Robust estimation method and system for photovoltaic model parameters under local shielding condition
By constructing an equivalent model of a photovoltaic array under partial shading conditions, selecting key parameters and optimizing them using a robust loss function, the accuracy and reliability issues of the photovoltaic array model under partial shading conditions were solved, achieving high-precision parameter identification and real-time engineering performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-31
- Publication Date
- 2026-03-10
AI Technical Summary
Existing photovoltaic array models have poor accuracy and reliability under partial shading conditions. Traditional identification methods are unreliable under complex working conditions, and traditional least squares methods will produce serious deviations in estimation results when faced with significant errors, making it difficult to meet the real-time requirements of engineering.
An equivalent model of a photovoltaic array under partial shading conditions is constructed. Key parameter sets are selected through parameter estimability and sensitivity analysis. A robust loss function of multi-dimensional residuals is used for optimization. The model is solved by combining a nonlinear optimization algorithm and the estimated value of the model is output.
It significantly improves the accuracy and reliability of photovoltaic models under complex lighting conditions and data error interference, can accurately describe multi-peak output characteristics, meet the real-time requirements of engineering, and reduce operation and maintenance costs.
Smart Images
Figure CN121637756A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of photovoltaic power generation system, and particularly relates to a robust estimation method and system for photovoltaic model parameters under local shading conditions. BACKGROUND
[0002] With the rapid development of social economy and the rapid improvement of living standards, human demand for energy is growing. However, due to the rapid consumption of fossil fuels and the serious destruction of the ecological environment, traditional energy has become increasingly difficult to meet the needs of human production and life. Photovoltaic power generation can generate electricity using solar energy and does not consume non-renewable energy. With the development of photovoltaic power generation technology, photovoltaic cells as the core unit of photovoltaic modules have gradually attracted more and more attention from experts and scholars in recent years.
[0003] In the field of photovoltaic power generation, accurate modeling of photovoltaic arrays is the basis for performance evaluation, maximum power point tracking and optimal control. However, there are still some significant bottlenecks in the existing technology. For example, key parameters such as series resistance Rs and diode quality factor A cannot be directly measured and need to rely on identification, while traditional identification methods are unreliable under complex working conditions. For example, under local shading conditions, the output characteristic curve of the photovoltaic array presents multiple peaks and steps, and the traditional single exponential model cannot accurately describe it. For example, sensor errors, environmental noise and other random errors and significant errors are inevitably contained in the measurement data, which will seriously distort the identification results of the traditional least squares method, and the weighted least squares method will produce serious bias in the face of significant errors, and the calculation is complex, which is difficult to meet the real-time requirements of engineering. Therefore, the existing photovoltaic array model has poor accuracy and reliability. SUMMARY
[0004] Therefore, in view of the above deficiencies, it is necessary to propose a robust estimation method and system for photovoltaic model parameters under local shading conditions to improve the accuracy and reliability of the photovoltaic model under complex lighting conditions and data error interference.
[0005] In a first aspect, the present application provides a robust estimation method for photovoltaic model parameters under local shading conditions, comprising:
[0006] Based on the output characteristic relationship of photovoltaic cells and arrays, an equivalent model of photovoltaic arrays under local shading conditions is constructed;
[0007] Based on parameter estimability or sensitivity analysis, the parameter set of the equivalent model of the photovoltaic array is screened to obtain a key parameter set;
[0008] The objective function is constructed by summing a plurality of dimensional residuals through a robust loss function, taking the set of key parameters as the only optimization variable; wherein the plurality of dimensional residuals in the objective function include at least two of a voltage residual, a current residual, an illumination residual, and a temperature residual;
[0009] Solving the objective function outputs an estimated value of the model.
[0010] Preferably, the equivalent model of the photovoltaic array under the local shading condition comprises:
[0011] A physical model of the photovoltaic cell under uniform illumination is obtained by describing the relationship between the output current and the output voltage of a single photovoltaic cell based on a physical mechanism.
[0012] An equivalent model of the photovoltaic array under the local shading condition is constructed; wherein the equivalent model of the photovoltaic array is composed of a plurality of sub-strings in parallel, each sub-string is composed of a plurality of photovoltaic cells in series, and a bypass diode is connected in parallel to each cell.
[0013] Preferably, the relationship between the output current I and the output voltage V of the physical model of the photovoltaic cell under uniform illumination is described as follows: ;
[0014] In the formula, is the photo-generated current, is the reverse saturation current of the diode, A is the diode quality factor, is the series resistance, is the parallel resistance, is the thermal voltage;
[0015] In the equivalent model of the photovoltaic array under the local shading condition, the output equation of the i-th photovoltaic cell is represented as: ; ;
[0016] In the formula, is the output current of the i-th photovoltaic cell, is the output voltage of the i-th photovoltaic cell, is the series resistance of the i-th photovoltaic cell, is the photo-generated current of the i-th photovoltaic cell, is the reverse saturation current of the diode of the i-th photovoltaic cell, is the voltage of the diode of the i-th photovoltaic cell, is the current flowing through the diode of the i-th photovoltaic cell, A i is the diode quality factor of the i-th photovoltaic cell, is the parallel resistance of the i-th photovoltaic cell;
[0017] The output current I and the output voltage V of the whole array are expressed as follows: ; ;
[0018] wherein N is the number of parallel sub-arrays.
[0019] Preferably, the parameter set of the photovoltaic array equivalent model is screened based on the parameter estimability, comprising:
[0020] constructing the parameter set of the photovoltaic array equivalent model and calculating the partial derivative matrix of the model output to the parameters to obtain a Jacobian matrix;
[0021] calculating a Fisher information matrix based on the Jacobian matrix and judging the parameter identifiability thereof by a condition number;
[0022] for the ill-conditioned parameter set determined, the parameter with the highest contribution degree is gradually proposed according to the contribution degree of each parameter to the condition number to obtain a key parameter set.
[0023] Preferably, the parameter set of the photovoltaic array equivalent model is screened based on the sensitivity analysis, comprising:
[0024] the relative sensitivity of each parameter in the parameter set of the photovoltaic array equivalent model is calculated by using the following calculation formula: ;
[0025] wherein, is the relative sensitivity of the ith parameter, I is the model output current.
[0026] the parameter with the relative sensitivity greater than a preset sensitivity threshold is determined as a key parameter.
[0027] Preferably, the objective function is expressed as follows: ;
[0028] wherein, is the residual of the model output voltage and the measured voltage, is the residual of the model output current and the measured current, is the residual of the model output light and the measured light, is the residual of the model output temperature and the measured temperature, is a robust loss function, and M is the total number of data groups.
[0029] Preferably, the robust loss function is a Cauchy loss function, a Correntropy loss function, a German-McClure loss function, a Welsch loss function, or a Lorentzian loss function.
[0030] Preferably, the solving the objective function comprises:
[0031] solving the objective function by using a nonlinear optimization algorithm; wherein the nonlinear optimization algorithm comprises a gradient descent method and a Levenberg-Marquardt algorithm.
[0032] In a second aspect, the present application provides a robust estimation system of photovoltaic model parameters under a local shading condition, comprising a data acquisition module, a model construction module, a parameter set screening module, a robust optimization solving module, and a model verification module.
[0033] The data acquisition module is configured to acquire multiple groups of measured data of output voltage, output current, light intensity, and temperature of a photovoltaic array during operation.
[0034] The model construction module is configured to construct an equivalent circuit model and output a parameter set to be identified.
[0035] The parameter set screening module is configured to screen the parameter set to be identified to obtain a key parameter set.
[0036] The robust optimization solving module is configured to construct an objective function, load a robust loss function, and call an optimization algorithm to solve the objective function, and output an estimated value of the model.
[0037] The model verification module is configured to substitute the estimated value into the equivalent circuit model to generate a complete output characteristic curve of the photovoltaic array, and compare and verify the complete output characteristic curve with the measured data.
[0038] In a third aspect, the present application provides a photovoltaic power generation system, comprising a photovoltaic array, a maximum power point tracking controller, an energy storage unit, and a robust estimation system of photovoltaic model parameters under a local shading condition as described in the second aspect; wherein the photovoltaic array provides original data input for the robust estimation system of photovoltaic model parameters under a local shading condition; the model parameters output by the robust estimation system of photovoltaic model parameters under a local shading condition are provided to the maximum power point tracking controller, which is used to accurately track a global maximum power point under a local shading condition, and output maximum power to the energy storage unit or a power grid.
[0039] From the above technical solution, the robust estimation method and system of photovoltaic model parameters under local shading conditions provided by the embodiment of the application can be known, when the robust estimation of photovoltaic model parameters is performed, first, based on the output characteristic relationship of the photovoltaic cell and array, the equivalent model of the photovoltaic array under the local shading condition is constructed, then based on the parameter estimability or sensitivity analysis, the parameter set of the equivalent model of the photovoltaic array is screened; further, taking the key parameter set screened as the only optimization variable, the objective function is constructed by the sum of the robust loss of multi-dimensional residual, and the objective function is solved, and the estimated value of the output model is combined. As can be seen, the present scheme can be used to accurately describe the multi-peak output characteristic by constructing the model under the local shading condition. Further, by screening the parameters, the rationality of the parameter set and the well-posedness of the identification problem can be ensured; finally, by the robust estimation for parameter identification, the negative influence of significant error can be effectively inhibited, so as to ensure the stability of the identification process and the accuracy of the result. The present scheme fundamentally solves the problem of large data error interference caused by improper selection of parameter set, and can significantly improve the accuracy and reliability of the photovoltaic model under complex light conditions and data pollution. BRIEF DESCRIPTION OF DRAWINGS
[0040] Figure 1 The flowchart of the robust estimation method of photovoltaic model parameters under local shading conditions provided by the embodiment of the application.
[0041] Figure 2 The schematic diagram of the robust estimation system of photovoltaic model parameters under local shading conditions provided by the embodiment of the application.
[0042] Figure 3 The schematic diagram of the photovoltaic power generation system provided by the embodiment of the application. DETAILED DESCRIPTION
[0043] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiments will be briefly introduced below, and obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can also obtain other drawings according to these drawings without creating labor.
[0044] As shown in Figure 1 The present application provides a robust estimation method of photovoltaic model parameters under local shading conditions, which can include the following steps:
[0045] Step 101: based on the output characteristic relationship of the photovoltaic cell and array, the equivalent model of the photovoltaic array under the local shading condition is constructed;
[0046] Step 102: Based on parameter estimability or sensitivity analysis, the parameter set of the photovoltaic array equivalent model is screened to obtain the key parameter set;
[0047] Step 103: Using the set of key parameters as the sole optimization variable, construct an objective function by summing the robust losses of multi-dimensional residuals; wherein, the multi-dimensional residuals in the objective function include at least two of the following: voltage residual, current residual, illuminance residual, and temperature residual;
[0048] Step 104: Solve the objective function and output the estimated value of the model.
[0049] In this embodiment, when performing robust estimation of photovoltaic model parameters, an equivalent photovoltaic array model under partial shading conditions is first constructed based on the output characteristic relationship between photovoltaic cells and the array. Then, based on parameter estimability or sensitivity analysis, the parameter set of the equivalent photovoltaic array model is screened. Furthermore, using the screened key parameter set as the sole optimization variable, an objective function is constructed by summing the robust losses of multi-dimensional residuals, and this objective function is solved, combined with the estimated value of the output model. Therefore, this scheme, by constructing a model under partial shading conditions, can accurately describe multi-peak output characteristics. Furthermore, by screening parameters, the rationality of the parameter set and the well-posedness of the identification problem can be ensured. Finally, parameter identification through robust estimation can effectively suppress the negative impact of significant errors, ensuring the stability of the identification process and the accuracy of the results. This scheme fundamentally solves the major problems of improper parameter set selection and large amounts of data error interference, significantly improving the accuracy, reliability, and engineering practical value of photovoltaic models under complex lighting conditions and data contamination.
[0050] The following provides a more detailed explanation of each step.
[0051] Step 101: Based on the relationship between the output characteristics of photovoltaic cells and arrays, construct an equivalent model of photovoltaic arrays under partial shading conditions.
[0052] This step aims to establish a full-condition model covering both uniform illumination and partial shading, providing accurate mathematical identification for parameter identification. Specifically, constructing the equivalent model of the photovoltaic array under partial shading conditions can be achieved through the following two sub-steps:
[0053] (1) Based on the physical mechanism, the relationship between the output current and voltage of a single photovoltaic cell is described, resulting in a physical model of the photovoltaic cell under uniform illumination. The relationship between the output current I and output voltage V of the physical model of the photovoltaic cell under uniform illumination is described as follows: ;
[0054] In the formula, for the photo-generated current, for the diode reverse saturation current, A is the diode quality factor, for the series resistance, for the parallel resistance, for the thermal voltage.
[0055] (2) constructing an equivalent model of a photovoltaic array under local shading conditions; wherein the photovoltaic array equivalent model is composed of multiple sub-strings in parallel, each sub-string is composed of multiple photovoltaic cells in series, and a bypass diode is connected in parallel to each cell. Wherein the output equation of the i-th photovoltaic cell in the equivalent model of the photovoltaic array under local shading conditions is represented as: ; ;
[0056] In the formula, is the output current of the i-th photovoltaic cell, is the output voltage of the i-th photovoltaic cell, is the series resistance of the i-th photovoltaic cell, is the photo-generated current of the i-th photovoltaic cell, is the diode reverse saturation current of the i-th photovoltaic cell, is the voltage of the diode of the i-th photovoltaic cell, is the current flowing through the diode of the i-th photovoltaic cell, A i is the diode quality factor of the i-th photovoltaic cell, is the parallel resistance of the i-th photovoltaic cell;
[0057] The output current I and the output voltage V of the entire array are represented as follows: ;
[0058] In the formula, N is the number of parallel sub-strings.
[0059] In this embodiment, considering the multi-peak characteristics caused by shading, a hierarchical approach of substring division and diode judgment is adopted. For substring division, photovoltaic cells with consistent illumination intensity and temperature in the photovoltaic array can be classified into one substring, each substring is composed of several cells in series, and each cell is connected in reverse parallel with one bypass diode, and the rated current of the cell is selected and matched to prevent hot spot effect. Further, the substrings with the same shading mode are connected in parallel to form a subarray, and the entire photovoltaic array is composed of several subarrays connected in parallel. For the state judgment of the bypass diode, it is determined by the terminal voltage of the photovoltaic cell and the light condition, so as to simulate the multi-peak output characteristics under local shading. Specifically, when the photocurrent of a certain cell in the substring is less than that of other cells, the cell becomes a current bottleneck, and the terminal voltage reversely rises to the diode conduction voltage, at which time the diode is regarded as a small resistor; when the photocurrent of all cells is consistent, the diode terminal voltage is lower than the conduction voltage, and it is regarded as a large resistor.
[0060] In this embodiment, the parameter set of the model can be, for example, .
[0061] Step 102, based on the parameter estimability or sensitivity analysis, the parameter set of the photovoltaic array equivalent model is screened to obtain a key parameter set.
[0062] In this step, the purpose is to analyze the parameter identifiability and the influence degree on the model output, eliminate the unidentifiable or low sensitivity parameters, and ensure the rationality of the parameter set. Specifically, when the parameter set of the photovoltaic array equivalent model is screened based on the parameter estimability, it can be realized by the following way:
[0063] Constructing the parameter set of the photovoltaic array equivalent model, and calculating the partial derivative matrix of the model output to the parameters to obtain the Jacobian matrix;
[0064] Calculating the Fisher information matrix based on the Jacobian matrix, and judging the parameter identifiability by the condition number;
[0065] For the ill-conditioned parameter set determined, the parameter with the highest contribution degree is gradually proposed according to the contribution degree of each parameter to the condition number, to obtain the key parameter set.
[0066] In this embodiment, first, the partial derivative matrix of the model output to each parameter can be calculated based on the measured data of voltage, current, illumination, temperature, etc., that is, the Jacobian matrix is obtained. For example, the Jacobian matrix can be represented as follows: ;
[0067] Further, the Fisher information matrix is calculated where W is a weight matrix, which can be set according to sensor accuracy. Then, a condition number judgment can be performed: the parameter identifiability is judged by the ratio of the maximum singular value and the minimum singular value of F, cond(F). If cond(F) is less than 10 3 , the parameter set is identifiable; if cond(F) is greater than 10 6 , the parameter set is ill-conditioned, and there is an unidentifiable parameter. Subsequently, for the ill-conditioned parameter set, the contribution degree of each parameter to the condition number is calculated, such as the partial derivative absolute value of R sh is the smallest, the contribution degree is the highest, and the parameter with the highest contribution degree is gradually removed until cond(F) is less than 103. For example, the key parameter set to be identified can be finally screened out, and Rsh and I0 are fixed by an empirical formula or a standard value to reduce the optimization dimension and improve efficiency.
[0068] When the parameter set of the photovoltaic array equivalent model is screened based on sensitivity analysis, the following methods can be used:
[0069] The relative sensitivity of each parameter in the parameter set of the photovoltaic array equivalent model is calculated by using the following calculation formula: ;
[0070] In the formula, S is the relative sensitivity of the ith parameter, and I is the output current of the model.
[0071] The parameters with a relative sensitivity greater than a preset sensitivity threshold are determined as key parameters.
[0072] For example, the relative sensitivity formula is used to calculate: , , , The parameters with a relative sensitivity greater than 0.5, i.e., the parameters with a significant impact on the output, are retained, and the parameters with a sensitivity less than 0.3, i.e., the parameters with a weak impact, are removed, and finally the key parameter set is screened out.
[0073] Of course, in some embodiments, when the parameter screening is performed, the screening method based on parameter estimability and the screening method based on sensitivity analysis can be combined to avoid missing key parameters by mutual verification. For example, the sensitivity method screens a parameter as a key parameter, and the screening method based on parameter estimability screens the parameter as a fixed parameter, which causes the parameter to be missed, affecting the final model output result.
[0074] Step 103, taking the key parameter set as the only optimization variable, constructing an objective function through the sum of multi-dimensional residual robust loss; wherein the multi-dimensional residual in the objective function includes at least two of voltage residual, current residual, illumination residual and temperature residual.
[0075] In this step, the loss function is designed for gross errors in the measurement data, and the optimization problem is constructed to realize stable estimation of the parameters. Specifically, considering the significant errors in the measurement data, the parameter identification problem is constructed as an optimization problem by using robust estimation method, and the objective function constructed can be expressed as follows: ;
[0076] In the formula, is the residual of the model output voltage and the measured voltage, is the residual of the model output current and the measured current, is the residual of the model output illumination and the measured illumination, is the residual of the model output temperature and the measured temperature, is the robust loss function, and M is the total number of data groups.
[0077] In this embodiment, the robust loss function is Cauchy loss function, Correntropy loss function, German-McClure loss function, Welsch loss function or Lorentzian loss function. In addition, the robust loss function contains an adjustment parameter, and the value of the adjustment parameter is set by the relative efficiency criterion, so that the robust estimator reaches the predetermined relative efficiency under the assumption that the measurement error obeys the normal distribution. For example, the robust loss function is Correntropy loss function, and its core form is: ρ(e)∝1-exp(-e² / (2σ²)), where σ is the adjustment parameter, and the predetermined relative efficiency is 95%.
[0078] Step 104, solving the objective function to output the estimated value of the model.
[0079] When solving the objective function, a nonlinear optimization algorithm can be used to solve the objective function; wherein the nonlinear optimization algorithm includes gradient descent method and Levenberg-Marquardt algorithm. For example, when the LM algorithm which takes into account the stability of gradient descent and the convergence speed of Newton method is used to solve the objective function, the main process of solving can be as follows:
[0080] (1) Initial value setting: set A0=1.5, R s0 =0.1;
[0081] (2) Iterative update: first calculate the Jacobian matrix J and the residual vector , and then adjust the damping factor ; wherein the initial damping factor can be set to 0.01, and if the objective function decreases, the damping factor is reduced, and vice versa. Further, update the parameters , I is the unit matrix, and W is the loss function weight matrix.
[0082] (3) Convergence judgment: when the parameter update amount or the objective function change , stop iteration and output the optimal parameter estimate value.
[0083] In addition, as shown in Figure 2 , the application also provides a robust estimation system for photovoltaic model parameters under local shading conditions, comprising: a data acquisition module 201, a model construction module 202, a parameter set screening module 203, a robust optimization solving module 204 and a model verification module 205:
[0084] The data acquisition module 201 is configured to acquire multiple groups of measured data of output voltage, output current, light intensity and temperature of the photovoltaic array in the running process;
[0085] The model construction module 202 is configured to construct an equivalent circuit model and output a to-be-identified parameter set;
[0086] The parameter set screening module 203 is configured to screen the to-be-identified parameter set to obtain a key parameter set;
[0087] The robust optimization solving module 204 is configured to construct an objective function, load a robust loss function and call an optimization algorithm to solve the objective function, and output an estimate value of the model;
[0088] The model verification module 205 is configured to substitute the estimate value into the equivalent circuit model to generate a complete output characteristic curve of the photovoltaic array, and compare and verify with the measured data.
[0089] Further, as shown in Figure 3 , the application also provides a photovoltaic power generation system, comprising: a photovoltaic array 301, a maximum power point tracking controller 302, an energy storage unit 303 and a robust estimation system for photovoltaic model parameters under local shading conditions 200 as described above; wherein the photovoltaic array 301 provides raw data input for the robust estimation system for photovoltaic model parameters under local shading conditions 200; the model parameters output by the robust estimation system for photovoltaic model parameters under local shading conditions 200 are provided to the maximum power point tracking controller 302, which is used to accurately track the global maximum power point under local shading conditions, and output the maximum power to the energy storage unit 303 or the power grid.
[0090] The effect of the present solution is further illustrated below with specific examples:
[0091] (1) Model establishment: take a photovoltaic array with 4 sub-strings in parallel and 7 cells in series in each sub-string as an example. Set 2 cells to be shaded and the light intensity to be reduced to 30% of the normal value. According to the model, the mathematical relationship is established.
[0092] (2) Data collection: under the conditions of array temperature 32°C and normal part light intensity 450 W / m², 52 groups of voltage-current data are collected by changing the load resistance . Among them, 5 groups of significant errors with amplitude of 15%-30% are randomly introduced to simulate abnormal data.
[0093] (3) Parameter identification
[0094] 1) Set the to-be-identified parameters as θ=[A, Rs].
[0095] 2) Respectively use the traditional weighted least squares method and the multiple robust estimators (such as Correntropy, German-McClure) of the present solution to construct the objective function and solve.
[0096] 3) The objective function of WLS is: .
[0097] 4) The objective function of German-McClure estimator is: .
[0098] (4) Result analysis
[0099] The performance index adopts mean square error: .
[0100] The results show that when there are significant errors, the MSE of WLS is 0.1623, while the MSE of robust estimators such as Correntropy and German-McClure can be reduced to 0.0555, and the accuracy is significantly improved. The fitting degree of the corrected I-V curve to the true curve is much higher than that of the WLS method.
[0101] From the above, the robust estimation method and system for photovoltaic model parameters under local shading conditions provided by the present solution at least have the following beneficial effects:
[0102] (1) Parameter identification accuracy is improved significantly: By analyzing the parameter estimability and sensitivity, the key parameters are screened, avoiding the interference of unidentifiable parameters, and the rationality of the parameter set is improved. In the measurement data containing gross errors, the mean square error (MSE) of model parameter estimation is also significantly reduced compared with the traditional weighted least squares method, and the estimation bias is controlled within a small degree.
[0103] (2) Strong adaptability of the model to local shading conditions: The sub-string-sub-array hierarchical model can accurately describe the multi-peak output characteristics of the photovoltaic array under local shading. In the cell shading scenario, the fitting degree R 2 of the current-voltage curve to the measured data is greater than 0.99, which is much higher than the fitting effect of the traditional model.
[0104] (3) Strong anti-gross error capability: By using robust loss functions (such as Correntropy, German-McClure, etc.), the significant errors (gross errors) in the measurement data are punished to saturation. When the measurement data contains certain gross errors, it can still output high-precision parameters stably, solving the problem of serious deviation of the estimation result of the traditional method under data pollution.
[0105] (4) Wide adaptability to working conditions: Through the adaptive selection strategy of the loss function (switching Correntropy, Cauchy, Welsch, etc. according to the proportion of gross errors), various scenes from slight interference to serious data pollution can be covered, ensuring stable identification performance under different shading degrees and different measurement errors.
[0106] (5) Real-time performance meets engineering requirements: By using the Levenberg-Marquardt (LM) optimization algorithm, the convergence time of parameter identification is significantly reduced, which can be directly integrated into the maximum power point tracking (MPPT) controller to meet the real-time control requirements of the photovoltaic power generation system.
[0107] (6) Reducing system operation and maintenance costs: Accurate model parameters can support the MPPT controller to quickly track the global maximum power point under local shading, improving the power generation efficiency; at the same time, the robust design reduces the probability of system failure caused by parameter errors, reducing the equipment operation and maintenance costs.
[0108] The application also provides a computing device comprising a memory and a processor, the memory storing executable code, and the processor executing the executable code to run the method of any of the above embodiments.
[0109] The application also provides a computer-readable storage medium storing a computer program, which, when executed in a computer, causes the computer to run the method of any of the above embodiments.
[0110] The device embodiment of the present application and the method embodiment are based on the same inventive concept, and the detailed description can be referred to the method embodiment, which will not be described here again.
[0111] The modules or units in the device embodiments of the present application can be combined, divided, and deleted according to actual needs. The above disclosure is only the preferred embodiments of the present application, and of course cannot limit the scope of the rights of the present application. Those skilled in the art can understand that all or part of the processes of the above embodiments are implemented, and equivalent changes made according to the claims of the present application still belong to the scope covered by the present application.
Claims
1. A method for robust estimation of photovoltaic model parameters under partial shading conditions, characterized in that, The method comprises the following steps: Based on the output characteristic relationship of photovoltaic cells and arrays, an equivalent model of photovoltaic arrays under local shading conditions is constructed; Based on parameter estimability or sensitivity analysis, a set of key parameters is obtained by screening the parameter set of the equivalent model of the photovoltaic array; A target function is constructed by taking the set of key parameters as the only optimization variable and by using the sum of robust losses of multidimensional residuals; wherein the multidimensional residuals in the target function include at least two of voltage residuals, current residuals, illumination residuals and temperature residuals; The target function is solved to output the estimated value of the model.
2. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 1, characterized in that, The construction of the equivalent model of the photovoltaic array under the local shading condition comprises: A physical model of a photovoltaic cell under uniform illumination is obtained by describing the relationship between the output current and voltage of a single photovoltaic cell based on the physical mechanism; An equivalent model of a photovoltaic array under local shading conditions is constructed; wherein the equivalent model of the photovoltaic array is composed of multiple parallel sub-strings, each sub-string is composed of multiple photovoltaic cells in series, and a bypass diode is connected in parallel to each cell.
3. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 2, characterized in that, The relationship between the output current I and the output voltage V of the physical model of the photovoltaic cell under uniform illumination is described as follows: ； wherein is the photo-generated current, is the diode reverse saturation current, A is the diode quality factor, is the series resistance, is the parallel resistance, is the thermal voltage; In the equivalent model of the photovoltaic array under the local shading condition, the output equation of the i-th photovoltaic cell is represented as: ； ； wherein is the output current of the i-th photovoltaic cell, is the output voltage of the i-th photovoltaic cell, is the series resistance of the i-th photovoltaic cell, is the photo-generated current of the i-th photovoltaic cell, is the diode reverse saturation current of the i-th photovoltaic cell, is the voltage of the diode of the i-th photovoltaic cell, is the current flowing through the diode of the i-th photovoltaic cell, A i is the diode quality factor of the i-th photovoltaic cell, is the parallel resistance of the i-th photovoltaic cell; The output current I and the output voltage V of the entire array are represented as follows: ； ； In the formula, N is the number of parallel sub-strings.
4. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 1, characterized in that, The screening of the parameter set of the equivalent model of the photovoltaic array based on parameter estimability comprises: A parameter set of the equivalent model of the photovoltaic array is constructed, and a partial derivative matrix of the model output with respect to the parameters is calculated to obtain a Jacobian matrix; The Fisher information matrix is calculated based on the Jacobian matrix, and the parameter identifiability is judged by the condition number; For the ill-conditioned parameter set, the parameter with the highest contribution degree is gradually proposed according to the contribution degree of each parameter to the condition number, and a set of key parameters is obtained.
5. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 1, characterized in that, The screening of the parameter set of the equivalent model of the photovoltaic array based on sensitivity analysis comprises: The relative sensitivity of each parameter in the parameter set of the equivalent model of the photovoltaic array is calculated by using the following calculation formula: ； wherein is the relative sensitivity of the i-th parameter , I is the model output current; The parameters with a relative sensitivity greater than a preset sensitivity threshold are determined as key parameters.
6. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 1, characterized in that, The target function is represented as follows: ； wherein, is the residual of the model output voltage and the measured voltage, is the residual of the model output current and the measured current, is the residual of the model output illumination and the measured illumination, is the residual of the model output temperature and the measured temperature, is the robust loss function, and M is the total number of data sets.
7. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 6, characterized in that, The robust loss function is a Cauchy loss function, a Correntropy loss function, a German-McClure loss function, a Welsch loss function or a Lorentzian loss function.
8. The method for robust estimation of photovoltaic model parameters under partial shading conditions according to claim 1, characterized in that, The solving of the target function comprises: A nonlinear optimization algorithm is used to solve the target function; wherein the nonlinear optimization algorithm includes a gradient descent method and a Levenberg-Marquardt algorithm. 9.A system for robust estimation of photovoltaic model parameters under partial shading conditions, characterized in that, The method comprises the following steps: A data acquisition module, a model construction module, a parameter set screening module, a robust optimization solving module and a model verification module are provided: The data acquisition module is configured to obtain multiple groups of measured data of the output voltage, the output current, the illumination intensity and the temperature of the photovoltaic array during operation; The model construction module is configured to construct an equivalent circuit model and output a set of parameters to be identified; The parameter set screening module is configured to screen the to-be-recognized parameter set to obtain a key parameter set; The robust optimization solving module is configured to construct a target function, load a robust loss function, and call an optimization algorithm to solve the target function, and output an estimated value of the model; The model verification module is configured to substitute the estimated value into the equivalent circuit model to generate a complete output characteristic curve of the photovoltaic array, and compare and verify the complete output characteristic curve with measured data.
10. A photovoltaic power system, characterized by, Comprise: A photovoltaic array, a maximum power point tracking controller, an energy storage unit, and the robust estimation system of photovoltaic model parameters under a local shading condition according to claim 9; wherein the photovoltaic array provides original data input for the robust estimation system of photovoltaic model parameters under the local shading condition; the model parameters output by the robust estimation system of photovoltaic model parameters under the local shading condition are provided to the maximum power point tracking controller, which is used to accurately track a global maximum power point under the local shading condition, and output maximum power to the energy storage unit or a power grid.