A distributed photovoltaic irradiation data checking and correcting method

Abstract

The present application relates to the technical field of data verification, especially relates to a kind of distributed photovoltaic irradiation data verification and correction method.The method first obtains grid-connected power station operating data, and the actual irradiance of each point is deduced based on photovoltaic module physical model;Subsequently, the reference point is screened from multiple dimensions, and the total deviation of measured and original simulated irradiance is decoupled into regional system deviation factor, local shading deviation factor and installation equipment deviation factor.On this basis, combined with historical meteorological statistics experience, the prior distribution of candidate point correction factor is established, the posterior mean and variance of each point correction factor are calculated by using Bayes formula to fuse measured likelihood information.Finally, the original data is calibrated using the expected value of correction factor, and the confidence interval under the preset confidence level is output.

Description

Technical Field

[0001] This invention relates to the field of data verification technology, and in particular to a method for verifying and correcting distributed photovoltaic irradiation data. Background Technology

[0002] Distributed photovoltaic (PV) power stations are typically installed on rooftops of urban buildings, in industrial parks, or in areas with complex terrain. Accurate assessment of their radiation resources is crucial for power station site selection, power generation forecasting, and financing evaluation. Currently, the industry primarily relies on spatial interpolation from ground-based meteorological stations or satellite inversion models (such as PVGIS and Solargis) to obtain PV radiation data. However, ground-based meteorological stations are sparsely distributed and mostly located in open suburban areas, making it difficult to capture radiation differences under urban microclimates. While satellite inversion models possess global coverage capabilities, they cannot effectively identify microscale shading losses caused by tall buildings, vegetation, and terrain at kilometer-level spatial resolution.

[0003] While there are attempts to verify irradiance by measuring the actual power of grid-connected power plants in existing technologies, the restoration process is subject to significant uncertainties due to the coupling interference of multiple non-radiative factors, such as inverter efficiency fluctuations, component temperature losses, hardware aging and dust accumulation, and installation angle deviations, in the energy conversion chain of photovoltaic systems. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a method for verifying and correcting distributed photovoltaic irradiance data, aiming to improve the pain points of low accuracy of distributed photovoltaic radiation data, difficulty in identifying shading, and inability to quantify risks in complex urban micro-environments.

[0005] This invention provides the following technical solution: a method for verifying and correcting distributed photovoltaic irradiance data, comprising: Obtain the operation data and engineering parameters of grid-connected photovoltaic power stations within the target area, and infer the actual irradiance of each power station location based on the physical model of photovoltaic modules; Based on the data quality evaluation indicators, verification benchmark points were selected from each power station location; The total deviation between the actual irradiance reached at the verification benchmark point and the original simulated irradiance at the corresponding location is calculated, and the total deviation is decoupled into a multi-dimensional deviation factor set through clear sky window filtering and three-dimensional occlusion analysis. Based on historical statistical experience, a prior distribution of the correction factor for each candidate point is established. A likelihood function is constructed using the multidimensional deviation factor set. The posterior distribution of the correction factor for each candidate point is calculated in combination with the prior distribution. Based on the posterior distribution, the expected value of the correction factor for each candidate point and its uncertainty interval are determined. The original simulated irradiation data corresponding to the candidate sites are corrected using the expected value of the correction factor to generate irradiation resource distribution data and output the corresponding uncertainty quantification results.

[0006] Preferably, the step of back-calculating the actual irradiance at each power station location based on the photovoltaic module physical model includes: Obtain the nominal power and temperature coefficient of the component under standard test conditions; Obtain the real-time output power of each power station and, in conjunction with the inverter conversion efficiency, reverse-calculate the DC-side power. The DC-side power is compensated for environmental degradation using the temperature coefficient and component performance degradation rate to obtain the received light power; The received light power is inverted into the actual effective irradiance on the component surface based on the single diode equivalent circuit model.

[0007] Preferably, the step of selecting verification reference points from the power station locations includes: Each power station location was screened based on four dimensions: data integrity, runtime, equipment stability, and environmental stability, eliminating locations that did not meet the preset conditions. Calculate the comprehensive data quality score for the remaining power station locations, and determine the locations with scores higher than the preset threshold as the verification benchmark points.

[0008] Preferably, the step of decoupling the total deviation into a multi-dimensional deviation factor set includes: Calculate the total deviation between the actual irradiance reached at the verification reference point and the original simulated irradiance at the corresponding location; Filter time windows under clear sky conditions, and separate regional system deviation factors related to meteorological models based on the total deviation within clear sky periods; Based on the three-dimensional spatial geometric parameters of the verification benchmark point and the digital surface model of the surrounding buildings, the ray tracing algorithm is used to simulate the shadow occlusion throughout the year and calculate the local occlusion deviation factor caused by the occlusion of the surrounding environment. The remaining deviation after removing the regional system deviation factor and the local occlusion deviation factor from the total deviation is included in the installation and equipment deviation factor.

[0009] Preferably, the step of establishing the prior distribution of the correction factor for each candidate site includes: The deviation distribution characteristics between historical meteorological observation data and satellite inversion data of the target area are obtained as global prior constraints. Based on the spatial distance and environmental similarity between each candidate point and each verification benchmark point, a spatial interpolation method is used to determine the spatial co-estimation of the correction factor as a local prior constraint. By combining the global prior constraints and the local prior constraints, the prior mean and prior variance of the correction factors for each candidate point are set to generate a prior distribution model that follows a normal distribution.

[0010] Preferably, the step of calculating the posterior distribution of the correction factor for each candidate point based on the prior distribution includes: Based on the multi-dimensional deviation factor set of the verification benchmark, a likelihood function for the correction factor is constructed. Using the prior distribution as prior information for the correction factor, and combining it with the likelihood function, probability fusion is performed using Bayes' theorem to calculate the posterior distribution of the correction factor for each candidate point. The posterior mean and posterior variance of the posterior distribution are extracted and used as the expected value and uncertainty measure of the correction factor for each candidate point, respectively.

[0011] Preferably, the step of correcting the original simulated irradiation data corresponding to the candidate sites includes: Obtain the original simulated irradiation data and the corresponding expected value of the correction factor for the candidate sites; The expected value of the correction factor is used as a calibration operator to apply to the original simulated irradiation data to obtain the corrected irradiation distribution data. By combining the uncertainty measure of the candidate points and mapping the probability density function, the confidence interval of the corrected irradiance distribution data at a preset confidence level is determined.

[0012] Preferably, the method further includes a model dynamic update step: Monitor the real-time operational data of the verification benchmark point and calculate the rate of change of the multi-dimensional deviation factor set over time; When the rate of change exceeds a preset threshold, the parameter adjustment of the prior distribution is initiated, and the expected value of the correction factor of each candidate point is updated based on the adjusted prior distribution. The irradiation resource distribution data are incrementally corrected using the updated expected value of the correction factor.

[0013] The present invention has the following beneficial effects: 1. This invention establishes a physical reverse model of photovoltaic modules, enabling the accurate reconstruction of electrical performance data at the power station end, affected by various interference factors such as ambient temperature, inverter efficiency, and equipment aging, into the actual irradiance reaching the module surface. This reverse inference from the energy production end to the resource input end provides a measured benchmark for distributed scenarios.

[0014] 2. This invention achieves structured processing of complex error sources by separating the total observation bias into regional system bias factors, local occlusion bias factors, and installation equipment bias factors. Utilizing three-dimensional spatial geometric modeling and ray tracing algorithms, this invention can quantify the instantaneous radiation loss caused by building occlusion in urban micro-scale environments.

[0015] 3. This invention employs a Bayesian inference framework, successfully fusing prior information from regional historical meteorological statistics with real-time decoupled physical observation likelihood evidence. This method not only outputs a double-calibrated expected value of irradiance, but more importantly, it outputs a measure of uncertainty in resource assessment through posterior variance. Attached Figure Description

[0016] Figure 1 This is a flowchart of a distributed photovoltaic irradiation data verification and correction method proposed in this invention. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 In the first embodiment of the present invention, a method for verifying and correcting distributed photovoltaic irradiance data is provided, such as... Figure 1 As shown, it includes the following steps: S1. Obtain the operation data and engineering parameters of grid-connected photovoltaic power stations in the target area, and back-calculate the actual irradiance of each power station location based on the physical model of photovoltaic modules. Preferably, the step of back-calculating the actual irradiance at each power station location based on the photovoltaic module physical model includes: Obtain the nominal power and temperature coefficient of the component under standard test conditions; Obtain the real-time output power of each power station and, in conjunction with the inverter conversion efficiency, reverse-calculate the DC-side power. The DC-side power is compensated for environmental degradation using the temperature coefficient and component performance degradation rate to obtain the received light power; The received light power is inverted into the actual effective irradiance on the component surface based on the single diode equivalent circuit model.

[0019] Specifically, the physical parameters of the photovoltaic modules in each grid-connected photovoltaic power station within the target area are first obtained. Referring to the parameter benchmarks for photovoltaic array design in GB 50797-2012, the physical parameters of the modules under standard test conditions (STC, i.e., irradiance) are obtained. Battery temperature Nominal power under AM1.5 spectrum and power temperature coefficient This benchmark is used to define the upper limit of the photoelectric conversion efficiency of a component under ideal conditions.

[0020] Real-time data collection of photovoltaic power plants at preset time steps Measured power generation on the AC side inside To eliminate the interference of system integration losses on resource assessment, the real-time conversion efficiency of the inverter is introduced. and the overall system loss coefficient (Covering transformer losses, line losses, and unbalanced losses). First, the AC-side power is converted back to DC-side power. : ; Subsequently, combined with the sampling time step Calculate the average real-time power on the DC side during this period. ; This will then convert the accumulated energy at the grid connection end back into the instantaneous power at the module end.

[0021] Given that the semiconductor characteristics of photovoltaic cells are severely affected by temperature, it is necessary to monitor the real-time power output on the DC side. Decoupling process: ; This leads to the light-receiving power driven solely by irradiance intensity. .

[0022] Based on the law of conservation of energy and the photoelectric conversion ratio, the decoupled light-receiving power With the nominal power of the components Comparison was performed, combining the standard irradiance constant. The actual effective irradiance of this location on the component surface during the corresponding time period was derived. ; Here In a physical sense, it is equivalent to the average irradiance during that period.

[0023] S2. Based on the data quality evaluation indicators, the verification benchmark points are selected from each power station location; Preferably, the step of selecting verification reference points from the power station locations includes: Each power station location was screened based on four dimensions: data integrity, runtime, equipment stability, and environmental stability, eliminating locations that did not meet the preset conditions. Calculate the comprehensive data quality score for the remaining power station locations, and determine the locations with scores higher than the preset threshold as the verification benchmark points.

[0024] Specifically, based on the acquired operational data and engineering parameters, each power station location is initially filtered from the following four dimensions: Data integrity verification: This involves statistically analyzing the sampling frequency and data retention rate of power plant operation data within a preset time period. Data points where the proportion of missing data due to unstable communication links, gateway packet loss, or monitoring system malfunctions exceeds a preset threshold are removed.

[0025] Runtime verification: Filter locations whose grid-connected operation time meets the preset duration, such as exceeding one full meteorological year. Remove locations that are in the grid-connected commissioning period, trial operation period, or have insufficient operating sample size to eliminate unsteady performance fluctuations during the equipment break-in period.

[0026] Equipment stability assessment: Monitor the operating efficiency dispersion of the power plant over historical periods. By analyzing the consistency of DC-side power conversion with AC-side power, identify and eliminate points with abnormal inverter efficiency, frequent component tripping, or severe string mismatch, ensuring that the electrical performance output meets the preset deviation range of the single diode physical model.

[0027] Environmental stability identification: By comparing historical power curves with a regional clear-sky background model, changes in the light environment at the monitoring point are identified. Sites with newly added tall buildings blocking the light, invasive vegetation growth, or long-term severe dust accumulation causing sudden performance changes are eliminated to ensure that the benchmark point represents the natural irradiance patterns of the target area.

[0028] Using the remaining locations from the initial selection, a data quality evaluation model is constructed. This model assigns weight coefficients based on the contribution of the four dimensions mentioned above, and calculates the comprehensive score for each location. The calculation formula is logically as follows: ; in, For the weights of each dimension, The minimum-maximum standardized scores for the corresponding dimensions are then calculated. The overall score for each location is then calculated. Compared with the system's preset quality threshold A comparison is performed. Power station locations with scores higher than the threshold are formally designated as verification benchmarks.

[0029] S3. Calculate the total deviation between the actual irradiance reached at the verification benchmark point and the original simulated irradiance at the corresponding location, and decouple the total deviation into a multi-dimensional deviation factor set through clear sky window filtering and three-dimensional occlusion analysis. Preferably, the step of decoupling the total deviation into a multi-dimensional deviation factor set includes: Calculate the total deviation between the actual irradiance reached at the verification reference point and the original simulated irradiance at the corresponding location; Filter time windows under clear sky conditions, and separate regional system deviation factors related to meteorological models based on the total deviation within clear sky periods; Based on the three-dimensional spatial geometric parameters of the verification benchmark point and the digital surface model of the surrounding buildings, the ray tracing algorithm is used to simulate the shadow occlusion throughout the year and calculate the local occlusion deviation factor caused by the occlusion of the surrounding environment. The remaining deviation after removing the regional system deviation factor and the local occlusion deviation factor from the total deviation is included in the installation and equipment deviation factor.

[0030] Specifically, each verification reference point exist The actual irradiance obtained by back-calculation from the physical model at any time Compared with the original simulated irradiance at the corresponding location Perform time series alignment and calculate total deviation. : ; Among them, the original simulated irradiance It is based on the latitude and longitude coordinates of the target area, utilizing simulation data from third-party meteorological databases including, but not limited to, PVGIS, Solargis, Meteonorm, or NASA-SSE. Because the aforementioned raw simulation data are often limited by spatial resolution and lack of local microenvironment modeling during the generation process, they cannot accurately reflect the complex actual irradiance conditions of distributed photovoltaic sites. Therefore, the calculated total deviation... In essence, it includes the combined effects of meteorological source system errors, local environmental obstruction errors, and power station performance deviations.

[0031] Since systematic bias typically manifests as a global scaling effect, it primarily stems from systematic drift in satellite inversion algorithms or numerical weather prediction models. Therefore, the clear sky index can be used... Using the ratio of measured irradiance to extraterrestrial irradiance as a criterion, in one feasible implementation, a selection process is adopted to screen for those that meet the criteria. Sample points that last for more than 3 hours are defined as clear sky windows. This window eliminates the interference of random cloud fluctuations on deviation decoupling. Therefore, the regional system deviation factor... The calculation formula is as follows: ; Among them, the regional system deviation factor This reflects the correction ratio of the meteorological background values ​​for the region. The median calculation is used to eliminate noise caused by occasional shading or momentary sensor anomalies at individual points, ensuring that this factor represents the common characteristics of the region.

[0032] Subsequently, a microscale 3D scene was constructed using the 3D spatial geometric parameters of the verification benchmark (including the support installation height, tilt angle, and azimuth angle) and the digital surface model of surrounding buildings. A ray tracing algorithm was then used to simulate the sun's annual trajectory, determining whether the light beam was interrupted by surrounding obstacles at each moment, and generating a binarized shadow function. ,in Represents no obstruction. This represents occlusion. Therefore, the calculation formula for the deviation separation step is as follows: ; If the measured value is within If the value is significantly lower than the simulated value after system bias correction within a certain period, then the difference is attributed to the local occlusion bias factor.

[0033] After eliminating meteorological and environmental errors, the remaining deviations reflect the power station's own health and installation quality. The calculation formula for the deviation analysis steps is as follows: ; If there is a deviation The system exhibits a cosine distribution characteristic related to the solar altitude angle. The optimal tilt angle and orientation are determined using the least squares method for fitting, and the installation angle deviation factor is calculated. If the residuals exhibit a long-term stable negative shift, the attenuation factor is calculated using the annualized performance loss rate method. The comprehensive installation and equipment deviation factor is: Based on the above steps, the total deviation is decoupled into a multi-dimensional deviation factor set { }

[0034] S4. Based on historical statistical experience, establish the prior distribution of the correction factor for each candidate point. Preferably, the step of establishing the prior distribution of the correction factor for each candidate site includes: The deviation distribution characteristics between historical meteorological observation data and satellite inversion data of the target area are obtained as global prior constraints. Based on the spatial distance and environmental similarity between each candidate point and each verification benchmark point, a spatial interpolation method is used to determine the spatial co-estimation of the correction factor as a local prior constraint. By combining the global prior constraints and the local prior constraints, the prior mean and prior variance of the correction factors for each candidate point are set to generate a prior distribution model that follows a normal distribution.

[0035] Specifically, historical meteorological station data (usually more than 10 years) or high-confidence satellite imagery data are collected within the target area, and satellite inversion irradiance at the same location and time resolution is extracted. Compared with ground-measured irradiance Define the proportional deviation. This ratio reflects the systematic bias characteristics of the satellite inversion model in this region.

[0036] Calculate the mean of the proportion deviations of all samples. and variance : ; ; The normal distribution As a global prior for the correction factor, it represents the average bias characteristics at the regional scale and does not depend on specific locations.

[0037] Using the selected benchmark points and their corresponding correction factors, the local prior of the candidate points is estimated using spatial interpolation. Preferably, inverse distance weighted interpolation is used. Let the candidate points be... The coordinates are There are K nearby calibration reference points, and their correction factor values ​​are... The spatial distance is Then the local prior mean of the candidate point. for: ; Local prior variance The variance is given by the interpolation error estimate.

[0038] Treating global and local priors as two independent information sources, a Bayesian conjugate update is used to fuse them, resulting in the fused prior distribution of candidate site correction factors. : ; ; This fusion method automatically balances the prior mean between the two constraints: when local information is reliable, the prior is more biased towards local estimates; when local information is missing or variance is large, the prior regresses to the global average level. Finally, the adjustment factors for each candidate point are... prior distribution All follow a normal distribution This serves as the input for subsequent Bayesian inference.

[0039] S5. Construct a likelihood function using the multi-dimensional deviation factor set, calculate the posterior distribution of the correction factor for each candidate point by combining the prior distribution, and determine the expected value of the correction factor for each candidate point and its uncertainty interval based on the posterior distribution. Preferably, the step of calculating the posterior distribution of the correction factor for each candidate point based on the prior distribution includes: Based on the multi-dimensional deviation factor set of the verification benchmark, a likelihood function for the correction factor is constructed. Using the prior distribution as prior information for the correction factor, and combining it with the likelihood function, probability fusion is performed using Bayes' theorem to calculate the posterior distribution of the correction factor for each candidate point. The posterior mean and posterior variance of the posterior distribution are extracted and used as the expected value and uncertainty measure of the correction factor for each candidate point, respectively.

[0040] Specifically, suppose there are M verification benchmark points within the target area. For the j-th benchmark point, its multi-dimensional deviation factor set, such as the comprehensive correction factor, has been obtained through step S3. This observation can be represented as the true correction factor. With observation noise sum: ; in The observation variance can be estimated based on the bias-decoupled residuals. For candidate locations... Its correction factor The likelihood function is defined as: given Under the assumption that all benchmark deviation factors were observed The probability density of the candidate points. To simplify the calculation, it is usually assumed that the observations of each reference point are independent of each other, and that the true correction factor of each candidate point satisfies the consistency relationship with the observation value of the reference point; that is, it is assumed that the observation values ​​of each reference point revolve around the same true value. Fluctuation. The likelihood function at this point is: ; The prior distribution generated in step S4 is used as the prior information for the correction factor. From S4, we know the candidate points... Correction factor Follows a normal prior distribution: ; According to Bayes' theorem, the posterior distribution is proportional to the product of the prior distribution and the likelihood function: ; Since both the prior distribution and the likelihood function are normally distributed, and the posterior distribution is also normally distributed, their parameters can be directly calculated analytically using Bayes' theorem. Let the joint observation accuracy of all reference points be... The weighted mean of the observations is Then the mean and variance of the posterior distribution are: ; ; During this fusion process, when the candidate point is close to the high-precision benchmark, the likelihood function will dominate and the posterior distribution will converge towards the measured value; when the observed data is sparse or noisy, the posterior distribution will automatically regress to the prior distribution.

[0041] S6. Use the expected value of the correction factor to correct the original simulated irradiation data corresponding to the candidate points, generate irradiation resource distribution data, and output the corresponding uncertainty quantification results.

[0042] Preferably, the step of correcting the original simulated irradiation data corresponding to the candidate sites includes: Obtain the original simulated irradiation data and the corresponding expected value of the correction factor for the candidate sites; The expected value of the correction factor is used as a calibration operator to apply to the original simulated irradiation data to obtain the corrected irradiation distribution data. By combining the uncertainty measure of the candidate points and mapping the probability density function, the confidence interval of the corrected irradiance distribution data at a preset confidence level is determined.

[0043] Specifically, candidate locations are extracted from the original simulated irradiation sequences provided by third-party databases such as PVGIS and Solargis. Simultaneously, the expected value of the correction factor, calculated from the Bayesian posterior distribution in step S5, is retrieved. This expected value has eliminated regional system bias, local occlusion, and equipment performance interference.

[0044] Correction factor expectation value As a calibration operator, it is applied to the original simulated irradiation data. In practice, the original sequence is corrected hourly using a multiplication factor or incremental compensation, calculated as follows: ; Generated corrected irradiance distribution data The original simulation model was corrected for prediction deviations in the complex microscale environment of the city, making it closer to the actual available solar radiation intensity on the ground.

[0045] To support the financing-grade assessment of distributed photovoltaic projects, a quantitative reliability analysis of the corrected data is required. This analysis should be combined with the posterior variance output from step S5. Construct a normal probability density function with the corrected irradiance as the mean. ,in This is the standard deviation of irradiance fluctuation obtained by mapping the posterior variance. Based on the user-preset confidence level (e.g., 95%), it is integrated on the probability density curve to determine the range of irradiance fluctuation. If the uncertainty measure is large, the generated confidence interval will be wide, suggesting that the correction result may fluctuate due to the distance of the point from the base station or the complexity of the environment; if the variance is small, the confidence interval will converge, proving that the correction result has extremely high certainty.

[0046] Preferably, the method further includes a model dynamic update step: Monitor the real-time operational data of the verification benchmark point and calculate the rate of change of the multi-dimensional deviation factor set over time; When the rate of change exceeds a preset threshold, the parameter adjustment of the prior distribution is initiated, and the expected value of the correction factor of each candidate point is updated based on the adjusted prior distribution. The irradiation resource distribution data are incrementally corrected using the updated expected value of the correction factor.

[0047] Specifically, continuously monitor the real-time operational data of each calibration benchmark point, and repeat steps S1 to S3 to periodically calculate the latest multi-dimensional deviation factor set. Define the rate of change of the deviation factors over time. ,in This indicates the change of the multidimensional bias factor set over time. The rate of change quantifies the drift of regional systematic errors, changes in local shading environments, and fluctuations in equipment degradation rates.

[0048] Set preset change threshold (e.g., 5%). When the calculated rate of change... At that time, the system automatically initiates parameter adjustments for the prior distribution, including: The observed deviation drift is compensated into the original prior mean to form a new prior mean that reflects the current environmental state. If the rate of change is drastic, the prior variance should be appropriately increased to introduce more randomness and give the likelihood function a higher correction weight. Based on the adjusted prior distribution, the Bayesian inference in step S5 is re-executed to calculate the expected value of the updated candidate site correction factor.

[0049] Finally, the updated expected value of the correction factor is used to incrementally correct the existing irradiation resource distribution data.

[0050] Example 2 This embodiment takes a typical industrial and commercial area in a city as the target area. The building layout in this area is complex, and the photovoltaic points are significantly affected by the shading of surrounding tall buildings and urban aerosols.

[0051] In existing technologies, the following schemes are commonly used to obtain the irradiance of distributed photovoltaic systems: Scheme A is based on spatial interpolation of ground meteorological stations, which uses measured data from the most recent typical meteorological station closest to the project site for spatial extrapolation; Scheme B is based on numerical simulation of satellite inversion models, which uses imagery data from geostationary orbit satellites combined with gridded simulation data generated by atmospheric transport models.

[0052] This embodiment performs in-depth verification and adaptive correction on the original simulation data of Scheme B described above. The specific process is as follows: Step 1: Benchmark Observation and Physical Inversion Several representative grid-connected commercial and industrial rooftop power stations were selected within the target area as verification benchmarks. Engineering parameters such as component nominal power, inverter rated efficiency, and power temperature coefficient were obtained. Using the single-diode physical reverse model described in this invention, the real-time output power collected by each power station was converted into the actual irradiance reaching the component surface.

[0053] Step 2: Multidimensional physical decoupling of total deviation By comparing the original simulation data with the actual irradiance obtained from physical inversion, the following decoupling operation is performed: Filter the clear sky time window within the region, calculate the average deviation ratio between the measured value and the simulated value, and identify the regional system deviation factor caused by the satellite inversion algorithm; A three-dimensional scene of the target area is constructed using a digital surface model. Combined with the geometric parameters of the component installation, the sun's trajectory is simulated using a ray tracing algorithm to identify and calculate the amount of irradiance energy loss caused by the shading of surrounding buildings. After deducting system and shading deviations, the remaining residuals are aggregated into equipment deviation factors that account for performance degradation caused by installation angle deviations and component dust accumulation.

[0054] Step 3: Posterior correction based on Bayesian inference For candidate sites within the region to be developed, a prior probability distribution composed of historical meteorological patterns and geographical environmental similarity is established. The decoupled multi-dimensional bias factor set is input into the likelihood function, and probability fusion is performed using Bayes' theorem. This process can automatically identify the environmental characteristics of different candidate sites and calculate the expected value of the correction factor for that specific location.

[0055] Step 4: Correcting the output and quantifying uncertainty The original simulated sequence is calibrated using the expected value of the correction factor, and the confidence interval at a preset confidence level is output based on the variance characteristics of the posterior distribution, thus completing the risk quantification of resource assessment.

[0056] Table 1. Comparison of the effects of this solution with existing technologies

[0057] The comparison between this scheme and existing technologies is shown in Table 1, based on the measured verification of sample points within the target area. As can be seen from the comparison data in Table 1, the scheme of this invention achieves a leap from kilometer-level regional grids to meter-level building points in spatial analysis. Its core advantage lies in the deep coupling of the single-diode physical inverse model and the ray tracing algorithm, reducing the average absolute error to within 5%, completely solving the shortcomings of Scheme A and Scheme B in failing to perceive local time-varying shading and urban microclimate system deviations. Simultaneously, based on the confidence interval of the Bayesian posterior distribution output, it provides mathematical support for the transformation of distributed photovoltaics from qualitative estimation to quantitative risk assessment (such as the P90 / P95 index). Combined with its self-learning dynamic update mechanism, it significantly improves the accuracy and financing-grade confidence of the power plant's full life-cycle resource assessment.

[0058] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for verifying and correcting distributed photovoltaic irradiance data, characterized in that, include: Obtain the operation data and engineering parameters of grid-connected photovoltaic power stations within the target area, and infer the actual irradiance of each power station location based on the physical model of photovoltaic modules; Based on the data quality evaluation indicators, verification benchmark points were selected from each power station location; The total deviation between the actual irradiance reached at the verification benchmark point and the original simulated irradiance at the corresponding location is calculated, and the total deviation is decoupled into a multi-dimensional deviation factor set through clear sky window filtering and three-dimensional occlusion analysis. Based on historical statistical experience, a prior distribution of the correction factor for each candidate point is established. A likelihood function is constructed using the multidimensional deviation factor set. The posterior distribution of the correction factor for each candidate point is calculated in combination with the prior distribution. Based on the posterior distribution, the expected value of the correction factor for each candidate point and its uncertainty interval are determined. The original simulated irradiation data corresponding to the candidate sites are corrected using the expected value of the correction factor to generate irradiation resource distribution data and output the corresponding uncertainty quantification results. The multi-dimensional deviation factor set includes regional system deviation factors, local occlusion deviation factors, and installation and equipment deviation factors. The step of decoupling the total deviation into a multi-dimensional deviation factor set includes: Calculate the total deviation between the actual irradiance reached at the verification reference point and the original simulated irradiance at the corresponding location; Filter time windows under clear sky conditions, and separate regional system deviation factors related to meteorological models based on the total deviation within clear sky periods; Based on the three-dimensional spatial geometric parameters of the verification benchmark point and the digital surface model of the surrounding buildings, the ray tracing algorithm is used to simulate the shadow occlusion throughout the year and calculate the local occlusion deviation factor caused by the occlusion of the surrounding environment. The remaining deviation after removing the regional system deviation factor and the local occlusion deviation factor from the total deviation is included in the installation and equipment deviation factor.

2. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The steps for back-calculating the actual irradiance at each power station location based on the physical model of photovoltaic modules include: Obtain the nominal power and temperature coefficient of the component under standard test conditions; Obtain the real-time output power of each power station and, in conjunction with the inverter conversion efficiency, reverse-calculate the DC-side power. The DC-side power is compensated for environmental degradation using the temperature coefficient and component performance degradation rate to obtain the received light power; The received light power is inverted into the actual effective irradiance on the component surface based on the single diode equivalent circuit model.

3. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The steps for selecting verification reference points from the various power station locations include: Each power station location was screened based on four dimensions: data integrity, runtime, equipment stability, and environmental stability, eliminating locations that did not meet the preset conditions. Calculate the comprehensive data quality score for the remaining power station locations, and determine the locations with scores higher than the preset threshold as the verification benchmark points.

4. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The steps to establish the prior distribution of the correction factor for each candidate site include: The deviation distribution characteristics between historical meteorological observation data and satellite inversion data of the target area are obtained as global prior constraints. Based on the spatial distance and environmental similarity between each candidate point and each verification benchmark point, a spatial interpolation method is used to determine the spatial co-estimation of the correction factor as a local prior constraint. By combining the global prior constraints and the local prior constraints, the prior mean and prior variance of the correction factors for each candidate point are set to generate a prior distribution model that follows a normal distribution.

5. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The steps for calculating the posterior distribution of the correction factor for each candidate point based on the prior distribution include: Based on the multi-dimensional deviation factor set of the verification benchmark, a likelihood function for the correction factor is constructed. Using the prior distribution as prior information for the correction factor, and combining it with the likelihood function, probability fusion is performed using Bayes' theorem to calculate the posterior distribution of the correction factor for each candidate point. The posterior mean and posterior variance of the posterior distribution are extracted and used as the expected value and uncertainty measure of the correction factor for each candidate point, respectively.

6. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The steps for correcting the original simulated irradiation data corresponding to the candidate sites include: Obtain the original simulated irradiation data and the corresponding expected value of the correction factor for the candidate sites; The expected value of the correction factor is used as a calibration operator to apply to the original simulated irradiation data to obtain the corrected irradiation distribution data. By combining the uncertainty measure of the candidate points and mapping the probability density function, the confidence interval of the corrected irradiance distribution data at a preset confidence level is determined.

7. The method for verifying and correcting distributed photovoltaic irradiance data according to claim 1, characterized in that, The method also includes a model dynamic update step: Monitor the real-time operational data of the verification benchmark point and calculate the rate of change of the multi-dimensional deviation factor set over time; When the rate of change exceeds a preset threshold, the parameter adjustment of the prior distribution is initiated, and the expected value of the correction factor of each candidate point is updated based on the adjusted prior distribution. The irradiation resource distribution data are incrementally corrected using the updated expected value of the correction factor.

Images