virtual current under the current terminal voltage. The Newton-Raphson iteration method directly solves the original physical equations online, decoupling the convergence process from the specific numerical values ​​of the parameters. and Regardless of the updated value, the consistency and accuracy of the solution process are maintained, and it is fully compatible with the technical architecture of "online model parameter correction," ensuring the corrected parameters are applied. , With the inherent model , , Thermoelectricity and the current measured terminal voltage Substituting the terms, by moving all terms on the right-hand side of the equation to one side to construct a function zero, the equation can be solved using current: ; The Newton-Raphson method is chosen because the implicit equations are continuously differentiable within the working interval and meet the fast convergence condition. Compared with the lookup table method, which relies on fixed parameters and is difficult to adapt to model updates, the Newton-Raphson method maintains the consistency of the solution after parameter changes, thus forming a synergy with "online parameter correction".

[0091] Current to be determined Simultaneously appearing in the index item The exponential position and resistance voltage drop term in In this process, nonlinear coupling of current to itself is formed, and an explicit solution cannot be obtained through algebraic transformation. Therefore, the Newton-Raphson iteration method is used for numerical solution.

[0092] Since the equation is a nonlinear implicit equation, if a lookup table method is used, the table needs to be regenerated after each parameter update. If a simple fixed-point iteration is used, convergence may be slow or unstable when the exponential term changes drastically. Therefore, the Newton-Raphson iteration is used, which uses the residual and derivative information at the current estimation point to determine the next correction amount, ensuring that the solution process simultaneously considers real-time performance and convergence accuracy. Dividing the current residual by the derivative yields the correction step size along the tangent direction to reach the zero point. Geometrically, this is the intersection of the tangent at the current point and the horizontal axis. The Newton-Raphson iteration approximates the true solution by linearly approximating f(I) at the current estimate. The iterative format is constructed using tangent extrapolation. ; in To make the first Sub-iteration current The residual obtained by substituting into the equation, This is the value of the derivative function.

[0093] Since the correction step size of Newton's iteration depends on the local slope of the residual relative to the change in current, if the derivative is obtained solely through numerical difference, it will increase the computational load and introduce errors in the selection of the difference step size. Directly performing analytical differentiation on the implicit equation ensures that a definite local slope is obtained in each iteration. To obtain the derivative value in the correction step size, analytical calculation is required. right The derivative function, by taking the derivative term by term, considers that both the exponential and resistance terms contain... Using the derivative formula: ; When the current estimate deviates significantly from the true solution, the exponential term becomes larger and the derivative becomes more negative, which affects the correction step size. , Maintain a reasonable amplitude. When the current approaches the true solution, the residual approaches zero, and the correction step size naturally decreases, ensuring fine convergence of the iteration.

[0094] Since the virtual current is used for power calculation in the current control cycle, the iteration process cannot stop too early, causing current errors to propagate to the MPPT, nor can it iterate indefinitely, increasing the burden on the control cycle. Instead of using a fixed number of iterations as the sole stopping condition, the change in current estimates between two consecutive iterations is used as the convergence criterion. Iteration terminates when the current estimate has stabilized. The initial value of the iteration is taken as the virtual current value output in the previous control cycle. Hot start is achieved by utilizing the temporal continuity of the photovoltaic array output current between adjacent control cycles, which can significantly reduce the number of iterations. During the iteration process, the absolute value of the current change between two consecutive iterations is used as the criterion to determine if it is less than a preset convergence threshold. As the stopping criterion, the expression is: ; in The preset convergence threshold is used as the stopping criterion, which directly reflects the stability of the current estimation. When the change is small enough, it indicates that the current has stopped moving, that is, the equation is approximately satisfied, and the iteration is terminated.

[0095] Since MPPT control requires determining the voltage adjustment direction based on the current power change and does not rely on physical current sensors to obtain the current, the virtual current obtained from the modified model inversion needs to be converted into a power quantity that can be used by the control algorithm. Based on the real-time terminal voltage and virtual current within the same control cycle, the current Iv obtained after convergence is the virtual current. This is because the calculation uses... and This is the latest value after being corrected by transient disturbance data. The virtual current accurately reflects the true physical state of the photovoltaic array after shading. By directly multiplying the virtual current by the real-time terminal voltage, the power calculation formula is used: ; The current cycle power is obtained. This power value carries the physical description of the shadow effect of the correction model. It is more resistant to circuit noise and offset than the measurement value of the direct current sensor. This allows the maximum power point tracking algorithm to perform maximum power point search based on power information that conforms to the actual characteristics of the component, even without a physical current sensor, and achieve adaptive tracking under dynamic shadow changes.

[0096] Unlike existing technologies that calculate virtual current based on fixed parameter models or rely on physical current sensors to obtain current feedback, this method does not directly use the uncorrected photovoltaic model. Instead, it rewrites the series and parallel resistances, corrected by transient disturbance data, into the single diode model and solves the virtual current at the current operating point in real time through Newton-Raphson iteration.

[0097] Considering the issues of model parameter shifts after shadowing disturbances, the inability of fixed models to reflect current array characteristics, and the potential introduction of noise and zero drift by current sensors, the virtual current calculation is based on the constraints of the latest identified parameters and real-time terminal voltage. The resulting power value no longer depends solely on fixed empirical models or external sensor measurements, but is obtained by real-time inversion from the corrected physical model, thereby improving the consistency between power calculation and maximum power point tracking control under dynamic shadowing conditions.

[0098] Based on the data distribution characteristics of photovoltaic arrays under transient disturbance conditions, a unified design is adopted for the entire process of disturbance triggering, data construction, control variable constraints, parameter selection, and model solving. The calculation steps are not selected independently, but are formed by the constraints of data characteristics at each level: disturbance triggering determines the data entry conditions, control freezing ensures the singleness of data sources, parameter selection makes the identification problem meet the solvability conditions, recursive calculation matches the data temporal structure, and online solving ensures the consistency of model and parameter updates. This forms a calculation path driven by data characteristics, which is different from the processing methods based on fixed processes or empirical rules in existing technologies. Instead, it is a structured calculation process formed under the constraints of data distribution.

[0099] This embodiment also provides a voltage adaptive solar charging control system, including: a voltage acquisition module, a disturbance freezing module, an inversion calculation module, a trend verification module, and an adaptive adjustment module.

[0100] The voltage acquisition module is used to acquire the real-time terminal voltage of the photovoltaic array and calculate the voltage drop slope and trend of the real-time terminal voltage.

[0101] The disturbance freeze module is used to determine whether a transient disturbance event is triggered based on the voltage drop slope and trend. When a transient disturbance event is triggered, the maximum power point tracking operation is paused.

[0102] The inversion calculation module is used to call the preset photovoltaic array mechanism model to perform mathematical inversion calculation on the real-time terminal voltage to obtain the virtual current, and combine the real-time terminal voltage to calculate the real-time power of the current cycle.

[0103] The trend verification module is used to extract historical voltage and power data from previous cycles, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage regulation direction, and update the regulation direction within non-transient disturbance event cycles.

[0104] The adaptive adjustment module is used to calculate the adaptive dynamic step size based on the voltage error between the real-time terminal voltage and the target reference voltage, combined with a preset power law function, and adjust the output voltage of the photovoltaic array based on the voltage adjustment direction and the adaptive dynamic step size.

[0105] This embodiment also provides a computer device applicable to the voltage adaptive solar charging control method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the voltage adaptive solar charging control method proposed in the above embodiment.

[0106] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0107] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the voltage-adaptive solar charging control method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0108] In summary, this invention determines the voltage regulation direction by extracting historical data from previous cycles and combining it with the current real-time power for multi-point weighted trend verification. It utilizes historical trends in the time series to smooth out instantaneous noise interference, fundamentally suppressing control oscillations caused by measurement noise and ensuring absolute accuracy of the regulation direction. Based on the voltage error and using the nonlinear characteristics of a preset power-law function, it calculates an adaptive dynamic step size to adjust the output voltage. This allows the system to automatically adopt a large step size for rapid response when far from the target point, and automatically reduce the step size for fine-tuning when approaching the target point. This completely solves the inherent contradiction in traditional fixed-step-size algorithms where "high speed leads to large oscillations, and high precision leads to slow response." While significantly improving the dynamic response speed, it greatly reduces power loss and oscillation amplitude in steady state.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A voltage-adaptive solar charging control method, characterized in that: include: Collect the real-time terminal voltage of the photovoltaic array and calculate the voltage drop slope and trend of the real-time terminal voltage; The voltage drop slope and trend are used to determine whether a transient disturbance event has been triggered. When a transient disturbance event is triggered, the maximum power point tracking operation is paused. The virtual current is obtained by mathematically inverting the real-time terminal voltage using a pre-set photovoltaic array mechanism model, and the real-time power of the current cycle is calculated in combination with the real-time terminal voltage. Extract historical voltage and power data from previous cycles, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage regulation direction, and update the regulation direction within non-transient disturbance event cycles; Based on the voltage error between the real-time terminal voltage and the target reference voltage, an adaptive dynamic step size is calculated using a preset power law function, and the output voltage of the photovoltaic array is adjusted based on the voltage adjustment direction and the adaptive dynamic step size.

2. The voltage adaptive solar charging control method as described in claim 1, characterized in that: The transient disturbance event judgment step includes: The real-time voltage drop amplitude is compared with the preset voltage drop threshold, and the output power change trend is compared with the undisturbed power trend predicted based on the current photovoltaic array physical model. When the drop exceeds the threshold and the power trend deviation exceeds the preset range, a transient disturbance event is determined to have occurred, and a disturbance activation flag is generated.

3. The voltage adaptive solar charging control method as described in claim 2, characterized in that: The disturbance activation flag includes: In response to the disturbance activation flag, the update of the maximum power tracking duty cycle is stopped, the current duty cycle is maintained, the maximum power tracking state is switched to the tracking freeze state, and the data buffer for transient voltage and transient power is activated.

4. The voltage adaptive solar charging control method as described in claim 3, characterized in that: The data buffer includes: Using the trigger time of the disturbance activation flag as a synchronization reference, steady-state voltage and power data for a first preset duration before the disturbance occurs, and transient voltage and power data for a second preset duration including voltage drop and recovery edge after the disturbance occurs are extracted from the buffer and spliced ​​together to form a transient identification dataset.

5. The voltage adaptive solar charging control method as described in claim 4, characterized in that: The transient identification dataset includes: Using a single diode model, the partial derivatives of each characteristic parameter of the photovoltaic array with respect to the output power are calculated analytically, and the variation of the partial derivatives within the time period covered by the transient identification dataset is compared. The parameter with the largest change range is selected as the dominant parameter to be identified. The dominant parameter to be identified includes at least series resistance and parallel resistance.

6. The voltage adaptive solar charging control method as described in claim 5, characterized in that: The dominant parameters to be identified include: When a transient disturbance event is detected, the timing of the disturbance activation flag is used as the synchronization reference. The terminal voltage sequence and the corresponding power sequence of the photovoltaic array are collected only when the maximum power point tracking control is in a frozen state. Based on the disturbance activation flag, a transient identification dataset is constructed from the collected data. The transient identification dataset includes: steady-state voltage and power data within a first preset time before the disturbance occurs, and transient voltage and power data within a second preset time after the disturbance occurs, including voltage drop edge and recovery edge. The terminal voltage sequence in the transient identification dataset is used as the input of the recursive least squares algorithm, and the corresponding power sequence is used as the output reference. A recursive fitting model is established using the current-voltage implicit equation of a single diode model, wherein the parameters to be estimated in the recursive fitting model are limited to the dominant parameters to be identified, including series resistance and parallel resistance. During the recursive process, sampled data from the transient identification dataset are introduced point by point. The residual between the model output power and the output reference is calculated based on the current parameter estimate. The gain vector is updated based on the residual to correct the parameter estimate, thereby minimizing the cumulative sum of squared residuals within the range of the transient identification dataset. The transient identification dataset is formed by splicing the steady-state data before the disturbance and the transient data after the disturbance, so that the input data simultaneously includes the parameter-sensitive change range and the steady-state constraint range, thereby improving the convergence speed and identification accuracy of the dominant parameter to be identified. After traversing the entire transient identification dataset, the converged series resistance and parallel resistance estimates are output, and these estimates are written into the photovoltaic array physical model to replace the original parameters and obtain the corrected photovoltaic array physical model.

7. The voltage adaptive solar charging control method as described in claim 6, characterized in that: The revised photovoltaic array physical model includes: During the detection of transient disturbance events, the maximum power point tracking control algorithm is kept in a frozen state to keep the output duty cycle of the DC-DC converter constant, thereby acquiring the terminal voltage sequence and power sequence of the photovoltaic array under fixed operating point conditions; Based on the data collected during the freeze period, a steady-state pause dataset for verifying the photovoltaic array physical model is formed, and the updated series resistance and parallel resistance parameters are read from the corrected photovoltaic array physical model. The real-time terminal voltage, the updated series resistance and parallel resistance parameters, and the preset photocurrent parameters, reverse saturation current parameters and ideal factor parameters in the photovoltaic array physical model are all substituted into the current-voltage implicit equation of the single diode model. To address the nonlinear coupling characteristic where current simultaneously appears in the exponential and resistance voltage drop terms in the implicit equation, the Newton-Raphson iterative method is employed for numerical solution. In each iteration, the residual and derivative information of the equation are calculated, and the current estimate is updated. During the iteration process, the current change between two adjacent iterations is lower than the preset convergence threshold as the stopping condition, and the converged current solution is used as the virtual current value. The calculation of the virtual current is based on the photovoltaic array physical model that has been corrected by transient perturbation data, so that the virtual current can reflect the actual parameter changes after the perturbation, thereby improving the accuracy of power calculation under dynamic illumination conditions. The virtual current is multiplied by the real-time terminal voltage to obtain the current cycle power, and the power is input into the maximum power point tracking control algorithm for subsequent duty cycle adjustment.

8. A voltage-adaptive solar charging control system, based on the voltage-adaptive solar charging control method according to any one of claims 1 to 7, characterized in that: It includes a voltage acquisition module, a disturbance freezing module, an inversion calculation module, a trend verification module, and an adaptive adjustment module; The voltage acquisition module is used to acquire the real-time terminal voltage of the photovoltaic array and calculate the voltage drop slope and trend of the real-time terminal voltage. The disturbance freezing module is used to determine whether a transient disturbance event is triggered based on the voltage drop slope and change trend. When a transient disturbance event is triggered, the maximum power point tracking operation is paused. The inversion calculation module is used to call a preset photovoltaic array mechanism model to perform mathematical inversion calculation on the real-time terminal voltage to obtain the virtual current, and combine the real-time terminal voltage to calculate the real-time power of the current cycle. The trend verification module is used to extract historical voltage data and historical power data from the previous period, combine them with real-time power, perform multi-point weighted trend verification to determine the voltage adjustment direction, and update the adjustment direction within the non-transient disturbance event period. The adaptive adjustment module is used to calculate the adaptive dynamic step size based on the voltage error between the real-time terminal voltage and the target reference voltage, combined with a preset power law function, and adjust the output voltage of the photovoltaic array based on the voltage adjustment direction and the adaptive dynamic step size.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the voltage adaptive solar charging control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the voltage adaptive solar charging control method according to any one of claims 1 to 7.