New energy power prediction credibility generation method, device and equipment, and storage medium

Abstract

The application discloses a new energy power prediction credibility generation method and device, equipment and storage medium, relates to the new energy power generation technical field, and includes: obtaining the multi-weather source prediction data, historical power data, historical prediction data and station basic information of a target station; time alignment, unit unification and missing data filling preprocessing are carried out on the multi-weather source data; the multi-source prediction difference is calculated for the same time and same weather element, and the multi-level divergence including average, range and robust divergence is formed; the prediction risk level is evaluated in combination with the multi-level divergence, the station type, the importance of the weather element, the prediction lead time and the historical error sample; the risk is quantified into a credibility index and a corresponding grade, and strategies such as conventional output, conservative correction or risk warning are matched. The application can quantify the prediction reliability, identify high-uncertainty weather processes, and improve the support capability of power prediction for business decision-making.

Description

Technical Field

[0001] This application relates to the field of new energy power generation technology, and in particular to a method, apparatus, equipment and storage medium for generating new energy power prediction reliability. Background Technology

[0002] Renewable energy power forecasting is a crucial foundation for renewable energy grid connection and dispatch, power generation planning, electricity spot trading, and production and operation auxiliary decision-making. With the continuous growth of wind and solar power installed capacity, the problems of renewable energy output being strongly influenced by meteorological conditions, exhibiting large fluctuations and high randomness, are becoming increasingly prominent. In actual operation, forecasting systems typically integrate numerical weather prediction data from multiple sources. These various meteorological sources differ in model frameworks, initial field assimilation methods, spatial resolution, and physical parameterization schemes, often leading to significant discrepancies in the forecast results for future meteorological elements at the same weather station.

[0003] Related technologies mostly focus on improving the accuracy of point forecasts, or use simple averaging and fixed weighting to fuse multiple meteorological sources. These technologies cannot quantitatively reflect the reliability of the power forecast results themselves. They ignore the divergence information between meteorological sources during the fusion process, making it difficult to effectively identify highly uncertain weather processes. Fixed weighting fusion models are prone to failure in extreme weather and scenarios with rapid fluctuations in power output. Furthermore, they lack an effective mapping relationship between meteorological source divergence and forecast risks, making it impossible to match differentiated outputs and risk management strategies based on the forecast reliability. This makes it difficult to fully support the risk decision-making needs of scheduling, trading, and other businesses. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for generating the reliability of new energy power prediction. It can solve the problems in related technologies such as ignoring discrepancy information due to the fusion of multiple meteorological sources, the inability to quantitatively determine the reliability of power prediction, and the lack of a mapping relationship between discrepancies and prediction risks, as well as differentiated handling strategies.

[0005] According to a first aspect of this application, a method for generating the reliability of new energy power prediction is provided, comprising:

[0006] Acquire multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station; The multi-meteorological source forecast data is preprocessed, and the preprocessing includes at least time alignment, unit unification, and missing data completion. For the same forecast time and the same meteorological element, the differences between different meteorological sources are calculated, and a multi-level divergence degree is formed based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree. The risk level of the current forecast result is assessed by comprehensively considering the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data. The risk level is quantified into a prediction credibility index and its corresponding credibility level; Based on the confidence level, a corresponding strategy is selected from a variety of preset post-processing strategies and executed. The variety of post-processing strategies includes at least: when the confidence level is high, a conventional prediction output strategy is adopted; when the confidence level is low, a more conservative correction strategy or a risk warning is triggered.

[0007] According to a second aspect of this application, a new energy power prediction reliability generation device is provided, comprising: The acquisition module is configured to acquire multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station; The preprocessing module is configured to preprocess the multi-meteorological source forecast data, and the preprocessing includes at least time alignment, unit unification, and missing data completion. The calculation module is configured to calculate the differences between different meteorological sources for the same forecast time and the same meteorological element, and to form a multi-level divergence degree based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree. The assessment module is configured to assess the risk level of the current forecast result by comprehensively considering the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data. The quantization module is configured to quantify the risk level into a prediction confidence index and its corresponding confidence level. The execution module is configured to select and execute a corresponding strategy from a variety of preset post-processing strategies based on the confidence level. The variety of post-processing strategies includes at least: using a conventional prediction output strategy when the confidence level is high, and using a more conservative correction strategy or triggering a risk warning when the confidence level is low.

[0008] According to a third aspect of this application, an electronic device is provided, comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which are executed by at least one processor to enable the at least one processor to execute the new energy power prediction reliability generation method described in the first aspect above.

[0009] According to a fourth aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the new energy power prediction reliability generation method described in the first aspect above.

[0010] According to a fifth aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the new energy power prediction reliability generation method as described in the first aspect above.

[0011] This application addresses the problems in related technologies, such as ignoring divergence information due to multi-meteorological source data fusion, inability to quantitatively determine the reliability of power predictions, lack of mapping relationship between divergence and prediction risk, and lack of differentiated handling strategies. By preprocessing multi-meteorological source data to calculate the differences between different meteorological sources and forming multi-level divergence degrees, and combining station type, importance of meteorological elements, prediction lead time and historical error samples to assess the risk level of the current prediction, the risk is quantified into a prediction reliability index and corresponding level, and differentiated post-processing strategies are matched according to the reliability level. This achieves the technical effects of quantitatively characterizing the reliability of predictions, effectively identifying high-uncertainty weather processes, adapting output and risk handling to different reliability scenarios, and improving the ability of power prediction results to support business decision-making.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0013] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating a method for generating the reliability of new energy power prediction provided in an embodiment of this application; Figure 2 This is a schematic diagram of a new energy power prediction reliability generation device provided in an embodiment of this application. Detailed Implementation

[0015] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0016] The following description, with reference to the accompanying drawings, describes a method, apparatus, device, and storage medium for generating reliability of new energy power prediction according to embodiments of this application.

[0017] Figure 1 This is a flowchart illustrating a method for generating the reliability of new energy power prediction, provided in an embodiment of this application.

[0018] like Figure 1 As shown, the method includes the following steps: Step 101: Obtain multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station.

[0019] In some embodiments, multi-meteorological source forecast data for the target renewable energy power station is acquired. Multi-meteorological sources refer to meteorological forecast data provided by different numerical weather prediction systems or different meteorological service providers for the same renewable energy power station at the same forecast time. These data typically include meteorological elements such as wind speed, wind direction, irradiance, cloud cover, temperature, air pressure, and humidity. For wind farms, the focus is on collecting forecast data related to wind speed, wind direction, air pressure, temperature, and air density; for photovoltaic power stations, the focus is on collecting forecast data related to total irradiance, direct irradiance, diffuse irradiance, cloud cover, temperature, and humidity.

[0020] The system simultaneously acquires historical power data, historical forecast data, and basic information about the target power plant. Historical power data consists of records of the actual power generation of the plant over various historical periods, while historical forecast data represents the power prediction results for the corresponding historical periods. Basic information about the power plant includes its installed capacity, type, time characteristics, and necessary equipment status data. Specifically, the power plant type is divided into two categories: wind farms and photovoltaic power plants.

[0021] By collecting multi-source meteorological data that is appropriate for the type of station and combining it with complete historical operational data and basic station attribute data, a comprehensive and station-specific data source can be provided for subsequent reliability analysis, ensuring the business adaptability and data support capabilities of the subsequent calculation process.

[0022] Step 102: Preprocess the multi-meteorological source forecast data. The preprocessing includes at least time alignment, unit unification, and missing data completion.

[0023] In some embodiments, the acquired multi-source weather forecast data is preprocessed to eliminate format differences and data defects between different data sources, forming a standardized dataset that meets the requirements of subsequent analysis. Different numerical weather prediction systems or meteorological service providers have different data output rules; preprocessing is a fundamental step to ensure the comparability and usability of multi-source data.

[0024] Time alignment addresses the issues of inconsistent data frequencies and mismatched time points among various meteorological sources by uniformly converting all forecast data into the standard time granularity required for power prediction. This ensures a one-to-one correspondence between data from different sources at the same point in time, guaranteeing subsequent multi-source comparative analysis for the same forecast time. Unit standardization involves standardizing the units of measurement for various meteorological elements such as wind speed, temperature, air pressure, and irradiance, adopting industry-standard units to eliminate interference from dimensional differences in subsequent data calculations.

[0025] Missing data filling addresses the issue of missing time-series data caused by abnormal data transmission or interrupted model calculations. It employs methods such as interpolation of nearby times and reference filling from similar meteorological sources to ensure the continuity and completeness of time-series data from each meteorological source, thus avoiding the impact of missing local data on the continuity and accuracy of subsequent analysis.

[0026] By using standardized preprocessing procedures, differences in time dimension, unit of measurement and data integrity among multiple meteorological data sources can be effectively eliminated, ensuring the comparability and usability of multi-source data and providing accurate and reliable basic data support for subsequent meteorological discrepancy analysis.

[0027] Step 103: For the same forecast time and the same meteorological element, calculate the differences between different meteorological sources, and form a multi-level divergence degree based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree.

[0028] In some embodiments, for the same forecast time and the same meteorological element, the forecast results of different meteorological sources are compared pairwise one by one to calculate the forecast difference between each pair of meteorological sources on that element. After all meteorological sources have completed pairwise comparisons, multiple sets of difference results for the corresponding meteorological element at that time are formed, which serve as the basis data for calculating the degree of divergence.

[0029] Based on the above discrepancies, a multi-level divergence degree is constructed to characterize the degree of inconsistency in forecasts from multiple meteorological sources from different dimensions. The average divergence degree is obtained by averaging all pairwise discrepancies and reflects the overall level of difference among multiple meteorological sources on this meteorological element. The range divergence degree is obtained by extracting the maximum value among all pairwise discrepancies and reflects the magnitude of the difference between the two most divergent meteorological forecasts, characterizing the boundary of extreme divergence. The robust divergence degree is obtained by weakening the influence of individual anomalous meteorological sources that significantly deviate from the overall trend. This avoids interference from single anomalous forecast data on the overall divergence judgment, making the divergence degree results more consistent with the actual divergence level of mainstream meteorological forecasts.

[0030] By forming a multi-level divergence system consisting of average divergence, range divergence, and robust divergence, the degree of forecast inconsistency of multiple meteorological sources can be comprehensively quantified from three dimensions: overall level, extreme boundary, and disturbance resistance robustness. This system fully covers both conventional and anomalous data scenarios, providing multi-dimensional and highly reliable divergence characteristics for subsequent forecast risk assessment.

[0031] Step 104: Based on the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data, assess the risk level of the current forecast result.

[0032] In some embodiments, the risk level of the current power forecast results is assessed by comprehensively considering multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from historical power data and historical forecast data. This transforms the data differences between meteorological sources into uncertainty judgments directly related to power forecasting operations, enabling the risk assessment results to directly reflect the possibility of deviations in the forecast results.

[0033] During the assessment, the importance priority of meteorological elements was determined based on the type of power station. The influence of meteorological elements on the output of different types of renewable energy power stations varied significantly. Wind farm output changes were primarily driven by wind speed; therefore, wind speed discrepancies were of the highest importance, with discrepancies in wind direction, air pressure, and temperature serving as supplementary assessment criteria. Photovoltaic power station output was directly related to irradiance conditions; therefore, irradiance and cloud cover discrepancies were of higher importance, with temperature discrepancies serving as a secondary assessment criterion. The importance of meteorological elements could be adjusted according to seasonal and temporal changes, while also incorporating the impact of forecast lead time. The longer the forecast lead time, the higher the uncertainty of meteorological condition evolution, and the corresponding increase in forecast risk.

[0034] Historical error samples contain the actual prediction deviation patterns under similar divergence scenarios in the past, providing data support for risk assessment. By analyzing the prediction error magnitude and deviation tendency corresponding to different divergence patterns in history, the multi-level divergence characteristics of the current moment are matched with historical patterns. Combined with the correction of meteorological element weights and prediction lead time, the risk level assessment conclusion of the current prediction result is finally obtained.

[0035] By integrating and evaluating multiple factors, the limitations of simply analyzing meteorological dispersion are overcome. This allows risk assessment to be deeply linked to the power output characteristics of the power station, the timeliness of forecasts, and historical error patterns. The evaluation results are more in line with the actual business scenarios of power forecasting and can accurately reflect the deviation risk of the forecast results.

[0036] Step 105: Quantify the risk level into a prediction credibility index and its corresponding credibility level.

[0037] In some embodiments, the risk level of the assessed current forecast result is quantified and converted to generate a standardized forecast credibility index. The forecast credibility index is a comprehensive indicator used to describe the reliability of a new energy power forecast result. It is expressed using a scoring method from zero to one hundred. The higher the score, the higher the credibility of the power forecast result for the corresponding period, and the more suitable it is as a decision-making basis for scheduling, trading, operation and maintenance, etc.; the lower the score, the higher the forecast deviation risk for the corresponding period, and the more cautious its use in business applications. There is a clear inverse correlation between risk level and credibility index: the higher the risk level, the lower the corresponding credibility index score; the lower the risk level, the higher the corresponding credibility index score.

[0038] Based on the credibility index, corresponding credibility levels are formed according to preset score ranges. Credibility levels are typically divided into five tiers: high credibility, relatively high credibility, medium credibility, low credibility, and very low credibility, each corresponding to a fixed range of index scores. The resulting credibility levels are then linked to the corresponding power prediction results at that time, providing an intuitive hierarchical indicator of the prediction results' reliability.

[0039] By transforming risk levels into quantifiable credibility indices and clear credibility grades, abstract uncertainty judgments can be converted into intuitively identifiable and standardized output results. This provides a clear basis for matching subsequent differentiated processing strategies and also helps business personnel quickly grasp the reliability of forecasts at different times, thereby improving the business support capabilities of forecast results.

[0040] Step 106: Based on the confidence level, select and execute the corresponding strategy from a variety of preset post-processing strategies. The variety of post-processing strategies includes at least: using a conventional prediction output strategy when the confidence level is high, and using a more conservative correction strategy or triggering a risk warning when the confidence level is low.

[0041] In some embodiments, based on the obtained confidence level, a corresponding strategy is selected from a variety of preset post-processing strategies and executed, so that the power prediction results with different levels of reliability are matched with the appropriate processing method, thus forming a connection between confidence judgment and business application.

[0042] When the confidence level is high, a conventional forecast output strategy is adopted, directly outputting the original power forecast result. In this scenario, the forecasts from multiple meteorological sources have little divergence, corresponding to a low risk of forecast deviation. The forecast results can meet the usage requirements of routine business such as grid connection scheduling and transaction settlement, without the need for additional adjustments.

[0043] When the confidence level is low, a more conservative correction strategy than the conventional strategy is adopted, or a risk warning is triggered. Conservative correction aims to reduce the risk of prediction deviation by constraining the output range of the original power prediction results, making the output results closer to the reasonable range of actual power generation at the power plant. Risk warnings identify and alert relevant personnel during periods of low confidence, conveying information about prediction uncertainty and prompting them to reserve a risk buffer when formulating power generation plans and making trading decisions.

[0044] By matching different post-processing strategies with credibility levels, the prediction reliability judgment can be transformed into an output and handling method that can be directly implemented, taking into account both prediction efficiency in high-credibility scenarios and risk control in low-credibility scenarios, thereby improving the ability of power prediction results to support business decisions.

[0045] Compared with related technologies, in this embodiment, multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station are acquired; the multi-meteorological source forecast data is preprocessed, including at least time alignment, unit unification, and missing data completion; for the same forecast time and the same meteorological element, the differences between different meteorological sources are calculated, and a multi-level divergence degree is formed based on the differences, including at least average divergence degree, range divergence degree, and robust divergence degree; the risk level of the current forecast result is assessed by comprehensively considering the multi-level divergence degree, power station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data; the risk level is quantified into a forecast credibility index and its corresponding credibility level; according to the credibility level, a corresponding strategy is selected from a variety of preset post-processing strategies and executed, including at least: when the credibility level is high, a conventional forecast output strategy is adopted; when the credibility level is low, a more conservative correction strategy or a risk warning is triggered. It can solve the problems in related technologies such as ignoring divergent information due to the fusion of multiple meteorological sources, the inability to quantitatively judge the reliability of power prediction, the lack of mapping relationship between divergence and prediction risk and differentiated handling strategies, and achieve the technical effects of quantitatively characterizing the reliability of prediction, effectively identifying weather processes with high uncertainty, adapting to output and risk handling in different reliability scenarios, and improving the ability of power prediction results to support business decision-making.

[0046] As a specific implementation of this application, based on the basic scheme, the calculation of differences between different meteorological sources is further limited, including: For continuous meteorological variables such as wind speed, irradiance, temperature, humidity, and air pressure, the difference is represented by the absolute difference between the predicted values ​​of the meteorological variables. For annular variables such as wind direction, the difference is represented by the smallest included angle.

[0047] Specifically, when calculating the forecast differences between different meteorological sources, differentiated calculation rules are adopted based on the numerical attributes and physical characteristics of meteorological elements to ensure that the difference results can truly reflect the degree of discrepancy in meteorological forecasts.

[0048] For continuous meteorological variables such as wind speed, irradiance, temperature, humidity, and air pressure, their values ​​are distributed along a linear scale and do not exhibit cyclical jumps. Therefore, the absolute difference between the predicted values ​​of two meteorological sources at the same time for the same element is directly used to represent the discrepancy between the two predictions. The magnitude of the absolute difference is positively correlated with the degree of prediction divergence; the larger the difference, the greater the deviation between the predictions of the two meteorological sources for that meteorological element. The calculation logic closely follows the physical change patterns of these elements, and the process is simple and the results are intuitive.

[0049] For cyclical variables like wind direction, which represent direction in the form of angles and exhibit a circular, cyclical value range (zero degrees and 360 degrees correspond to the same wind direction), directly using linear absolute difference calculations can lead to misinterpretations of significant discrepancies between wind direction predictions close to zero degrees and 360 degrees. Therefore, for cyclical variables, the minimum included angle is used to represent the difference. The angle value corresponding to the shortest arc between two wind direction angles is calculated as the difference result, eliminating calculation biases caused by the cyclical numerical boundaries and accurately reflecting the actual degree of inconsistency in wind direction predictions.

[0050] By matching and adapting different calculation methods to meteorological elements with different attributes, the boundary calculation error of cyclic variables can be effectively avoided, ensuring the accuracy and physical rationality of the calculation of differences in multi-meteorological source predictions, and providing reliable basic data support for the subsequent construction of multi-level divergence.

[0051] As a specific implementation of this application, based on the basic scheme, the robust divergence degree is further obtained in the following way: Abnormal meteorological sources that significantly deviate from the overall trend will be downgraded or removed. The robustness of divergence is calculated with reference to the median or mainstream meteorological judgment of all meteorological sources.

[0052] Specifically, robustness of divergence is used to reduce the interference of individual anomalous meteorological sources that deviate significantly from the overall trend on the overall divergence assessment results, avoid the distortion of the overall divergence level by a single anomalous data, and improve the anti-interference ability and assessment reliability of the divergence index.

[0053] Robust divergence can be obtained by either reducing the weight of anomalous sources or removing them. First, based on pre-defined deviation judgment rules, anomalous meteorological sources that significantly deviate from the predicted trends of most meteorological sources are identified. Reducing the weight of anomalous sources in pairwise difference calculations weakens their impact on the final divergence result; removing anomalous sources directly excludes them from the current divergence calculation sample, calculating the divergence only based on the remaining normal meteorological sources, thus completely eliminating interference from anomalous data.

[0054] Robust divergence can also be calculated using the median or mainstream meteorological judgment as a reference. When using the median as a reference, first calculate the median of all meteorological source forecasts for the same meteorological element at the same time. Use the median as a benchmark to calculate the difference between each meteorological source's forecast and the benchmark value. Then, obtain the robust divergence based on the statistical analysis of these differences. The median itself has strong resistance to outliers and can avoid the influence of extreme values ​​on the benchmark. When using mainstream meteorological judgment as a reference, first select a set of mainstream meteorological sources whose forecast results fall within a concentrated distribution range. Use the forecast benchmark value of this set as a reference, and calculate the divergence only based on the meteorological sources within the set, excluding marginal data sources that deviate from the mainstream judgment. This makes the divergence result more closely reflect the actual divergence level of most meteorological sources.

[0055] By obtaining the degree of divergence through multiple robust processing methods, the interference of abnormal meteorological sources on the calculation of the degree of divergence can be effectively suppressed, the stability and reliability of the degree of divergence index can be improved, the misjudgment of the reliability of prediction caused by the abnormality of data from a single meteorological source can be avoided, and more accurate divergence characteristics can be provided for subsequent risk assessment.

[0056] As a specific implementation of this application, based on the basic scheme, it further includes, before comprehensively assessing the risk level of the current prediction results, the following: The average divergence, the range divergence, and the robust divergence are scaled uniformly to make the divergence of different meteorological elements comparable.

[0057] Specifically, before comprehensively assessing the risk level of the current forecast results, the average divergence, range divergence, and robust divergence of each meteorological element should be standardized in terms of scale. Different meteorological elements have different physical dimensions and numerical fluctuation ranges. The units of measurement and the normal fluctuation range of elements such as wind speed, irradiance, temperature, and cloud cover are significantly different. The original divergence values ​​of various elements are not directly comparable. If they are directly used for comprehensive risk assessment, the difference in dimensions and numerical scales will lead to an imbalance in element weights, and will fail to accurately reflect the actual impact of the divergence of various elements on the risk of power forecasting.

[0058] Scale unification uses the historical fluctuation range, physical allowable fluctuation threshold, or operational experience threshold of each meteorological element as a benchmark. Through standardized transformation, the original divergence degrees of different dimensions are mapped to a unified dimensionless numerical range, eliminating differences in numerical scale among various elements. During the transformation process, the mapping rules can be adjusted according to the operational impact of each element to ensure that the unified divergence degree values ​​can equitably reflect their contribution to forecast risk. After scale unification, the multi-level divergence degrees of different meteorological elements have an equally comparable numerical basis and can jointly participate in subsequent comprehensive uncertainty assessments.

[0059] By standardizing the scale, the differences in the dimensions and fluctuation ranges of different meteorological elements are eliminated, so that the degree of divergence of various elements has a unified measurement standard. This can avoid the problem of weight imbalance when multi-element fusion assessment and ensure the rationality and accuracy of subsequent risk assessment results.

[0060] As a specific implementation of this application, based on the basic scheme, the type of the site is further specified to include wind farms or photovoltaic power stations; the assessment of the risk level of the current forecast result by comprehensively considering multi-level divergence, site type, importance of meteorological elements, forecast lead time, and historical error samples includes: When the site type is a wind farm, the meteorological elements are prioritized from high to low as follows: wind speed divergence, wind direction divergence, air pressure divergence, and temperature divergence. In the high wind speed range, the contribution weight of the wind speed divergence to the risk level is increased. When the site type is a photovoltaic power station, the meteorological factors are prioritized from high to low as follows: irradiance divergence, cloud cover divergence, and temperature divergence. In cloudy or rapidly moving cloud scenarios, the contribution weight of irradiance divergence and cloud cover divergence to the risk level is increased.

[0061] Specifically, when comprehensively assessing the risk level of the current forecast results, differentiated meteorological element importance rules are matched based on the type of power plant, which includes both wind farms and photovoltaic power plants. The dominant factors for power output differ significantly among different types of renewable energy power plants, and the impact of forecast discrepancies among various meteorological elements on power forecast errors varies. Therefore, a tiered importance priority system is used to match the power output characteristics of the corresponding power plants.

[0062] When the site type is a wind farm, the priority of meteorological elements, from highest to lowest, is: wind speed divergence, wind direction divergence, air pressure divergence, and temperature divergence. Wind speed is the core meteorological element determining wind power output, and its forecast divergence has the most significant impact on power prediction accuracy; therefore, it is set as the highest priority. In high wind speed ranges, wind power output is significantly more sensitive to wind speed changes; even small differences in wind speed forecasts can cause substantial power deviations. Therefore, in this scenario, the contribution weight of wind speed divergence to risk level is increased, strengthening the risk characterization capability of wind speed divergence in high wind speed ranges, and making the risk assessment results more closely reflect the actual output patterns of wind power.

[0063] When the site type is a photovoltaic (PV) power station, the meteorological factors are prioritized from highest to lowest as follows: irradiance discrepancy, cloud cover discrepancy, and temperature discrepancy. Irradiance and cloud cover directly determine the solar radiation conditions of the PV site and are core factors affecting PV output; therefore, they are set as high priority. In cloudy or rapidly moving cloud scenarios, prediction deviations in irradiance and cloud cover can easily cause rapid fluctuations in PV output. In this scenario, the contribution weight of irradiance discrepancy and cloud cover discrepancy to the risk level is increased to amplify the risk impact of core factors in highly volatile scenarios and ensure the adaptability of risk assessment to typical complex weather scenarios.

[0064] By setting differentiated priorities for meteorological elements for wind farms and photovoltaic power plants, and dynamically increasing the contribution weight of core elements in corresponding high-risk scenarios, risk assessments can be deeply aligned with the output characteristics of different types of power plants, accurately reflect the actual impact of differences in meteorological elements on prediction errors, and effectively improve the accuracy and scenario adaptability of risk level assessment results.

[0065] As a specific implementation of this application, based on the basic solution, the method further defines the historical error samples extracted from the historical power data and the historical prediction data, and assesses the risk level of the current prediction result, including: Establish a historical error sample library that includes historical meteorological source divergence characteristics, historical power prediction results, historical actual power, prediction lead time, station type, seasonal and time period characteristics; The learning model establishes a mapping relationship between the divergence features in the historical error sample library and the prediction error risk. The learning model includes at least one of the gradient boosting tree model, random forest model, logistic regression model, or piecewise rule model. The current multi-level divergence degree and related features are input into the mapping relationship to obtain the probability that the current prediction result will have a large error, which serves as the basis for assessing the risk level.

[0066] Specifically, when extracting historical error samples from historical power data and historical forecast data and assessing the risk level of current forecast results, a historical error sample library is first constructed. This library includes historical meteorological source divergence characteristics, historical power forecast results, historical actual power, forecast lead time, station type, and seasonal and time-period characteristics. Historical meteorological source divergence characteristics correspond to the divergence data of multiple meteorological sources calculated at various historical moments. Historical actual power and historical power forecast results are matched one-to-one by time to statistically analyze actual forecast deviations under different scenarios. Forecast lead time, station type, season, and time-period characteristics serve as scenario classification dimensions, enabling the sample library to cover the correspondence between divergence and error under different operating conditions, providing comprehensive sample support for learning risk patterns.

[0067] A learning model is used to establish a mapping relationship between divergence features within a historical error sample database and the risk of prediction errors. The learning model can employ at least one of the following: gradient boosting tree model, random forest model, logistic regression model, or piecewise rule model. The model takes divergence features and scene features from the sample database as input and outputs the probability of prediction error occurrence or the magnitude of the error. Through training, the model uncovers the error risk patterns corresponding to different divergence modes. The model output is not the power prediction value itself, but rather the probability of a significant error in the current prediction result. This risk result is independent of the power prediction curve and can serve as an additional confidence indicator for the prediction curve.

[0068] After the mapping relationship is constructed, the multi-level divergence degree at the current moment and the corresponding site type, prediction lead time, seasonal period and other related features are input into the trained mapping relationship to calculate the probability of a large error in the current prediction result, which is used as the basis for assessing the risk level of the current prediction result.

[0069] By constructing a multi-dimensional historical error sample library and establishing a quantitative mapping between divergence characteristics and error risks using a learning model, the error patterns accumulated from historical operations can be transformed into the basis for current risk assessment, improving the objectivity and accuracy of risk level assessment. The output independent risk results can be directly used in conjunction with power prediction results, facilitating the business side to quickly identify prediction deviation risks.

[0070] As a specific implementation of this application, the method embodiment further includes, based on the basic solution, the following additional features: During online operation, the historical error sample library and meteorological source capability profile are continuously updated using newly generated power prediction and actual power data.

[0071] Specifically, during the online operation of the method, iterative updates are continuously completed through feedback of measured data to ensure the long-term adaptability of risk assessment and credibility judgment. After the end of each prediction cycle, the power prediction results for the corresponding period are matched with the actual power generation data of the power plant, and combined with the multi-meteorological source divergence characteristics calculated during the same period, to form newly added operational sample data.

[0072] By continuously updating the historical error sample library with newly added data, new meteorological source divergence characteristics, power prediction results, and actual power data are combined with corresponding prediction lead times, station types, and seasonal and time-period characteristics to form standard error samples, which are then added to the sample library. The continuous expansion of the sample library can continuously cover more weather scenarios and operating conditions, adapting to error pattern shifts caused by meteorological model adjustments and changes in station equipment status, ensuring that the mapping relationship of error risk always closely matches the actual operational performance of the stations.

[0073] Simultaneously, the newly added data is used to continuously update the meteorological source capability profile, and the prediction accuracy, deviation trend, and frequency of anomalies of each meteorological source under different recent scenarios are statistically analyzed to dynamically grasp the changes in the prediction capability of each meteorological source. The updated meteorological source capability profile can provide real-time basis for the identification of anomalies and weight adaptation in subsequent divergence calculations, ensuring the rationality and accuracy of divergence assessment.

[0074] By continuously updating the historical error sample database and meteorological source capability profiles online, a closed-loop operation mechanism is formed, which includes prediction output, measured feedback, and model iteration. This enables the credibility assessment logic to continuously adapt to the dynamic changes in the station's operating status and meteorological source performance, avoiding the attenuation of assessment accuracy caused by model solidification, and ensuring the long-term stability of the method and the reliability of the assessment results.

[0075] Figure 2 This is a schematic diagram of the structure of a new energy power prediction reliability generation device provided in an embodiment of this application, as shown below. Figure 2 As shown, it includes: The acquisition module 201 is configured to acquire multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target new energy power station.

[0076] The preprocessing module 202 is configured to preprocess the multi-meteorological source forecast data, and the preprocessing includes at least time alignment, unit unification and missing data completion.

[0077] The calculation module 203 is configured to calculate the differences between different meteorological sources for the same forecast time and the same meteorological element, and to form a multi-level divergence degree based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree.

[0078] The assessment module 204 is configured to assess the risk level of the current forecast result by comprehensively considering the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data.

[0079] The quantization module 205 is configured to quantify the risk level into a prediction credibility index and its corresponding credibility level.

[0080] The execution module 206 is configured to select and execute a corresponding strategy from a variety of preset post-processing strategies based on the confidence level. The variety of post-processing strategies include at least: using a conventional prediction output strategy when the confidence level is high, and using a more conservative correction strategy or triggering a risk warning when the confidence level is low.

[0081] It should be noted that other corresponding descriptions of the functional units involved in the new energy power prediction reliability generation device provided in this embodiment can be found in [reference]. Figure 1 The corresponding descriptions in [the document] will not be repeated here.

[0082] Based on the above, Figure 1 The embodiment illustrates a method for generating reliability for new energy power prediction. Correspondingly, this embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method. Figure 1 This paper presents a method for generating credibility of new energy power prediction.

[0083] Based on the above, Figure 1 The embodiment illustrates a method for generating reliability for new energy power prediction. Correspondingly, this embodiment also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 This paper presents a method for generating credibility of new energy power prediction.

[0084] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0085] Based on the above, Figure 1 A method for generating credibility of new energy power prediction is shown, and Figure 2 To achieve the above objectives, the present application also provides an electronic device, such as a personal computer or a server, in the illustrated virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 This paper presents a method for generating credibility of new energy power prediction.

[0086] In some embodiments, the aforementioned physical device may further include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit such as a keyboard, etc., and optionally, a USB interface, a card reader interface, etc. In some embodiments, the network interface may include a standard wired interface, a wireless interface (such as a Wi-Fi interface), etc.

[0087] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0088] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0089] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for generating credibility of new energy power prediction, characterized in that, include: Acquire multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station; The multi-meteorological source forecast data is preprocessed, and the preprocessing includes at least time alignment, unit unification, and missing data completion. For the same forecast time and the same meteorological element, the differences between different meteorological sources are calculated, and a multi-level divergence degree is formed based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree. The risk level of the current forecast result is assessed by comprehensively considering the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data. The risk level is quantified into a prediction credibility index and its corresponding credibility level; Based on the confidence level, a corresponding strategy is selected from a variety of preset post-processing strategies and executed. The variety of post-processing strategies includes at least: when the confidence level is high, a conventional prediction output strategy is adopted; when the confidence level is low, a more conservative correction strategy or a risk warning is triggered.

2. The method for generating credibility of new energy power prediction according to claim 1, characterized in that, The calculation of differences between different meteorological sources includes: For continuous meteorological variables such as wind speed, irradiance, temperature, humidity, and air pressure, the difference is represented by the absolute difference between the predicted values ​​of the meteorological variables. For annular variables such as wind direction, the difference is represented by the smallest included angle.

3. The method for generating credibility of new energy power prediction according to claim 1, characterized in that, The robustness degree of divergence is obtained through the following method: Abnormal meteorological sources that significantly deviate from the overall trend will be downgraded or removed. The robustness of divergence is calculated with reference to the median or mainstream meteorological judgment of all meteorological sources.

4. The method for generating credibility of new energy power prediction according to claim 1, characterized in that, Before comprehensively assessing the risk level of the current forecast results, the following is also included: The average divergence, the range divergence, and the robust divergence are scaled uniformly to make the divergence of different meteorological elements comparable.

5. The method for generating credibility of new energy power prediction according to claim 1, characterized in that, The site types include wind farms or photovoltaic power stations; the assessment of the risk level of the current forecast results, based on a comprehensive multi-level divergence, site type, importance of meteorological elements, forecast lead time, and historical error samples, includes: When the site type is a wind farm, the meteorological elements are prioritized from high to low as follows: wind speed divergence, wind direction divergence, air pressure divergence, and temperature divergence. In the high wind speed range, the contribution weight of the wind speed divergence to the risk level is increased. When the site type is a photovoltaic power station, the meteorological factors are prioritized from high to low as follows: irradiance divergence, cloud cover divergence, and temperature divergence. In cloudy or rapidly moving cloud scenarios, the contribution weight of irradiance divergence and cloud cover divergence to the risk level is increased.

6. The method for generating credibility of new energy power prediction according to claim 1, characterized in that, The step of extracting historical error samples from the historical power data and the historical prediction data, and assessing the risk level of the current prediction result, includes: Establish a historical error sample library that includes historical meteorological source divergence characteristics, historical power prediction results, historical actual power, prediction lead time, station type, seasonal and time period characteristics; The learning model establishes a mapping relationship between the divergence features in the historical error sample library and the prediction error risk. The learning model includes at least one of the gradient boosting tree model, random forest model, logistic regression model, or piecewise rule model. The current multi-level divergence degree and related features are input into the mapping relationship to obtain the probability that the current prediction result will have a large error, which serves as the basis for assessing the risk level.

7. The method for generating credibility of new energy power prediction according to claim 6, characterized in that, Also includes: During online operation, the historical error sample library and meteorological source capability profile are continuously updated using newly generated power prediction and actual power data.

8. A new energy power prediction reliability generation device, characterized in that, include: The acquisition module is configured to acquire multi-meteorological source forecast data, historical power data, historical forecast data, and basic information of the target renewable energy power station; The preprocessing module is configured to preprocess the multi-meteorological source forecast data, and the preprocessing includes at least time alignment, unit unification, and missing data completion. The calculation module is configured to calculate the differences between different meteorological sources for the same forecast time and the same meteorological element, and to form a multi-level divergence degree based on the differences. The multi-level divergence degree includes at least the average divergence degree, the range divergence degree, and the robust divergence degree. The assessment module is configured to assess the risk level of the current forecast result by comprehensively considering the multi-level divergence, station type, importance of meteorological elements, forecast lead time, and historical error samples extracted from the historical power data and the historical forecast data. The quantization module is configured to quantify the risk level into a prediction confidence index and its corresponding confidence level. The execution module is configured to select and execute a corresponding strategy from a variety of preset post-processing strategies based on the confidence level. The variety of post-processing strategies includes at least: using a conventional prediction output strategy when the confidence level is high, and using a more conservative correction strategy or triggering a risk warning when the confidence level is low.

9. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the new energy power prediction reliability generation method according to any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the new energy power prediction reliability generation method according to any one of claims 1-7.

Images