CMIP6 mode output deviation correction method based on multi-mode multi-method multi-index set weighting, storage medium and equipment
By employing a CMIP6 model output bias correction method that weights multiple models, methods, and indicators, the problems of CMIP6 model output bias and the poor performance of single correction methods are solved, achieving higher accuracy in future climate prediction and energy resource assessment, and promoting the efficient utilization of wind and solar energy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HARBIN INST OF TECH
- Filing Date
- 2026-01-08
- Publication Date
- 2026-05-22
AI Technical Summary
The CMIP6 model output has biases, and existing studies mostly use a single evaluation index or a single bias correction method, which means that the correction effect needs to be improved.
A CMIP6 model output bias correction method based on multi-model, multi-method, and multi-index set weighting is adopted. By acquiring different global climate models and observational data, using multiple bias correction methods and multiple evaluation indicators, skill scores of different combinations are calculated to determine their weights in the final model output, and the final future model data is generated.
It significantly improves the accuracy and reliability of model output data, enhances the precision of future climate predictions, and promotes the efficient utilization of wind and solar energy and rational energy planning.
Smart Images
Figure CN122072676A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of meteorological forecasting technology and relates to a method, storage medium and device for correcting model output deviation. Background Technology
[0002] In the process of promoting the green and low-carbon transformation of the energy system, China's wind power and photovoltaic industries have developed rapidly. According to a report from the National Energy Administration, China's installed capacity of wind and photovoltaic power will surpass that of thermal power for the first time in the first quarter of 2025. Wind and photovoltaic power are gradually moving from a marginal position in the power supply structure to a dominant role, playing an increasingly important role in addressing the crisis of fossil fuel depletion and the challenges of global warming. Against this backdrop, it is necessary to conduct refined wind and solar resource assessments to utilize wind and solar power more efficiently and rationally. Resource assessment aims to determine the availability of wind and solar power generation potential in specific areas by analyzing the spatiotemporal distribution characteristics of climate parameters such as wind speed and irradiance, thereby providing a reliable basis for decision-making on the site selection, layout, and planning of wind and solar power plants. However, current mainstream resource assessment methods mainly rely on historical remote sensing data and lack attention to future trends in wind and solar resource reserves, which to some extent restricts the development of renewable energy technologies.
[0003] To quantitatively assess the future potential for wind and solar resource development, the most crucial task is obtaining future irradiance and wind speed data. The Coupled Model Intercomparison Project (CMIP), overseen by the Coupled Model Working Group of the World Climate Research Programme, addresses this issue and is currently in its sixth phase (CMIP6). This is because the Scenario Model Intercomparison Project, one of the 23 sub-projects approved by CMIP6, shares past, present, and future climate data simulated by global climate models by climate research institutions and teams from various countries. Therefore, using CMIP6 model data to analyze the evolution of future climate parameters such as wind speed and irradiance to assess future wind and solar energy resource endowments is not only a common and reasonable practice but has also attracted widespread attention from numerous scholars. Although many studies have explored the impact of climate change on the potential for wind and solar resource development based on CMIP6 data, many works have overlooked the limitations of systematic bias in model data. This is a well-known problem because global climate models struggle to accurately simulate the components of the Earth's climate system, including the interaction of the atmosphere, ocean, and land, and forcing data cannot objectively reflect changes in external factors affecting the climate system. Therefore, in order to improve the reliability of model data, bias correction is required when applying model data.
[0004] Model data bias correction methods are generally divided into two categories: dynamic downscaling and statistical downscaling. The former uses climate parameters simulated by global climate models as boundary conditions and inputs them into high-resolution regional climate models for simulation. Regional climate models can provide higher-resolution climate information for local areas and better capture local climate characteristics. However, because regional climate models require significant computational resources and time, they are difficult to provide model data over large areas and long time spans, so dynamic downscaling is not currently considered. The latter establishes a statistical relationship between historical model outputs and observational data and applies it to future model outputs. Although statistical downscaling ignores the physical processes of the climate system, its high computational efficiency and flexibility have made it a commonly used method for model data bias correction. However, most studies only use one bias correction method and a single evaluation index, ignoring the fact that no single method can consistently perform well on a specific evaluation index in all situations. Therefore, the correction effect of using a single evaluation index or a single bias correction method still needs improvement. Summary of the Invention
[0005] This invention aims to address the issue of deviations in the output of CMIP6 mode, as well as the need to improve the correction effect of using only a single evaluation index or a single deviation correction method.
[0006] A CMIP6 model output bias correction method based on multi-mode, multi-method, and multi-index set weighting includes the following steps:
[0007] S100. Obtain CMIP6 model output data for the target area simulated by different global climate models, including historical and future climate data; at the same time, obtain an observation data dataset with the same geographical range and time span, wherein the observation data dataset is a climate reanalysis dataset or a measured dataset.
[0008] S200. Divide the historical data and observation data output by the CMIP6 model into three parts: calibration dataset, validation dataset, and test dataset;
[0009] S300, Based on different CMIP6 model output bias correction methods, the CMIP6 model outputs in the validation dataset, test dataset, and future dataset are generated by correcting the statistical relationship between the CMIP6 model outputs in the calibration dataset and the observation data.
[0010] S400. Based on the observational data in the validation dataset and the bias-corrected CMIP6 model output, calculate the skill score for different model-correction method combinations, considering all models, all bias correction methods, all evaluation indicators, and all latitude and longitude point data. Then score based on skill Determine the weight of the CMIP6 mode output generated by different combinations in the final CMIP6 mode output. ;in, , , and These represent the pattern index, deviation correction index, latitude and longitude point index, and index, respectively. , , and These represent the number of models, the number of bias correction methods, the number of latitude and longitude points, and the number of evaluation indicators, respectively. Indicates the first The first mode uses the first The bias correction method in the first The first latitude and longitude point obtained Each evaluation indicator value;
[0011] S500, the CMIP6 mode output and its corresponding weights after correction for different combinations of biases in the test dataset and future dataset. The final CMIP6 mode output is obtained.
[0012] Furthermore, the observation data dataset uses the ERA5 reanalysis dataset.
[0013] Furthermore, in step S200, before dividing the historical data and observation data output by the CMIP6 model into three parts, the units of the CMIP6 model output and the observation data are unified.
[0014] Furthermore, in the process of unifying the CMIP6 mode output and the observation data units, when the spatial resolution of the CMIP6 mode output is inconsistent with the spatial resolution of the observation data, bilinear interpolation technology is used to resample the CMIP6 mode output to the spatial resolution of the observation data.
[0015] Furthermore, in step S100, the time resolution of both the CMIP6 model output data and the observation data is 1 day.
[0016] Furthermore, the weights .
[0017] Furthermore, the process of finally outputting the CMIP6 mode in step S500 includes:
[0018] according to Obtain the final historical CMIP6 mode output from the test dataset. In the formula For the first The first mode uses the first Historical CMIP6 mode output generated by the bias correction method;
[0019] Furthermore, the process of finally outputting the CMIP6 mode in step S500 also includes:
[0020] The final historical CMIP6 mode output is obtained by using this method. The final CMIP6 mode output in the future dataset is obtained in the same way.
[0021] A computer storage medium storing at least one instruction, which is loaded and executed by a processor to implement the CMIP6 mode output deviation correction method based on a multi-mode, multi-method, and multi-index set weighting.
[0022] A CMIP6 mode output deviation correction device based on multi-mode, multi-method, and multi-index set weighting, the device includes a processor and a memory, the memory stores at least one instruction, the at least one instruction is loaded and executed by the processor to implement the CMIP6 mode output deviation correction method based on multi-mode, multi-method, and multi-index set weighting.
[0023] The beneficial effects of this invention are as follows:
[0024] This invention proposes a CMIP6 model output bias correction method based on a multi-model, multi-method, and multi-indicator ensemble weighting. This method assigns different weights to model data generated by different model-correction method combinations, obtaining superior future model data compared to a single combination, significantly improving the accuracy of model output data. Specifically, the bias correction effect of different combinations is evaluated through multiple indicators, with a combination receiving a larger weight when its bias correction effect is better, and a smaller weight when a combination's ability to correct model output bias is weaker. This combination helps to fully leverage the concept of ensemble modeling, avoiding over-reliance on a single bias correction method and a single evaluation indicator, thereby improving the reliability of future climate prediction data. Furthermore, this invention has significant academic value and practical implications for improving the accuracy of future wind and solar resource assessments, promoting the efficient utilization of wind and solar energy, and formulating reasonable regional energy planning schemes. Attached Figure Description
[0025] Figure 1 The root mean square error (RMSE) plots are shown for different methods and the method of the present invention during the test period of the embodiments. Detailed Implementation
[0026] To address the issues of bias in CMIP6 model outputs and the reliance on single bias correction methods and evaluation indicators in many future wind and solar resource assessment studies, this invention proposes a CMIP6 model output bias correction method based on a weighted set of multiple models, methods, and indicators. The invention also proposes corresponding storage media and devices. First, different bias correction methods are used to correct historical and future data from different models. Next, the ability of different model-correction method combinations to improve the reliability of historical model data is quantified based on different evaluation indicators, and the weight of different model output combinations in the final model output is determined accordingly. Finally, based on the future model data generated by different combinations and their corresponding weights, the bias-corrected future model data is obtained. The invention will be described in detail below with reference to specific implementation methods. Specific implementation method one:
[0028] This implementation method is a CMIP6 model output bias correction method based on multi-mode, multi-method, and multi-index set weighting, specifically including the following steps:
[0029] S100, Download Data:
[0030] Download historical and future climate data (e.g., irradiance and 100-meter wind speed) for the target area from the Earth System Grid Consortium website, simulated by different global climate models. Simultaneously, download climate reanalysis data with the same geographical extent and time span as the historical CMIP6 data from the ERA5 website developed by the European Centre for Medium-Range Weather Forecasts (ECMWF). Both model data and reanalysis data have a 1-day temporal resolution. Furthermore, select only one climate change scenario for future climate data.
[0031] S200, Preprocessed data:
[0032] The units of CMIP6 mode output and ERA5 reanalysis data are standardized, and the historical data output from CMIP6 mode and the ERA5 reanalysis data are divided into three parts: calibration dataset, validation dataset, and test dataset. CMIP6 data includes historical datasets and future datasets. In this embodiment, the historical dataset is divided into calibration datasets, validation datasets, and test datasets. Similarly, the ERA5 reanalysis datasets within the same time span as CMIP6 are also divided into calibration datasets, validation datasets, and test datasets.
[0033] When the spatial resolution of the CMIP6 model output is inconsistent with the spatial resolution of the observation data, bilinear interpolation is used to resample the model output to the spatial resolution of the observation data.
[0034] S300. Generate pattern data after processing with the single-pattern correction method:
[0035] Assuming that the statistical relationship between the CMIP6 model output and the observed data remains unchanged in the historical and future periods, we can introduce model output bias correction methods such as additive linear scaling, quantile mapping, and random forest. Based on the statistical relationship between the model output and the observed data in the calibration dataset, we can generate the bias-corrected model output in the validation dataset, test dataset, and future dataset.
[0036] S400. Determine the weights of different mode-correction method combinations:
[0037] Based on the observational data in the validation dataset and the bias-corrected model output, the ability of different model-correction method combinations to correct model output bias is quantified and evaluated using multiple evaluation metrics. For example, root mean square error (RMSE) is used as an evaluation metric, and the specific calculation formula is as follows:
[0038] (1)
[0039] In the formula, This represents the calculated RMSE. This indicates climate parameters such as irradiance and wind speed. The subscript 'a' indicates climate parameters after bias correction, and the subscript 'o' indicates observational data. and These represent the data point index and the number of data points at the same latitude and longitude point, respectively. Generally speaking, the lower the RMSE calculated for a combination on the validation dataset, the better the bias correction effect of that combination.
[0040] Similarly, other negatively oriented positive evaluation metrics (i.e., the smaller the value of the evaluation metric, the better the bias correction performance, and all evaluation metrics are positive) can be used for evaluation, such as mean absolute error and mean absolute percentage error.
[0041] Furthermore, considering all models, all bias correction methods, all evaluation indicators, and all latitude and longitude point data, skill scores for different combinations are introduced to measure the relative performance of different model-correction method combinations in reducing model bias. The formulas for calculating the skill scores for different combinations are shown below:
[0042] (2)
[0043] In the formula, , , and These represent the pattern index, deviation correction index, latitude and longitude point index, and index, respectively. , , and These represent the number of models, the number of deviation correction methods, the number of latitude and longitude points, and the number of evaluation indicators, respectively. Indicates the first The first mode uses the first The bias correction method in the first The first latitude and longitude point obtained Each evaluation indicator value, Then it means the first The first mode uses the first The skill score for the bias correction method.
[0044] As can be seen from equation (2), both the numerator and denominator on the right-hand side of the equation are inverses of all evaluation indicators. This means that the lower the evaluation indicator value of a combination, the higher the skill score of that combination. The denominator introduces... This is to calculate the average of a certain evaluation index for all combinations at the same latitude and longitude point, and and The introduction of this feature is to reflect the bias correction performance of a particular combination across all latitude and longitude points and all evaluation indicators in the target area. After obtaining the skill scores for all combinations, the weight of the model output generated by different combinations in the final model output is determined by the following formula:
[0045] (3)
[0046] In the formula, For the first The first mode uses the first The weights corresponding to different bias correction methods. Undoubtedly, the better the bias correction effect of a certain mode-correction method combination, the greater the weight of that combination.
[0047] S500: Obtain historical and future model outputs after bias correction.
[0048] Based on the pattern outputs after different combinations of bias corrections and their corresponding weights in the test dataset, the final historical pattern output generated by the method of this invention is obtained:
[0049] (4)
[0050] In the formula, For the first The first mode uses the first The historical pattern output generated by the bias correction method This indicates the final historical pattern output generated by the method of this invention.
[0051] Equation (4) indicates that the final historical model output is a weighted sum of model outputs generated by different combinations of model-correction methods. Furthermore, different evaluation indicators are used to quantify the performance of the method of the present invention in improving the simulation accuracy of model output, and the results are compared with those of all single combinations. Similarly, the statistical relationship shown in Equation (4) is applied to the future model data generated by different combinations to generate the final future model output based on the method of the present invention.
[0052] The study area of this invention is located in the "Three Norths" region of China, including Northeast, North China, and Northwest China, encompassing 13 regions such as Xinjiang Uygur Autonomous Region and Jilin Province. Irradiance was chosen as the climate parameter for this study because it is the most important weather element affecting photovoltaic output. Hourly irradiance data for the Three Norths region of China from 1960 to 2014 were downloaded from the ERA5 website (daily irradiance data can be obtained by summing hourly irradiance data), and daily irradiance data from 16 global climate models were downloaded from the CMIP6 website. The 16 global climate models include ACCESS-CM2, ACCESS-ESM1-5, CanESM5, CESM2-WACCM, CMCC-CM2-SR5, CMCC-ESM2, FGOALS-g3, IITM-ESM, INM-CM4-8, INM-CM5-0, IPSL-CM6A-LR, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, MRI-ESM2-0, and NorESM2-LM. The historical data for CMIP6 covers 1960–2014, while the future data covers 2025–2059. The climate change scenario is SSP126. It is important to note that ERA5 irradiance is measured in J / m². 2 The unit of CMIP6 irradiance is W / m². 2 To maintain unit consistency, the irradiance units for both datasets need to be unified to W / m². 2 Given the varying spatial resolutions of the 16 global climate models, bilinear interpolation was used to resample them to the ERA5 spatial resolution of 0.25° × 0.25°. When model outputs and observational data for climate parameters were available, the time ranges for the calibration, validation, and test datasets were determined to be 1960–1989, 1990–2004, and 2005–2014, respectively. Furthermore, five simple model output bias correction methods were introduced: additive linear scaling, multiplicative linear scaling, variance scaling, quantile mapping, and quantile increment mapping. Evaluation metrics included RMSE and mean absolute error. It is worth noting that the python-cmethods package provides open-source functions for these five methods. Thus, the equation can be obtained. In , , and The values are 16, 5, 9174 and 2 respectively.
[0053] Table 1 shows the weights of the ACCESS-CM2 mode combined with the five bias correction methods. The table shows that, when correcting the output bias of the ACCESS-CM2 mode, the additive linear scaling method performs best among the five bias correction methods. Figure 1 The RMSE of different mode-correction method combinations and the method proposed in this invention were compared during the testing period. In the box plot, the middle horizontal line represents the mean of all combinations, with the two ends representing the 25th and 75th percentiles, respectively. It can be observed from the figure that the RMSE of the method proposed in this invention is significantly lower than that of a single combination. Therefore, compared to using only a single correction method to correct a single mode output, the method proposed in this invention can generate mode outputs with higher simulation accuracy.
[0054] Table 1. Weights of the five deviation correction methods combined with the ACCESS-CM2 mode in this embodiment.
[0055] Specific Implementation Method Two:
[0057] This embodiment is a computer storage medium that stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the CMIP6 mode output deviation correction method based on a multi-mode, multi-method, and multi-index set weighting.
[0058] It should be understood that the instructions include computer program products, software, or computerized methods corresponding to any method described in this invention; the instructions can be used to program computer systems or other electronic devices. Computer storage media may include readable media on which instructions are stored, and may include, but are not limited to, magnetic storage media, optical storage media; magneto-optical storage media include read-only memory (ROM), random access memory (RAM), erasable programmable memory (e.g., EPROM and EEPROM), and flash memory layers, or other types of media suitable for storing electronic instructions. Specific implementation method three:
[0060] This embodiment is a CMIP6 mode output deviation correction device based on multi-mode, multi-method, and multi-index set weighting. The device includes a processor and a memory. It should be understood that this includes any device described in this invention that includes a processor and a memory. The device may also include other units and modules that perform display, interaction, processing, control, and other functions through signals or instructions.
[0061] The memory stores at least one instruction, which is loaded and executed by the processor to implement the CMIP6 mode output deviation correction method based on a multi-mode, multi-method, and multi-index set weighting.
[0062] Those skilled in the art will understand that at least one stored instruction is a computer program product corresponding to a method or system. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.
[0063] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application, and can also be used with corresponding devices. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0064] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0066] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.
[0067] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.
[0068] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. A CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting, characterized in that, Includes the following steps: S100. Obtain CMIP6 model output data for the target area simulated by different global climate models, including historical and future climate data; at the same time, obtain an observation data dataset with the same geographical range and time span, wherein the observation data dataset is a climate reanalysis dataset or a measured dataset. S200. Divide the historical data and observation data output by the CMIP6 model into three parts: calibration dataset, validation dataset, and test dataset; S300, Based on different CMIP6 model output bias correction methods, the CMIP6 model outputs in the validation dataset, test dataset, and future dataset are generated by correcting the statistical relationship between the CMIP6 model outputs in the calibration dataset and the observation data. S400. Based on the observational data in the validation dataset and the bias-corrected CMIP6 model output, calculate the skill score for different model-correction method combinations, considering all models, all bias correction methods, all evaluation indicators, and all latitude and longitude point data. Then score based on skill Determine the weight of the CMIP6 mode output generated by different combinations in the final CMIP6 mode output. ;in, , , and These represent the pattern index, deviation correction index, latitude and longitude point index, and index, respectively. , , and These represent the number of models, the number of bias correction methods, the number of latitude and longitude points, and the number of evaluation indicators, respectively. Indicates the first The first mode uses the first The bias correction method in the first The first latitude and longitude point obtained Each evaluation indicator value; S500, the CMIP6 mode output and its corresponding weights after correction for different combinations of biases in the test dataset and future dataset. The final CMIP6 mode output is obtained.
2. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting as described in claim 1, characterized in that, The observation data dataset used is the ERA5 reanalysis dataset.
3. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting as described in claim 2, characterized in that, In step S200, before dividing the historical data and observation data output by the CMIP6 model into three parts, the units of the CMIP6 model output and the observation data are unified.
4. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting as described in claim 3, characterized in that, In the process of unifying the spatial resolution of the CMIP6 mode output and the spatial resolution of the observation data, when the spatial resolution of the CMIP6 mode output is inconsistent with the spatial resolution of the observation data, bilinear interpolation technology is used to resample the CMIP6 mode output to the spatial resolution of the observation data.
5. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting according to claim 4, characterized in that, In step S100, the time resolution of both the CMIP6 mode output data and the observation data is 1 day.
6. A CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting according to any one of claims 1 to 5, characterized in that, The weight .
7. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting as described in claim 6, characterized in that, The process of finally outputting the CMIP6 mode in S500 includes: according to Obtain the final historical CMIP6 mode output from the test dataset. In the formula For the first The first mode uses the first The historical CMIP6 mode output generated by the bias correction method.
8. The CMIP6 mode output bias correction method based on multi-mode, multi-method, and multi-index set weighting according to claim 7, characterized in that, The process of finally outputting the CMIP6 mode in S500 also includes: The final historical CMIP6 mode output is obtained by using this method. The final CMIP6 mode output in the future dataset is obtained in the same way.
9. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement the CMIP6 mode output deviation correction method based on a multi-mode, multi-method, and multi-index set weighting as described in any one of claims 1 to 8.
10. A CMIP6 mode output deviation correction device based on multi-mode, multi-method, and multi-index set weighting, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement the CMIP6 mode output deviation correction method based on a multi-mode, multi-method, and multi-index set weighting as described in any one of claims 1 to 8.