A multi-model combined hydrological forecasting method, system and computer storage medium

Abstract

The application discloses a kind of multi-model combination hydrological prediction method, system and computer storage medium, method includes: basic hydrological model screening, basic prediction sequence acquisition, multi-element decision support information selection, basic prediction sequence process similarity calculation, basic prediction sequence value deviation calculation, basic prediction sequence comprehensive weight calculation and basic prediction sequence weighted combination.The application evaluates and selects qualified basic hydrological model from multiple angles and obtains prediction sequence, selects multi-element decision support information from historical hydrological time series and fuses with prediction sequence to determine the weight of different basic prediction sequences, finally weighted combination multiple basic prediction sequences to obtain hydrological time series multi-model combination prediction result, by incorporating key decision support information in the process of determining multi-model combination prediction scheme, scientific and reliable hydrological time series multi-model combination prediction is realized.

Description

Technical Field

[0001] This invention relates to the field of hydrological forecasting technology, and in particular to a multi-model combined hydrological forecasting method, system, and computer storage medium. Background Technology

[0002] Hydrological forecasting is a fundamental challenge in hydrology. Numerous hydrological models are used in hydrological forecasting, each utilizing different data and providing valuable information from different perspectives. These models are often complementary. To improve the accuracy of hydrological forecasts, employing a multi-model combination approach can effectively avoid the risk of decision-making errors due to inappropriate model selection and also significantly enhance the stability of hydrological time series forecasts. Currently, multi-model combinations are often based directly on the arithmetic mean or weighted average of the predictions from multiple models, or by inputting the predictions from different models into a pre-constructed combined model for hydrological time series multi-model combined forecasting.

[0003] However, current methods of using multiple models in hydrological forecasting do not involve screening the basic hydrological models, posing a significant risk of introducing unreasonable or inapplicable basic hydrological models. Furthermore, there is a lack of decision-making support regarding which basic hydrological models to use in multi-model forecasting, severely impacting the effectiveness of hydrological forecasts. Summary of the Invention

[0004] This invention provides a multi-model combined hydrological forecasting method, system, and computer storage medium to solve the problems in the prior art.

[0005] On one hand, embodiments of the present invention provide a multi-model combined hydrological forecasting method, including:

[0006] By inputting forecast factors into multiple basic hydrological models, the corresponding basic forecast sequences are obtained.

[0007] Determine the hydrological and meteorological similarity years for the forecast period;

[0008] The process similarity between the average of the same period in hydrologically similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence is determined, and the corresponding similarity matrix is ​​obtained.

[0009] Determine the deviation of the values ​​of the average values ​​of the same period in hydrologically similar years, the average values ​​of the same period in meteorologically similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, and obtain the corresponding deviation matrix.

[0010] The comprehensive weights of the basic prediction sequences are determined based on the similarity matrix and the deviation matrix;

[0011] The basic prediction sequences are weighted and combined according to their weights to obtain the prediction results.

[0012] On the other hand, embodiments of the present invention provide a multi-model combined hydrological forecasting system, including:

[0013] The sequence prediction module is used to input forecast factors into multiple basic hydrological models to obtain the corresponding basic prediction sequences.

[0014] The similarity year determination module is used to determine the hydrological and meteorological similarity years for the forecast period;

[0015] The similarity determination module is used to determine the process similarity between the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence, and to obtain the corresponding similarity matrix.

[0016] The deviation determination module is used to determine the deviation of the values ​​of the average values ​​of the same period in hydrological similar years, the average values ​​of the same period in meteorological similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, and to obtain the corresponding deviation matrix.

[0017] The weight determination module is used to determine the comprehensive weight of the basic prediction sequence based on the similarity matrix and the deviation matrix;

[0018] The combined forecast module is used to perform weighted combination of basic forecast sequences according to weights to obtain forecast results.

[0019] On the other hand, embodiments of the present invention provide a computer storage medium storing a plurality of computer instructions, which are used to cause a computer to execute the above-described method.

[0020] The multi-model combined hydrological forecasting method, system, and computer storage medium of this invention have the following advantages:

[0021] 1. By simultaneously considering process similarity and magnitude deviation, decision support information is incorporated into the multi-model combined forecasting process of hydrological time series, resulting in more accurate, reliable, and stable forecast results.

[0022] 2. It is simple to implement and relies on relatively mature technologies.

[0023] 3. It can eliminate unqualified or inapplicable basic hydrological models.

[0024] 4. The weights of the multi-model combination change with the forecast period, realizing dynamic weight determination. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a multi-model combined hydrological forecasting method provided in an embodiment of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Before introducing the technical solution of this invention, some terms need to be explained.

[0029] Hydrological time series: The process by which hydrological elements change over time. Hydrological elements include floods, dry season runoff, and medium- and long-term runoff.

[0030] Hydrological models: Models used to describe the formation process of hydrological elements and to carry out hydrological time series forecasting.

[0031] Basic hydrological model: In multi-model combined forecasting scenarios, a single hydrological model is referred to as the basic hydrological model.

[0032] Multi-model combined forecasting: Combining multiple basic forecast sequences to obtain hydrological time series forecast results.

[0033] Multi-source information fusion: "Multi" refers to category, and "multi-source" refers to multiple categories of information. In this invention, it refers to historical hydrological time series information of different categories.

[0034] The average for the same period refers to the average value of historical data corresponding to the forecast period. For example, if the forecast period is from January to December 2022, the average for the same period over many years refers to the hydrological time series composed of the average value of January, the average value of February, ..., and the average value of December of all years from 2021 to the present.

[0035] Decision support: In this invention, after the hydrological model obtains the hydrological time series prediction results and before publishing the report, it is necessary to judge the reliability of the results and decide whether to publish the prediction results. This judgment and decision process is the decision support process; the information that can assist in the "judgment" and "decision" is decision support information.

[0036] Figure 1 A flowchart illustrating a multi-model combined hydrological forecasting method provided in an embodiment of the present invention. The embodiment of the present invention provides a multi-model combined hydrological forecasting method, including:

[0037] S100 inputs forecast factors into multiple basic hydrological models to obtain the corresponding basic forecast sequences.

[0038] For example, before inputting the forecast factors into multiple basic hydrological models in S100, the present invention further includes: screening the multiple basic hydrological models and retaining the basic hydrological models that meet the comprehensive evaluation indicators.

[0039] Specifically, this invention establishes a multi-dimensional evaluation system from the following three perspectives to evaluate and select qualified basic hydrological models for multi-model combination forecasting of hydrological time series: (1) whether it can characterize the information contained in the hydrological time series, (2) whether it can accurately predict the hydrological time series, and (3) whether it can reliably generalize from the verification period to the test period.

[0040] To facilitate comparison of the prediction performance of different basic hydrological models, the normalized Nash-Sutcliffe efficiency coefficient (NNSE), normalized root mean square error (NRMSE), and generalization error (GE) were used as multivariate evaluation indicators to evaluate and select the basic hydrological models. The calculation formulas are as follows:

[0041]

[0042]

[0043]

[0044] in, This represents the measured hydrological time series. This indicates a predicted hydrological time series. The values ​​represent the average of the measured hydrological time series, with subscripts C and T indicating the validation period and testing period, respectively. The values ​​of NNSE, NRMSE, and GE range from [0,1]. A higher NNSE value indicates that the basic hydrological model better expresses the information contained in the hydrological time series, while lower NRMSE and GE indicate higher accuracy and generalization ability of the basic hydrological model. After determining the specific values ​​of the above three multivariate evaluation indicators, this invention further screens the basic hydrological model based on the comprehensive evaluation index BMSI, calculated using the following formula:

[0045]

[0046] In this invention, basic hydrological models with BMSI > 0.5 are considered qualified models, and the selected basic hydrological models can be added to the subsequent hydrological time series multi-model combination forecast.

[0047] S110 determines the hydrological and meteorological similarity years for the forecast period.

[0048] For example, S110 specifically includes: determining the trajectory similarity between the historical hydrological sequence of each year and the historical hydrological sequence of the reference year; taking the year following the year with the highest trajectory similarity between the historical hydrological sequence and the historical hydrological sequence of the reference year as the hydrological similar year; determining the trajectory similarity between the historical meteorological sequence of each year and the historical meteorological sequence of the reference year; taking the year following the year with the highest trajectory similarity between the historical meteorological sequence and the historical meteorological sequence of the reference year as the meteorological similar year.

[0049] When selecting a reference year, the year preceding the forecast period can be used. To determine hydrological and meteorological similarity years, one of the following methods can be used to determine trajectory similarity: Hausdorff distance, edit distance, Fréchet distance, one-way distance, and multi-line positional distance.

[0050] Let's take the Hausdorff distance as an example. The Hausdorff distance is the maximum distance between the closest points of two trajectory lines, and the formula is:

[0051] d H (tr1,tr2)=max{h(tr1,tr2),h(tr2,tr1)}

[0052] Where h(tr1,tr2) is the one-way Hausdorff distance from trajectory line tr1 to trajectory line tr2, and h(tr2,tr1) is the one-way Hausdorff distance from trajectory line tr2 to trajectory line tr1. They are defined as follows:

[0053]

[0054]

[0055] Where p and q are the trajectory points of trajectory lines tr1 and tr2, respectively. The smaller the Hausdorff distance value, the more similar the trajectory lines tr1 and tr2 are.

[0056] After determining the hydrological and meteorological similar years, the corresponding historical hydrological time series can be determined. The historical hydrological time series provided in this embodiment of the invention include the average of the same period in hydrologically similar years, the average of the same period in meteorologically similar years, the multi-year average of the same period, and the average of the same period over the previous 5 years. These historical hydrological time series can serve as multivariate decision support information. The multi-year average of the same period can be obtained by calculating the multi-year average of the same period during the forecast period, while the average of the same period over the previous 5 years can be obtained by calculating the multi-year average of the same period over the previous 5 years.

[0057] S120 determines the process similarity between the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence, and obtains the corresponding similarity matrix.

[0058] For example, one of the following methods can be used to determine the process similarity between the average of hydrologically similar years, the average of meteorologically similar years, the average of multiple years, and the average of the previous 5 years and the basic prediction sequence, thereby obtaining a similarity matrix.

[0059] S130, determine the deviation of the values ​​of the average values ​​of the same period in hydrological similar years, the average values ​​of the same period in meteorological similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, and obtain the corresponding deviation matrix.

[0060] For example, one of the following can be used to determine the deviation of the values ​​of the average values ​​of the same period in hydrological similar years, the average values ​​of the same period in meteorological similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, thereby obtaining a deviation matrix.

[0061] Taking the normalized root mean square deviation as an example, the calculation formula is as follows:

[0062]

[0063] in, This refers to historical hydrological time series, such as the average of the same period in similar hydrological years, the average of the same period in similar meteorological years, the average of the same period over many years, and the average of the same period over the previous 5 years. This is the basic prediction sequence.

[0064] S140, determine the comprehensive weight of the basic prediction sequence based on the similarity matrix and the deviation matrix.

[0065] For example, S140 specifically includes: determining the corresponding basic weights using the entropy weight method based on the similarity matrix and the deviation matrix respectively; and combining the two basic weights to obtain the comprehensive weight.

[0066] The idea behind calculating weights using the entropy weight method is to use the similarity matrix or deviation matrix as the original matrix for weight determination:

[0067]

[0068] Where J represents the number of basic hydrological models obtained through screening, and K represents the amount of decision support information.

[0069] First, the original matrix R is normalized, and the normalized value p of the j-th basic hydrological model and the k-th decision support information is obtained. jk The calculation formula is as follows:

[0070]

[0071] Then calculate the moisture value e of the j-th basic hydrological model. j :

[0072]

[0073] Finally, the weight w of the j-th basic hydrological model is calculated. ij :

[0074]

[0075] i represents the weighting reorganization number, where i = 1, 2 in this invention. The formula for calculating the comprehensive weight of the j-th basic hydrological model is:

[0076]

[0077] I represents the number of weighted recombinations in the basic prediction sequence; in this invention, I = 2.

[0078] S150: The basic prediction sequences are weighted and combined according to their weights to obtain the prediction results.

[0079] For example, the formula for calculating the weighted combination is as follows:

[0080]

[0081] in, This is a multi-model combined forecast result for hydrological time series. Let j be the j-th basic prediction sequence.

[0082] This invention also provides a multi-model combined hydrological forecasting system, which includes:

[0083] The sequence prediction module is used to input forecast factors into multiple basic hydrological models to obtain the corresponding basic prediction sequences.

[0084] The similarity year determination module is used to determine the hydrological and meteorological similarity years for the forecast period;

[0085] The similarity determination module is used to determine the process similarity between the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence, and to obtain the corresponding similarity matrix.

[0086] The deviation determination module is used to determine the deviation of the values ​​of the average values ​​of the same period in hydrological similar years, the average values ​​of the same period in meteorological similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, and to obtain the corresponding deviation matrix.

[0087] The weight determination module is used to determine the comprehensive weight of the basic prediction sequence based on the similarity matrix and the deviation matrix;

[0088] The combined forecast module is used to perform weighted combination of basic forecast sequences according to weights to obtain forecast results.

[0089] This invention also provides a computer storage medium storing a plurality of computer instructions for causing a computer to execute the above-described method.

[0090] Experimental instructions

[0091] Basic hydrological models include process-driven models and data-driven models, such as the European Hydrological System Model (SHE), the two-parameter monthly water balance model (MWB), the gradient iterative decision tree (GBRT), the temporal convolutional neural network (TCN), the support vector machine (SVR), and the deep confidence neural network (DBN). Each model utilizes different forecasting factors. An experiment was conducted to predict the monthly runoff of a specific hydrological section in a certain watershed in 2022 using a multi-model combination.

[0092] Step 1: Using three evaluation indicators—NNSE, NRMSE, and GE—and the basic hydrological model comprehensive evaluation index BMSI, the basic hydrological models MWB, GBRT, TCN, SVR, and DBN were screened. The evaluation results are shown in Table 1. Based on whether the BMSI is greater than 0.5, the final basic hydrological models selected are GBRT, TCN, and DBN.

[0093] Table 1 Evaluation and Screening of Basic Hydrological Models

[0094]

[0095] Step 2: Using the basic hydrological models GBRT, TCN, and DBN obtained in Step 1, the monthly runoff of the target hydrological section during the forecast period of 2022 was predicted. The results are shown in Table 2.

[0096] Table 2. Monthly runoff forecast results for target hydrological sections in 2022

[0097]

[0098] Step 3: Select the average of the same period in hydrologically similar years, the average of the same period in meteorologically similar years, the average of the same period over many years, and the average of the same period in the previous 5 years as decision support information.

[0099] The Hausdorff distance was used to determine the hydrological and meteorological similarity years for the forecast period of the target hydrological section in 2022. The monthly runoff sequence of the target hydrological section in 2021 was selected as the reference sequence. The Hausdorff distance between the monthly runoff process of the target hydrological section in 2021 and the monthly runoff process of each year since the observation records began before 2020 was calculated and the results were arranged in ascending order, as shown in Table 3.

[0100] Table 3. Results of hydrological similarity year calculations based on Hausdorff distance.

[0101]

[0102] In this embodiment, only the most similar hydrological year is selected. If the Hausdorff distance is close, multiple hydrologically similar years can be selected, and the average of the same period of hydrologically similar years is calculated as decision support information. Table 3 shows that 2003 is most similar to the reference year 2021. Therefore, 2004 is selected as the hydrologically similar year for the target hydrological section forecast period of 2022, and the monthly runoff of the target hydrological section in 2004 is the average of the same period of hydrologically similar years.

[0103] The monthly precipitation of the target hydrological section's control basin in 2021 was selected as a reference series. The Hausdorff distances of the monthly precipitation in the target hydrological section's control basin in 2021 and the monthly precipitation events of each year since records began before 2020 were calculated and sorted in ascending order (the calculation process is similar to the selection of hydrologically similar years). From this, the meteorologically similar year for the forecast period of 2022 was selected as 2011. Therefore, the monthly runoff of the target hydrological section in 2011 is the average for the same period of meteorologically similar years. Note that the hydrologically similar year and the meteorologically similar year may be the same; in this case, both are considered together.

[0104] The multi-year average for the same period is the average runoff for the corresponding month since records began in 2021 and earlier. For example, the runoff for January is the average of the runoff for January in all years. The average for the same period over the previous 5 years is the average runoff for the corresponding month from 2017 to 2021. The final decision support information is shown in Table 4.

[0105] Table 4. Decision Support Information for Target Hydrological Section Forecast Period

[0106]

[0107] Step 4: Use Hausdorff distance to calculate the process similarity between the basic prediction sequence obtained in Step 2 and the decision support information, and obtain the similarity matrix, as shown in the following formula.

[0108]

[0109] Step 5: Use the normalized root mean square error to calculate the deviation between the basic prediction sequence obtained in Step 2 and the decision support information, and obtain the deviation matrix, as shown in the following formula.

[0110]

[0111] Step 6: Calculate the weights of the basic prediction sequences using the entropy weight method based on the similarity matrix and the deviation matrix. The weights of the basic hydrological model obtained based on the similarity matrix are:

[0112] w 1j =[0.39 0.24 0.37]

[0113] The weights of the basic hydrological model obtained based on the deviation matrix are:

[0114] w 2j =[0.42 0.20 0.38]

[0115] Based on the above two sets of weights, the comprehensive weight of the basic hydrological model is obtained as follows:

[0116] w = [0.405 0.22 0.375]

[0117] Step 7: Using the comprehensive weights obtained in Step 6, the basic prediction sequences obtained in Step 2 (see Table 2) are weighted and combined to obtain the multi-model combined forecast results for the target hydrological section in 2022, as shown in Table 5.

[0118] Table 5 shows the multi-model combined forecast results for the target hydrological sections in 2022.

[0119]

[0120] Although preferred embodiments of the invention 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 the invention.

[0121] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-model combined hydrological forecasting method, characterized in that, include: By inputting forecast factors into multiple basic hydrological models, the corresponding basic forecast sequences are obtained. Determine the hydrological and meteorological similarity years for the forecast period; The process similarity between the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence is determined, and the corresponding similarity matrix is ​​obtained. Determine the deviation of the values ​​of the average values ​​of the same period in hydrologically similar years, the average values ​​of the same period in meteorologically similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence, and obtain the corresponding deviation matrix. The comprehensive weight of the basic prediction sequence is determined based on the similarity matrix and the deviation matrix. The basic prediction sequences are weighted and combined according to the weights to obtain the prediction results. The step of determining the comprehensive weight of the basic prediction sequence based on the similarity matrix and the deviation matrix includes: Based on the similarity matrix and the deviation matrix, the corresponding basic weights are determined using the entropy weight method. The two basic weights are combined to obtain the combined weight; Before inputting the forecast factors into multiple basic hydrological models, the following steps are also included: The basic hydrological models described above are screened, and those that meet the comprehensive evaluation indicators are retained. When selecting the basic hydrological model, normalized Nash was used. The Sutcliffe efficiency coefficient (NNSE), normalized root mean square error (NRMSE), and generalization error (GE) are used as multivariate evaluation indicators. Basic hydrological models with a comprehensive evaluation index (BMSI) > 0.5 are considered qualified models. The subscript T indicates the testing period; in, This represents the measured hydrological time series. This indicates a predicted hydrological time series. This represents the average value of the measured hydrological time series, and the subscript C indicates the verification period.

2. The multi-model combined hydrological forecasting method according to claim 1, characterized in that, The determination of the hydrological similarity year and meteorological similarity year for the forecast period includes: Determine the similarity of the trajectory lines between the historical hydrological sequences of each year and the historical hydrological sequences of the reference year; The year following the year with the highest similarity between the trajectory of the historical hydrological sequence and the historical hydrological sequence of the reference year is taken as the hydrological similar year. Determine the similarity of the trajectory lines between the historical meteorological sequences of each year and the historical meteorological sequences of the reference year; The year following the year with the highest similarity between the historical meteorological sequence and the historical meteorological sequence of the reference year is taken as the meteorological similarity year.

3. The multi-model combined hydrological forecasting method according to claim 2, characterized in that, The similarity of trajectory lines is determined by one of the following: Hausdorff distance, edit distance, Fréchet distance, one-way distance, and multi-line position distance.

4. The multi-model combined hydrological forecasting method according to claim 1, characterized in that, The process of determining the similarity between the average of hydrologically similar years, the average of meteorologically similar years, the average of the same period over many years, and the average of the same period over the previous 5 years and the basic prediction sequence, and obtaining the corresponding similarity matrix, includes: The similarity matrix is ​​obtained by determining the process similarity between the average of hydrological similar years, the average of meteorological similar years, the average of multiple years, and the average of the previous 5 years and the basic prediction sequence using one of the following methods: Hausdorff distance, edit distance, Fréchet distance, one-way distance, and multi-line location distance.

5. The multi-model combined hydrological forecasting method according to claim 1, characterized in that, The determination of the deviations between the average values ​​of the same period in hydrologically similar years, the average values ​​of the same period in meteorologically similar years, the average values ​​of the same period over many years, and the average values ​​of the same period over the previous 5 years and the basic prediction sequence, to obtain the corresponding deviation matrix, includes: The deviation matrix is ​​obtained by using one of the following methods: normalized root mean square deviation, mean absolute deviation, and mean absolute deviation percentage, to determine the magnitude deviation of the average values ​​of the same period in hydrological similar years, the average values ​​of the same period in meteorological similar years, the average values ​​of the same period in multiple years, and the average values ​​of the same period in the previous 5 years from the basic prediction sequence.

6. A multi-model combined hydrological forecasting system, wherein the system applies the method of claim 1, characterized in that, include: The sequence prediction module is used to input forecast factors into multiple basic hydrological models to obtain the corresponding basic prediction sequences. The similarity year determination module is used to determine the hydrological and meteorological similarity years for the forecast period; The similarity determination module is used to determine the process similarity between the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years and the basic prediction sequence, and to obtain the corresponding similarity matrix. The deviation determination module is used to determine the deviation of the values ​​of the average of the same period in hydrological similar years, the average of the same period in meteorological similar years, the average of the same period in multiple years, and the average of the same period in the previous 5 years from the basic prediction sequence, and to obtain the corresponding deviation matrix. The weight determination module is used to determine the comprehensive weight of the basic prediction sequence based on the similarity matrix and the deviation matrix; The combined forecast module is used to perform weighted combination of the basic prediction sequences according to the weights to obtain the forecast results.

7. A computer storage medium, characterized in that, The computer storage medium stores a plurality of computer instructions, which are used to cause the computer to execute claim 1.

5. The method described in any one of the above.

Images