A photovoltaic power prediction method

Abstract

The present application relates to the field of photovoltaic power prediction, and particularly relates to a photovoltaic power prediction method, which obtains spectral response function data of different photovoltaic components in a photovoltaic power station, and measures spectral irradiance distribution of surfaces of each photovoltaic component in real time; according to the spectral response function data and the spectral irradiance distribution, effective spectral irradiance of each component type is calculated according to component type grouping; temperature coefficients of batches to which each photovoltaic component belongs are counted according to historical data, and a component-level temperature-power correction factor is established; string-level temperature correction power is obtained according to string topology relationship; the string-level temperature correction power, effective spectral irradiance and global irradiance data of numerical weather prediction are used to output a corrected power station-level photovoltaic power prediction value; the problem of low photovoltaic power prediction accuracy caused by inaccurate modeling due to photovoltaic component parameter difference and temperature influence in photovoltaic power station power prediction is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic power prediction, and particularly relates to a photovoltaic power prediction method. BACKGROUND

[0002] As an important part of renewable energy, the installed capacity of photovoltaic power generation continues to grow rapidly. However, the output power of photovoltaic power stations is significantly affected by environmental factors and the characteristics of the components themselves, resulting in greater uncertainty in power prediction. Accurate power prediction is a key technical basis for grid dispatching, energy storage system optimization and power market transactions, and is crucial to improving grid stability and economic benefits.

[0003] Currently, photovoltaic power prediction does not fully consider the spectral response differences and temperature coefficient dispersion of different batches and types of photovoltaic components. Especially in large-scale photovoltaic power stations, mixed batches of components lead to systematic error accumulation, significantly reducing prediction accuracy. And it does not fully consider the influence of dynamic factors such as photovoltaic component aging and changes in packaging material thermal resistance, resulting in a gradual increase in photovoltaic power error after long-term operation.

[0004] Therefore, there is an urgent need for a high-precision photovoltaic power prediction method that can deeply integrate the heterogeneity characteristics of components and dynamic thermodynamic behavior to improve the accuracy of photovoltaic power prediction. SUMMARY

[0005] (1) Technical problems to be solved

[0006] The purpose of the present application is to provide a photovoltaic power prediction method to solve the problem of low photovoltaic power prediction accuracy caused by inaccurate modeling of photovoltaic component parameter differences and temperature effects in photovoltaic power station power prediction.

[0007] (2) Technical solutions

[0008] To achieve the above purpose, the present application provides a photovoltaic power prediction method, which comprises:

[0009] S1, real-time acquisition of spectral response function data of different photovoltaic components in a photovoltaic power station, wherein the different photovoltaic components include at least two types of components whose spectral response characteristics differ by more than a preset difference threshold.

[0010] S2, real-time measurement of the spectral irradiance distribution incident on the surface of each photovoltaic component by a multispectral sensor network deployed above the photovoltaic array; according to the spectral response function data and the spectral irradiance distribution, the effective spectral irradiance of each component type is calculated by grouping according to component type.

[0011] S3, according to the historical data statistics each photovoltaic component belongs to the temperature coefficient of the batch, and establish component level temperature-power correction factor; according to the group string topological relationship, the temperature-power correction factor of all photovoltaic components in the same group string is superimposed to obtain the group string level temperature correction power.

[0012] S4, the group string level temperature correction power and effective spectral irradiance are input into the pre-trained photovoltaic power prediction model, and the global irradiance data of numerical weather prediction are combined to dynamically output the corrected power prediction value of the power station level photovoltaic power.

[0013] Further, the method of calculating the effective spectral irradiance of each component type according to the spectral response function data and the spectral irradiance distribution comprises:

[0014] The spectral irradiance distribution is discretized according to the preset wavelength interval, and the corresponding irradiance value and wavelength range in each wavelength interval are obtained.

[0015] The spectral response contribution value of each wavelength interval is calculated according to the response value of each wavelength interval in the spectral response function data of the component type and the measured irradiance value of the corresponding wavelength interval, and the spectral response contribution values of all wavelength intervals are weighted and summed to obtain the initial effective spectral irradiance of the component type.

[0016] According to the standard solar spectrum distribution data under the standard test condition, the initial effective spectral irradiance is normalized to obtain the effective spectral irradiance.

[0017] Further, the method of calculating the effective spectral irradiance of each component type according to the spectral response function data and the spectral irradiance distribution comprises:

[0018] According to the unique identification code of the photovoltaic component, the production batch information is associated, the temperature coefficient data set of each production batch photovoltaic component in the historical database is extracted, and the temperature coefficient mean value and dispersion parameter of each production batch photovoltaic component are calculated.

[0019] The backboard temperature data of the photovoltaic component is collected in real time, and the temperature-power correction factor of the photovoltaic component level is dynamically calculated according to the temperature coefficient mean value and dispersion parameter of the production batch to which it belongs; wherein the temperature-power correction factor of the production batch photovoltaic component whose dispersion exceeds the preset dispersion threshold is introduced into the dispersion compensation weight, and the dispersion compensation weight is negatively correlated with the dispersion parameter.

[0020] If the production batch to which the photovoltaic component belongs is mixed and the temperature coefficient difference exceeds the preset temperature coefficient difference threshold, then the temperature-power correction factor weight is adjusted according to the electrical connection relationship of the photovoltaic components in the group string.

[0021] When it is detected that the temperature coefficient dispersion of the production batch photovoltaic module in the string continuously exceeds the historical statistical range, a step temperature control excitation signal is applied to the production batch photovoltaic module, and the transient change slope of the temperature response curve is collected to inversely calculate the actual temperature coefficient, update the temperature coefficient mean and dispersion parameters of the production batch photovoltaic module, and regenerate the component-level temperature-power correction factor.

[0022] Further, the method of applying a step temperature control excitation signal to the production batch photovoltaic module when it is detected that the temperature coefficient dispersion of the production batch photovoltaic module in the string continuously exceeds the historical statistical range, and collecting the transient change slope of the temperature response curve to inversely calculate the actual temperature coefficient comprises:

[0023] The step temperature control excitation signal is applied through a temperature control actuator integrated in the backsheet of the photovoltaic module, and the transient change data of the backsheet temperature of the photovoltaic module is synchronously collected at a preset sampling frequency.

[0024] The transient change data is filtered and denoised to extract the response curve of each temperature step interval and calculate the instantaneous change slope and steady-state offset, and a temperature-time transfer function is constructed in combination with a dynamic heat capacity model of the photovoltaic module.

[0025] The actual temperature coefficient is inversely calculated and the confidence interval is calculated according to the matching of the temperature-time transfer function and a preset standard temperature coefficient response database through least squares method iteration optimization; and if the half-width of the confidence interval is less than a preset precision threshold, the actual temperature coefficient is output.

[0026] Further, the method of filtering and denoising the transient change data, extracting the response curve of each temperature step interval, calculating the instantaneous change slope and steady-state offset, and constructing a temperature-time transfer function in combination with a dynamic heat capacity model of the photovoltaic module comprises:

[0027] The starting point and the ending point of the temperature step interval are identified through a dynamic threshold segmentation algorithm, the response curve is fitted through cubic spline interpolation in each temperature step interval after segmentation, the instantaneous change slope of each sampling point on the response curve is calculated, and the maximum instantaneous change slope is extracted as a step response characteristic value.

[0028] A dynamic heat capacity model is established according to the material parameters and the structural parameters of the photovoltaic module; the step response characteristic value and the steady-state offset are input into the dynamic heat capacity model to obtain a parameter matrix of the temperature-time transfer function, and the parameter matrix is calibrated in real time through a Kalman filtering algorithm; and if the calibration residual of the parameter matrix exceeds a preset residual tolerance threshold, the adaptive update of the dynamic heat capacity model is triggered.

[0029] Further, the method of establishing a dynamic heat capacity model according to the material parameters and the structural parameters of the photovoltaic module comprises:

[0030] constructing a three-dimensional non-uniform grid model according to material parameters and structure parameters of the photovoltaic module, the material parameters being physical characteristic parameters of each constituent material of the photovoltaic module, and the structure parameters being geometric characteristic and physical layout parameters of the photovoltaic module;

[0031] correcting a degradation coefficient of the material parameters according to historical aging data of the photovoltaic module, the degradation coefficient including a thermal resistance annual growth rate of a packaging material and a decay curve of a cell thermal capacity, and generating an initial thermal resistance-thermal capacity correlation matrix according to the corrected material parameters.

[0032] iteratively optimizing the initial thermal resistance-thermal capacity correlation matrix and three-dimensional non-uniform grid model combined with real-time collected backboard temperature gradient distribution data to obtain dynamic thermal capacity parameters, and mapping and generating a dynamic thermal capacity model according to the optimized dynamic thermal capacity parameters.

[0033] Further, the method of mapping and generating a dynamic thermal capacity model according to the optimized dynamic thermal capacity parameters comprises:

[0034] mapping the optimized dynamic thermal capacity parameters to the three-dimensional non-uniform grid model according to grid nodes, extracting thermal resistance and thermal capacity values of each grid node and converting them into equivalent lumped parameters, and constructing a dynamic thermal capacity model according to the equivalent lumped parameters.

[0035] carrying out frequency domain and time domain double verification on the dynamic thermal capacity model, verifying amplitude-frequency characteristics and phase characteristics of a transfer function in the frequency domain through a sweep excitation test, verifying a root mean square value of a residual error between a predicted temperature and an actual temperature in the time domain through a step temperature control experiment, and binding the dynamic thermal capacity model with a unique identification code of the photovoltaic module if verification results of the frequency domain and time domain double verification meet a preset verification accuracy.

[0036] Further, the method of inputting the string-level temperature correction power and effective spectral irradiance into a pre-trained photovoltaic power prediction model and combining global irradiance data of a numerical weather forecast to dynamically output a corrected power station-level photovoltaic power prediction value comprises:

[0037] summarizing each string-level temperature correction power into a power station-level temperature correction power matrix according to a string topology relationship and a power station electrical connection diagram, fusing the power station-level temperature correction power matrix, effective spectral irradiance and global irradiance data of a numerical weather forecast into a multi-dimensional input feature vector after time sequence alignment processing, inputting the multi-dimensional input feature vector into a pre-trained photovoltaic power prediction model, and obtaining the photovoltaic power prediction model through mixed neural network training.

[0038] if an extreme weather event warning is detected, triggering a multi-scale time window prediction to obtain a final power station-level photovoltaic power prediction value.

[0039] The final power station level photovoltaic power prediction value is compared with real-time power monitoring data to obtain a power prediction error value, and if the power prediction errors of consecutive N time windows exceed a preset power error threshold, the pre-trained photovoltaic power prediction model is adaptively adjusted, and N is greater than or equal to 3.

[0040] Further, the method for triggering multi-scale time window prediction to obtain the final power station level photovoltaic power prediction value if an extreme weather event warning is detected comprises:

[0041] The multi-scale time window prediction comprises short-term prediction and medium and long-term prediction; the short-term prediction generates high-frequency prediction results in a minute-level time window; the medium and long-term prediction generates trend prediction results in a hour-level time window; and the prediction results of the short-term prediction and the medium and long-term prediction are dynamically weighted and fused according to the grade of the extreme weather event warning and the duration of the weather event to obtain the final power station level photovoltaic power prediction value.

[0042] (3) Beneficial effects

[0043] Compared with the prior art, the beneficial effects of the present application are:

[0044] 1. By considering the spectral response difference of different photovoltaic components, measuring the spectral irradiance distribution in real time, combining with the component level temperature monitoring and temperature coefficient correction, the prediction accuracy is greatly improved, especially in the complex power station system with mixed different photovoltaic components, high-precision photovoltaic power prediction is realized.

[0045] 2. A dynamic adaptive temperature compensation mechanism is established, the actual temperature coefficient is inversely calculated through a ladder type temperature control excitation signal, and a dynamic heat capacity model is constructed according to the component material properties and structure parameters, so that the real-time tracking and correction of the change of the component temperature coefficient with aging are realized, and the prediction deviation problem caused by the aging or batch difference of the component temperature coefficient which cannot be solved by the traditional method is solved.

[0046] 3. The extreme weather response capability is provided, the dynamic weighted fusion of the multi-scale time window prediction is combined with the adaptive adjustment mechanism of the model, and the power prediction stability and reliability under extreme weather conditions are effectively improved, so that more accurate decision basis is provided for power grid dispatching and photovoltaic power station management. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 It is a flowchart of a photovoltaic power prediction method of the embodiment 1 of the present application. DETAILED DESCRIPTION

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Before giving examples, it is necessary to explain the application scenarios of the present invention. The present invention is a photovoltaic power prediction method. The impact of the heterogeneity of photovoltaic module characteristics on power prediction is a fundamental problem that has been underestimated for a long time. Different photovoltaic modules (such as monocrystalline silicon and thin-film batteries) have significant differences in spectral response functions. Moreover, the temperature coefficient dispersion between mixed batches of photovoltaic modules is greater than that between photovoltaic modules in the same batch, resulting in low photovoltaic power prediction accuracy.

[0050] Example 1: As Figure 1 As shown in the figure, this embodiment provides a photovoltaic power prediction method, the method including:

[0051] S1. Real-time acquisition of spectral response function data of different photovoltaic modules in a photovoltaic power station. These different photovoltaic modules include at least two module types with spectral response characteristics differing by a preset difference threshold. Photovoltaic power stations typically use a mix of different types of photovoltaic modules, including monocrystalline silicon, polycrystalline silicon, amorphous silicon, CIGS thin-film, and other types. Different photovoltaic modules exhibit significant differences in their spectral response characteristics. The "preset difference threshold" is defined as a difference in the average quantum efficiency exceeding 5% or a difference in peak response wavelength exceeding 30nm in the main operating wavelength band (300-1100nm). For example, monocrystalline silicon modules typically achieve a quantum efficiency of 80-90% in the near-infrared region (780-1100nm), while some thin-film modules may only achieve 50-60% in this region. The peak response wavelength of monocrystalline silicon is approximately 950nm, while the peak response wavelength of some thin-film modules may be in the 500-600nm range. This difference in spectral response characteristics leads to significant differences in the power generation efficiency of different modules under the same irradiation conditions.

[0052] S2, measuring the spectral irradiance distribution incident to the surface of each photovoltaic module in real time through a multi-spectral sensor network deployed above the photovoltaic array; calculating the effective spectral irradiance of each module type according to the spectral response function data and the spectral irradiance distribution, and grouping according to the module type; deploying one sensor per 2-4 hectares of power station area in the multi-spectral sensor network, forming a triangular network structure between sensors to achieve optimal coverage. Each sensor can simultaneously monitor the spectral irradiance of 12-20 wavebands, and the waveband division is determined according to the key points of the characteristic curve of the photovoltaic module, covering the main working waveband of 300-1100nm. For example, in a typical 10MW photovoltaic power station, about 15-20 multi-spectral sensors can be deployed, distributed and installed in different areas of the power station to ensure comprehensive monitoring of the spectral distribution in the power station. The sensor acquisition frequency is 5-15 minutes per time when the weather is stable, and can be increased to 1-5 minutes per time under rapidly changing weather conditions.

[0053] S3, calculating the temperature coefficient of each photovoltaic module according to historical data, and establishing a temperature-power correction factor at the module level; according to the topology relationship of the string, superimposing the temperature-power correction factors of all photovoltaic modules in the same string to obtain a temperature correction power at the string level.

[0054] S4, inputting the temperature correction power at the string level and the effective spectral irradiance into a pre-trained photovoltaic power prediction model, and combining with the global irradiance data of numerical weather prediction to dynamically output a corrected photovoltaic power prediction value at the power station level.

[0055] The method of calculating the effective spectral irradiance of each module type according to the spectral response function data and the spectral irradiance distribution comprises:

[0056] Discretizing the spectral irradiance distribution according to a preset wavelength interval to obtain corresponding irradiance values and wavelength ranges in each wavelength interval; discretization divides the continuous spectral distribution into multiple discrete wavelength intervals, facilitating calculation and data processing. The division of the preset wavelength interval takes into account the spectral response characteristics of the photovoltaic module, and divides the main working waveband of 300-1100nm into 20 wavelength intervals, of which a finer division (about 10-15nm per interval) is used in the region with larger response change (such as the response peak region of monocrystalline silicon at 800-1000nm), and a coarser division (about 30-50nm per interval) is used in the region with smaller response change.

[0057] The spectral response contribution value of each wavelength interval is calculated according to the response value of each wavelength interval in the spectral response function data of the component type and the measured irradiance value of the corresponding wavelength interval, and the spectral response contribution values of all wavelength intervals are weighted and summed to obtain the initial effective spectral irradiance of the component type. For each wavelength interval, the spectral response contribution value of each wavelength interval is obtained by multiplying the response value of the wavelength interval in the spectral response function data of the photovoltaic component type with the measured irradiance value of the corresponding wavelength interval. For example, if the average spectral response value of a monocrystalline silicon component in the 950-965 nm wavelength interval is 0.72 A / W, and the measured average irradiance in this wavelength interval is 0.65 W / m 2 , then the spectral response contribution value of the monocrystalline silicon component in this wavelength interval is 0.468 A / m 2 . The spectral response contribution values of all wavelength intervals are weighted and summed according to the energy distribution proportion of each wavelength interval in the entire working spectrum to obtain the initial effective spectral irradiance of the component type.

[0058] The initial effective spectral irradiance is normalized to obtain the effective spectral irradiance according to the standard solar spectrum distribution data under standard test conditions. The initial effective spectral irradiance is normalized to obtain the final effective spectral irradiance according to the standard solar spectrum distribution data under standard test conditions (such as AM1.5G standard spectrum). The normalization formula is: effective spectral irradiance = initial effective spectral irradiance x (1000 / initial effective spectral irradiance under standard conditions), where 1000 represents the irradiance value (W / m 2 ) under standard test conditions. Normalization ensures that the effective spectral irradiance measured under different weather conditions and at different times can be directly compared with the performance under standard test conditions, providing a unified benchmark for power prediction.

[0059] The method for calculating the temperature coefficient of each batch of photovoltaic components according to historical data and establishing a component-level temperature-power correction factor includes:

[0060] According to the unique identification code of the photovoltaic module, the production batch information is associated, the temperature coefficient dataset of each production batch photovoltaic module in the historical database is extracted, and the temperature coefficient mean and dispersion parameter of each production batch photovoltaic module are counted; the unique identification code of the photovoltaic module is usually printed in the form of serial number or two-dimensional code on the back plate or frame of the module, and the power station management system records the correspondence between the module ID and the installation position through the module installation registration process. The historical database stores the temperature coefficient test value of each production batch photovoltaic module under different running time, including the initial test value and the periodic detection value. The temperature coefficient mean is calculated by the weighted average method, and the weight of the recent test value is higher than that of the early test value, so as to reflect the change trend of the photovoltaic module performance with time. The dispersion parameter is calculated by the weighted standard deviation, which reflects the dispersion degree of the temperature coefficient of the photovoltaic module in the same batch.

[0061] The back plate temperature data of the photovoltaic module is collected in real time, and the temperature-power correction factor of the photovoltaic module is dynamically calculated according to the temperature coefficient mean and the dispersion parameter of the production batch to which it belongs; wherein the temperature-power correction factor of the production batch photovoltaic module whose dispersion exceeds the preset dispersion threshold introduces a dispersion compensation weight, and the dispersion compensation weight is negatively correlated with the dispersion parameter; the back plate temperature data collection uses a distributed temperature sensor network, the sensor is installed at the center position of the module back plate, the precision is ±0.5℃, and the sampling frequency is synchronized with the irradiance data collection (5-15 minutes / time). The output power of the photovoltaic module is negatively correlated with the temperature, and the temperature coefficient represents the power attenuation ratio when the temperature rises by 1℃. The basic calculation formula of the temperature-power correction factor of the module is: correction factor = 1 + temperature coefficient mean × (measured back plate temperature-25℃), wherein 25℃ is the temperature of the standard test condition. For the production batch photovoltaic module whose dispersion exceeds the preset dispersion threshold, the temperature-power correction factor needs to introduce the dispersion compensation weight. The preset dispersion threshold is usually set to 0.05% / ℃ of the standard deviation of the temperature coefficient, and the dispersion compensation weight calculation formula is: dispersion compensation weight = 1-(dispersion parameter-preset dispersion threshold) × k, wherein k is an empirical adjustment coefficient. When the dispersion parameter approaches the preset dispersion threshold, the compensation weight approaches 1, and the influence on the correction factor is small; when the dispersion parameter significantly exceeds the preset dispersion threshold, the compensation weight decreases, reducing the dependence on the estimation of the temperature coefficient of the production batch photovoltaic module.

[0062] If the production batch of the photovoltaic module is mixed and the difference in temperature coefficient exceeds the preset temperature coefficient difference threshold, the temperature-power correction factor weight is adjusted according to the electrical connection relationship of the photovoltaic modules of different production batches in the string. If the production batch of the photovoltaic module is mixed and the difference in temperature coefficient exceeds the preset temperature coefficient difference threshold, which is usually set to 0.08% / ℃, the temperature-power correction factor weight is adjusted according to the electrical connection relationship of the photovoltaic modules of different production batches in the string. For the photovoltaic modules of different production batches connected in series, the minimum value method is used to determine the string current limiting factor. For the photovoltaic modules of different production batches connected in parallel, the capacity proportion of each production batch in parallel is weighted and averaged. It is ensured that in the case of mixed use of photovoltaic modules of different production batches, the temperature-power correction can accurately reflect the limiting effect of the actual electrical connection on the power output.

[0063] When it is detected that the temperature coefficient dispersion of the photovoltaic module of the production batch in the string continuously exceeds the historical statistical range, a stepwise temperature control excitation signal is applied to the photovoltaic module of the production batch, and the transient change slope of the temperature response curve is collected to inverse the actual temperature coefficient, update the mean value and dispersion of the temperature coefficient of the photovoltaic module of the production batch, and regenerate the component-level temperature-power correction factor. By analyzing the corresponding relationship between temperature change and power output, the current actual temperature coefficient is calculated, the temperature coefficient database of the production batch is updated, and a more accurate component-level temperature-power correction factor is regenerated.

[0064] The method for detecting that the temperature coefficient dispersion of the photovoltaic module of the production batch in the string continuously exceeds the historical statistical range, then applying a stepwise temperature control excitation signal to the photovoltaic module of the production batch, and collecting the transient change slope of the temperature response curve to inverse the actual temperature coefficient comprises:

[0065] A stepwise temperature control excitation signal is applied through a temperature control actuator integrated in the backsheet of the photovoltaic module, and the transient change data of the backsheet temperature of the photovoltaic module is synchronously collected at a preset sampling frequency. The temperature control actuator can be a thin film electrothermal element attached to the backsheet of the module, which can provide precisely controlled heating power. The stepwise temperature control excitation signal is designed as multiple consecutive temperature steps, with an initial step amplitude of 2-3℃, and subsequent steps adjusted according to the previous response, with a maximum of 5℃ to avoid thermal shock. Each step lasts for 5-10 minutes to ensure that the module temperature reaches a steady state. The transient change data of the backsheet temperature of the photovoltaic module is synchronously collected at a preset sampling frequency. The sampling frequency is set to 1-5Hz, which is sufficient to capture the dynamic process of temperature change. At the same time, the change data of the electrical parameters (open-circuit voltage, short-circuit current, maximum power point voltage and current) of the module are recorded to establish the corresponding relationship between temperature change and power change.

[0066] The transient change data is filtered and denoised, the response curve of each temperature step interval is extracted, and the instantaneous change slope and steady-state offset are calculated, and a temperature-time transfer function is constructed in combination with a dynamic thermal capacity model of the photovoltaic module; the filtering and denoising adopts wavelet transform or a moving average filter to remove measurement noise and environmental interference.

[0067] The actual temperature coefficient is inversely calculated by least square method iteration according to matching of the temperature-time transfer function and a preset standard temperature coefficient response database, and a confidence interval is calculated; if the half width of the confidence interval is less than a preset precision threshold, the actual temperature coefficient is output. The standard temperature coefficient response database contains typical temperature response characteristic parameters of different types and different aging degrees of components, and the most matched temperature coefficient value is found out by comparing the difference between the measured response and the standard response in combination with the least square method. Meanwhile, the confidence interval of the temperature coefficient is calculated to reflect the reliability of the estimation result. If the half width of the confidence interval is less than a preset precision threshold, for example, 0.01% / ℃, it is considered that the estimation result is reliable, and the actual temperature coefficient is output for updating the database; otherwise, the test time is increased or the test parameters are adjusted for re-measurement.

[0068] The method of filtering and denoising the transient change data, extracting the response curve of each temperature step interval, calculating the instantaneous change slope and steady-state offset, and constructing a temperature-time transfer function in combination with a dynamic thermal capacity model of the photovoltaic module comprises:

[0069] The starting point and the ending point of the temperature step interval are identified by a dynamic threshold segmentation algorithm, the response curve is obtained by cubic spline interpolation fitting in each temperature step interval segmented, the instantaneous change slope of each sampling point on the response curve is calculated, and the maximum instantaneous change slope is extracted as the step response characteristic value; the dynamic threshold segmentation algorithm automatically detects the boundary of the step signal according to the statistical characteristics of the temperature change rate. When the temperature change rate exceeds the threshold, it is marked as the starting point of the step, and after a certain time (usually 3-5 sampling periods) below the threshold, it is marked as the ending point. The response curve is obtained by cubic spline interpolation fitting in each temperature step interval segmented. The cubic spline interpolation ensures the continuity and smoothness of the curve, and reduces the influence of noise on subsequent analysis. The instantaneous change slope of each sampling point, that is, the ratio of the temperature change rate (dT / dt) to the power change rate (dP / dt), is calculated on the fitted curve. The maximum instantaneous change slope is extracted from all sampling points as the step response characteristic value, which reflects the maximum sensitivity of the photovoltaic module to temperature change.

[0070] The dynamic thermal capacity model is established according to material parameters and structural parameters of the photovoltaic module; the step response characteristic value and the steady-state offset are input into the dynamic thermal capacity model to obtain a parameter matrix of a temperature-time transfer function, and the parameter matrix is calibrated in real time through a Kalman filtering algorithm; if a calibration residual of the parameter matrix exceeds a preset residual tolerance threshold, adaptive updating of the dynamic thermal capacity model is triggered. The dynamic thermal capacity model is in the form of a thermoelectric equivalent network, contains multiple thermal resistance and thermal capacity elements, and simulates thermal conduction characteristics of each layer (glass, EVA, cell sheet, back plate, etc.) of the photovoltaic module. The step response characteristic value and the steady-state offset are input into the dynamic thermal capacity model to obtain a parameter matrix of a temperature-time transfer function. The parameter matrix contains all coefficients describing dynamic characteristics of the system, such as time constant, gain coefficient, etc. A second-order Kalman filtering algorithm is used to calibrate the parameter matrix in real time. An initial process noise covariance matrix is determined according to a historical temperature fluctuation standard deviation, a measurement noise covariance matrix is set according to accuracy of a temperature sensor, and a state transition matrix is determined according to a thermal time constant (usually 5-15 minutes) of the module. The Kalman filtering can effectively process measurement noise and model uncertainty, and continuously optimize parameter estimation. The adaptive updating process reevaluates the structure and parameters of the dynamic thermal capacity model, such as increasing the order of the dynamic thermal capacity model or adjusting the parameter range, to ensure that the current thermal dynamic characteristics of the photovoltaic module can be accurately described.

[0071] The method for establishing the dynamic thermal capacity model according to the material parameters and the structural parameters of the photovoltaic module comprises:

[0072] A three-dimensional non-uniform grid model is constructed according to material parameters and structural parameters of the photovoltaic module; the material parameters are physical characteristic parameters of each constituent material of the photovoltaic module, and the structural parameters are geometric characteristic and physical layout parameters of the photovoltaic module; the material parameters include physical characteristics such as thermal conductivity, specific heat capacity and density of each layer of material, and the structural parameters include geometric characteristics such as thickness and area of each layer and internal electrical connection mode. For example, the three-dimensional non-uniform grid model is divided into 10x10 to 20x20 units in the plane direction of the module, and is divided into 4-6 layers in the thickness direction, corresponding to the glass layer, the EVA layer, the cell sheet layer and the back plate layer. The grid density of each layer is inversely proportional to the thermal conductivity of the material, and the grid density of the material layer with lower thermal conductivity is higher, so as to improve the temperature gradient calculation accuracy.

[0073] According to the historical aging data of the photovoltaic module, the degradation coefficients of the material parameters are corrected, the degradation coefficients including the annual growth rate of the thermal resistance of the packaging material and the decay curve of the thermal capacity of the cell sheet, and an initial thermal resistance-thermal capacity correlation matrix is generated according to the corrected material parameters; the degradation coefficients of the material parameters are corrected according to the historical aging data of the photovoltaic module. The degradation coefficients mainly include the annual growth rate of the thermal resistance of the packaging material and the decay curve of the thermal capacity of the cell sheet. The thermal resistance of the packaging material (such as EVA) increases with time, and according to the industry experience data, the annual growth rate is about 2-4%, and the total growth after 5 years of use is about 15-25%; the thermal capacity of the cell sheet decays with aging, and the decay rate within 5 years is about 5-10%. The degradation coefficients are obtained by analyzing the temperature response characteristics of the module under different running time lengths, and are updated regularly. An initial thermal resistance-thermal capacity correlation matrix is generated according to the corrected material parameters, which is used to describe the heat transfer path and efficiency between each part of the module.

[0074] The initial thermal resistance-thermal capacity correlation matrix is combined with the three-dimensional non-uniform grid model and the real-time collected backboard temperature gradient distribution data to obtain dynamic thermal capacity parameters through iterative optimization; and a dynamic thermal capacity model is generated according to the optimized dynamic thermal capacity parameters. The iterative optimization adopts a gradient descent algorithm, and the optimization target is to minimize the mean square error between the predicted temperature and the measured temperature of the model. The optimization process usually requires 10-20 iterations, and the convergence criterion is that the parameter change of two consecutive iterations is less than 1%. The optimized dynamic thermal capacity parameters can accurately reflect the current thermal dynamic characteristics of the photovoltaic module.

[0075] The method of generating a dynamic thermal capacity model according to the optimized dynamic thermal capacity parameters comprises:

[0076] The optimized dynamic thermal capacity parameters are mapped to the three-dimensional non-uniform grid model according to the grid nodes, the thermal resistance and thermal capacity values of each grid node are extracted and converted into equivalent lumped parameters, and a dynamic thermal capacity model is constructed according to the equivalent lumped parameters; the optimized dynamic thermal capacity parameters are mapped to the three-dimensional non-uniform grid model according to the grid nodes, and the thermal resistance and thermal capacity values of each grid node are extracted. The mapping process maintains the spatial distribution information, ensuring that the model can accurately reflect the temperature gradient at different positions of the module. The extracted distributed thermal resistance and thermal capacity values are converted into equivalent lumped parameters through the principle of electrical-thermal equivalent circuit, which simplifies the calculation complexity while maintaining the dynamic characteristics. The equivalent lumped parameters include thermal inertia factor (reflecting temperature change rate) and order and coefficient of temperature-time transfer function. The thermal inertia factor is usually expressed in time constant, which reflects the time required for the module to reach temperature steady state; the order of the transfer function determines the complexity of the system response, which is usually 2-3 orders; the coefficient determines the weight of each order term, reflecting the influence degree of different time scale thermal processes.

[0077] The dynamic heat capacity model is verified in frequency domain and time domain. In the frequency domain, the amplitude-frequency characteristics and phase characteristics of the transfer function are verified by the sweep excitation test. In the time domain, the root mean square value of the residual error between the predicted temperature and the actual temperature is verified by the step temperature control experiment. If the verification results of the frequency domain and the time domain meet the preset verification accuracy, the dynamic heat capacity model is bound with the unique identification code of the photovoltaic module. The sweep excitation test is carried out in the frequency range of 0.001-0.1 Hz, and the influence amplitude and phase difference of different frequency temperature changes on power output are measured. The verification standard requires that the relative error of the model amplitude-frequency characteristics and the measured data in the test frequency range is not more than 3%, and the phase error is not more than 10°. The step temperature control experiment applies a temperature step of 2-5℃ for 30-60 minutes, and records the temperature and power change process. The verification standard requires that the root mean square error of temperature prediction is less than 0.5℃, and the relative root mean square error of power prediction is less than 2%. If the verification results of the frequency domain and the time domain meet the preset verification accuracy, the dynamic heat capacity model is bound with the unique identification code of the photovoltaic module, and is saved in the power station database as a special model of the module.

[0078] The method for inputting the string-level temperature corrected power and effective spectral irradiance into the pre-trained photovoltaic power prediction model, and combining the global irradiance data of numerical weather prediction to dynamically output the corrected power station-level photovoltaic power prediction value comprises:

[0079] The temperature-corrected power of each group string is summarized as a power station level temperature-corrected power matrix according to the group string topology and the power station electrical connection diagram. The power station level temperature-corrected power matrix, effective spectral irradiance and global irradiance data from numerical weather prediction are time series aligned and fused into a multi-dimensional input feature vector. The multi-dimensional input feature vector is input into a pre-trained photovoltaic power prediction model, which is trained by a hybrid neural network. The group strings and inverters in the photovoltaic power station form a complex topology network. The actual electrical connection relationship needs to be considered when integrating the group string level temperature-corrected power into the power station level power. The power station electrical connection diagram describes the parallel relationship between group strings, the configuration and connection mode of grid-connected inverters. The influence of the power change of any group string on the total output of the power station can be calculated through the power station electrical connection diagram. A hierarchical aggregation method is used in the summary process. The first layer is the component layer, which processes each photovoltaic component through the temperature-power correction factor. The second layer is the group string layer, which considers the series characteristics of the components within the group string and uses the lowest temperature-corrected power within the group string as the limiting factor. The third layer is the parallel group layer, which weights and sums the parallel connected group strings according to their respective rated capacities. The last layer is the inverter layer, which considers the efficiency curve of the inverter. The power station level temperature-corrected power matrix is a multi-dimensional data structure, including the time dimension (prediction time point) and the spatial dimension (group string location and topology relationship). Time series alignment is a key step to ensure the time consistency of multi-source data. For example, the time granularity of the power station level temperature-corrected power matrix is 15 minutes, the numerical weather prediction data is usually 1 hour granularity, and the spectral irradiance is 5-15 minute granularity. A unified time reference converts data of different granularities into a unified 15-minute interval. For data with coarse granularity, a cubic spline interpolation method is used for time refinement. For data with finer granularity, a time window average method is used for aggregation. The data fusion process uses feature engineering methods to construct a multi-dimensional input feature vector. The feature vector includes: power prediction values in the power station level temperature-corrected power matrix (main feature); effective spectral irradiance of each waveband and its time gradient (reflecting the trend of spectral change); global irradiance, cloud cover, wind speed, air temperature and other meteorological parameters from numerical weather prediction. All features are standardized to eliminate dimensional differences. The pre-trained photovoltaic power prediction model is trained by a hybrid neural network, which includes a time convolution network (TCN) and a long short-term memory network (LSTM). The time convolution network (TCN) is responsible for capturing short-term time patterns and processing rapid changes in irradiance and temperature. The long short-term memory network (LSTM) captures long-term dependencies and is used for periodic patterns. The photovoltaic power prediction model outputs the power station level photovoltaic power prediction values at different time points in the future and their 95% confidence intervals.

[0080] If an extreme weather event warning is detected, a multi-scale time window prediction is triggered to obtain the final power station level photovoltaic power prediction value; the extreme weather event includes weather conditions such as heavy rain, thunderstorm, sandstorm and strong convection that may cause the photovoltaic output to fluctuate dramatically. The extreme weather event warning is derived from real-time warning information released by the meteorological department and is divided into four levels (I-IV levels). The multi-scale time window prediction combines the advantages of short-term prediction and medium and long-term prediction to improve the prediction accuracy under extreme conditions.

[0081] The final power station level photovoltaic power prediction value is compared with the real-time power monitoring data to obtain a power prediction error value. If the power prediction error of N consecutive time windows exceeds the preset power error threshold, the pre-trained photovoltaic power prediction model is adaptively adjusted, and N≥3. The real-time power monitoring data comes from the power station SCADA system and records the real-time output power of each inverter and busbar of the power station. The power prediction error is calculated in two ways, i.e., relative error and absolute error. If the power prediction error of N consecutive time windows exceeds the preset power error threshold, such as the relative error exceeding ±5% or the absolute error exceeding 100 kW, the pre-trained photovoltaic power prediction model is triggered to adaptively adjust. The adaptive adjustment adopts an online incremental learning method, and the measured data of the last 30 days is used as the training set. The incremental update is performed every 7 days, the update learning rate is 10-30% of the basic learning rate, and a forgetting factor (0.9-0.95) is introduced to reduce the weight influence of the long-term data.

[0082] The method comprises the following steps:

[0083] The multi-scale time window prediction includes short-term prediction and medium and long-term prediction. The short-term prediction generates high-frequency prediction results in a minute-level time window. The medium and long-term prediction generates trend prediction results in a hour-level time window. According to the extreme weather event warning level and the weather event duration, the prediction results of the short-term prediction and the medium and long-term prediction are dynamically weighted and fused to obtain the final power station level photovoltaic power prediction value. The short-term prediction generates high-frequency prediction results in a minute-level time window (5-15 minutes), mainly based on real-time monitoring data and short-term weather change trend, and is suitable for capturing the rapid fluctuation of photovoltaic output. The medium and long-term prediction generates trend prediction results in a hour-level time window (1-24 hours), mainly based on numerical weather prediction and historical similar day mode, and is suitable for grasping the overall trend of photovoltaic output. The weighted fusion adopts a time-varying weight method, the short-term prediction weight is higher at a recent time point, and the medium and long-term prediction weight is higher at a long-term time point, so as to realize smooth transition from short-term accurate prediction to long-term trend prediction.

[0084] Finally, it should be noted that although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still make modifications to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to part of the technical features, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A photovoltaic power prediction method, characterized in that, The method comprises: Real-time acquisition of spectral response function data of different photovoltaic components in a photovoltaic power station, wherein the different photovoltaic components include at least two component types with a spectral response characteristic difference greater than a preset difference threshold; Real-time measurement of spectral irradiance distribution incident to the surface of each photovoltaic component by a multispectral sensor network deployed above the photovoltaic array; calculation of effective spectral irradiance of each component type according to the spectral response function data and the spectral irradiance distribution; According to historical data, the temperature coefficient of each photovoltaic component belonging to the batch is counted, and a component-level temperature-power correction factor is established; according to the string topology relationship, the temperature-power correction factors of all photovoltaic components in the same string are superimposed to obtain a string-level temperature correction power; The string-level temperature correction power and effective spectral irradiance are input into a pre-trained photovoltaic power prediction model, and combined with global irradiance data of numerical weather prediction, a corrected power station-level photovoltaic power prediction value is dynamically output; The method for calculating the effective spectral irradiance of each component type according to the spectral response function data and the spectral irradiance distribution comprises: Discretize the spectral irradiance distribution according to the preset wavelength interval to obtain the corresponding irradiance value and wavelength range in each wavelength interval; According to the response value of each wavelength interval in the spectral response function data of the component type and the measured irradiance value of the corresponding wavelength interval, the spectral response contribution value of each wavelength interval is calculated, and the weighted sum of the spectral response contribution values of all wavelength intervals is obtained. The initial effective spectral irradiance of the component type is obtained. According to the standard solar spectrum distribution data under the standard test conditions, the initial effective spectral irradiance is normalized to obtain the effective spectral irradiance.

2. The photovoltaic power prediction method of claim 1, wherein, The method for counting the temperature coefficient of each photovoltaic component belonging to the batch according to historical data, and establishing a component-level temperature-power correction factor comprises: According to the unique identification code of the photovoltaic component, the production batch information is associated, the temperature coefficient dataset of each production batch photovoltaic component in the historical database is extracted, and the temperature coefficient mean value and dispersion parameter of each production batch photovoltaic component are counted; Real-time acquisition of photovoltaic component backboard temperature data, and dynamic calculation of photovoltaic component-level temperature-power correction factor according to the temperature coefficient mean value and dispersion parameter of its belonging production batch; wherein the temperature-power correction factor of the production batch photovoltaic component with dispersion exceeding the preset dispersion threshold is introduced into the dispersion compensation weight, and the dispersion compensation weight is negatively correlated with the dispersion parameter; If the production batch to which the photovoltaic component belongs is mixed and the temperature coefficient difference exceeds the preset temperature coefficient difference threshold, then adjust the temperature-power correction factor weight according to the electrical connection relationship of different production batch photovoltaic components in the string. When it is detected that the temperature coefficient dispersion of the production batch photovoltaic module within the string continuously exceeds the historical statistical range, a step temperature control excitation signal is applied to the production batch photovoltaic module, and the transient change slope of the temperature response curve is collected to inversely calculate the actual temperature coefficient, update the temperature coefficient mean and dispersion parameters of the production batch photovoltaic module, and regenerate the component-level temperature-power correction factor.

3. A photovoltaic power prediction method according to claim 2, characterized in that, The method for inversely calculating the actual temperature coefficient by applying the step temperature control excitation signal to the production batch photovoltaic module and collecting the transient change slope of the temperature response curve when it is detected that the temperature coefficient dispersion of the production batch photovoltaic module within the string continuously exceeds the historical statistical range comprises: applying the step temperature control excitation signal through the temperature control actuator integrated in the backsheet of the photovoltaic module, and synchronously collecting the transient change data of the backsheet temperature of the photovoltaic module according to a preset sampling frequency; performing filtering and noise reduction processing on the transient change data, extracting the response curve of each temperature step interval, calculating the instantaneous change slope and steady-state offset, and constructing a temperature-time transfer function in combination with the dynamic thermal capacity model of the photovoltaic module; matching the temperature-time transfer function with a preset standard temperature coefficient response database, inversely calculating the actual temperature coefficient and calculating the confidence interval through least square method iteration optimization, and outputting the actual temperature coefficient if the half-width of the confidence interval is less than a preset precision threshold.

4. The photovoltaic power prediction method of claim 3, wherein, The method for performing filtering and noise reduction processing on the transient change data, extracting the response curve of each temperature step interval, calculating the instantaneous change slope and steady-state offset, and constructing a temperature-time transfer function in combination with the dynamic thermal capacity model of the photovoltaic module comprises: identifying the starting point and ending point of the temperature step interval through a dynamic threshold segmentation algorithm, fitting the response curve in each temperature step interval after segmentation through cubic spline interpolation, calculating the instantaneous change slope of each sampling point on the response curve, and extracting the maximum instantaneous change slope as the step response characteristic value; establishing a dynamic thermal capacity model according to the material parameters and structural parameters of the photovoltaic module, inputting the step response characteristic value and the steady-state offset into the dynamic thermal capacity model to obtain the parameter matrix of the temperature-time transfer function, and performing real-time calibration on the parameter matrix through Kalman filtering algorithm; if the calibration residual of the parameter matrix exceeds a preset residual tolerance threshold, triggering adaptive update of the dynamic thermal capacity model.

5. A photovoltaic power prediction method according to claim 4, characterized in that, The method for establishing a dynamic thermal capacity model according to the material parameters and structural parameters of the photovoltaic module comprises: constructing a three-dimensional non-uniform grid model according to the material parameters and structural parameters of the photovoltaic module; the material parameters are the physical property parameters of each constituent material of the photovoltaic module, and the structural parameters are the geometric features and physical layout parameters of the photovoltaic module; correcting the degradation coefficient of the material parameters according to the historical aging data of the photovoltaic module, the degradation coefficient including the thermal resistance annual growth rate of the packaging material and the decay curve of the cell thermal capacity, and generating an initial thermal resistance-thermal capacity correlation matrix according to the corrected material parameters; The initial thermal resistance-thermal capacity correlation matrix is combined with the three-dimensional non-uniform grid model and the real-time collected temperature gradient distribution data of the back plate to obtain dynamic thermal capacity parameters through iterative optimization; and a dynamic thermal capacity model is generated according to the optimized dynamic thermal capacity parameters.

6. A photovoltaic power prediction method according to claim 5, characterized in that, The method for generating the dynamic thermal capacity model according to the optimized dynamic thermal capacity parameters comprises: The optimized dynamic thermal capacity parameters are mapped to the three-dimensional non-uniform grid model according to the grid nodes, the thermal resistance and thermal capacity values of each grid node are extracted and converted into equivalent lumped parameters, and a dynamic thermal capacity model is constructed according to the equivalent lumped parameters; The dynamic thermal capacity model is verified in both the frequency domain and the time domain, the amplitude-frequency characteristics and phase characteristics of the transfer function are verified in the frequency domain through sweep excitation testing, and the root mean square value of the residual error between the predicted temperature and the actual temperature is verified in the time domain through step temperature control experiments; if the verification results of the verification in both the frequency domain and the time domain meet the preset verification accuracy, the dynamic thermal capacity model is bound with the unique identification code of the photovoltaic module.

7. The photovoltaic power prediction method of claim 1, wherein, The method for inputting the string-level temperature correction power and the effective spectral irradiance into the pre-trained photovoltaic power prediction model and dynamically outputting the corrected power station-level photovoltaic power prediction value in combination with the global irradiance data of the numerical weather forecast comprises: The string-level temperature correction power of each group is summarized into a power station-level temperature correction power matrix according to the group string topological relationship and the power station electrical connection diagram, the power station-level temperature correction power matrix, the effective spectral irradiance and the global irradiance data of the numerical weather forecast are processed in time sequence alignment and then fused into a multi-dimensional input feature vector, and the multi-dimensional input feature vector is input into the pre-trained photovoltaic power prediction model, which is obtained through mixed neural network training; If an extreme weather event warning is detected, a multi-scale time window prediction is triggered to obtain a final power station-level photovoltaic power prediction value; The final power station-level photovoltaic power prediction value is compared with the real-time power monitoring data to obtain a power prediction error value, and if the power prediction errors of continuous N time windows exceed a preset power error threshold, the pre-trained photovoltaic power prediction model is adaptively adjusted, N≥3.

8. The photovoltaic power prediction method of claim 7, wherein, The method for triggering a multi-scale time window prediction to obtain a final power station-level photovoltaic power prediction value if an extreme weather event warning is detected comprises: The multi-scale time window prediction comprises short-term prediction and medium and long-term prediction; the short-term prediction generates a high-frequency prediction result in a minute-level time window; the medium and long-term prediction generates a trend prediction result in a hour-level time window; according to the warning level of the extreme weather event and the duration of the weather event, the prediction results of the short-term prediction and the medium and long-term prediction are dynamically weighted and fused to obtain a final power station-level photovoltaic power prediction value.

Images