Ensemble forecasting method, system, device and storage medium for data missing

Abstract

The present invention provides a data-missing ensemble forecasting method, system, device and storage medium, the method comprising: obtaining a plurality of forecast source data, and initializing the forecast weight of each forecast source data to obtain an ensemble forecasting basic data set; performing data missing detection on each forecast source data in the ensemble forecasting basic data set, and updating the forecast weight of each forecast source data according to the corresponding missing detection result to obtain a first ensemble forecasting data set; performing credibility detection on each forecast source data in the first ensemble forecasting data set, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result to obtain a second ensemble forecasting data set; performing ensemble forecasting according to the second ensemble forecasting data set to obtain a forecasting result. The present invention can not only simply and effectively avoid the problems of data missing and low credibility, but also make full use of each forecast source data to provide an effective guarantee for the reliability and accuracy of ensemble forecasting.

Description

Technical Field

[0001] The present invention relates to the technical field of ensemble forecasting, and in particular to a data-missing ensemble forecasting method, system, computer equipment and storage medium. Background Art

[0002] Ensemble forecasting refers to a multi-value forecasting method that combines multiple sets of forecast results with different credibility for the same effective forecast time into one forecast result using statistical methods or machine learning methods to reduce uncertainty and randomness and improve the reliability and accuracy of the forecast results. It is widely used in the field of meteorological research. However, the quality of each forecast data source involved in the forecast will have a direct impact on the final ensemble forecast result. That is, when one or more forecast source data is completely missing or has low credibility, it is not conducive to the calculation and processing of the ensemble forecast, which in turn affects the quality of the forecast data.

[0003] At present, in order to ensure the reliability of the ensemble forecasting method, the missing data are generally discarded, or the prediction model is constructed using historical data, and then prediction and completion processing is performed, or the sliding average method is used to fill in the data. However, in actual applications, missing data is very common in multi-source data ensemble forecasts. If the abandonment method is directly adopted, the number of available forecast sources will be reduced and the quality of ensemble forecast data will be reduced. If the filling method is adopted, the error caused by the non-forecast source itself will be introduced due to the error generated by the filling method itself, making the filled data part have a larger deviation, affecting the credibility of the ensemble forecast data, and different formats of data require different filling methods, which makes the filling process complicated and has poor replicability.

[0004] Therefore, there is an urgent need to propose a simple and effective ensemble forecasting method for data missing that can fully utilize the data from various forecast sources and effectively avoid the problems of data missing and low credibility. Summary of the invention

[0005] The purpose of the present invention is to provide an ensemble forecasting method for data missing, which solves the application defects of the ensemble forecasting method for data missing in the prior art by effectively detecting the data missing situation and credibility of each forecast source and reasonably adjusting and allocating the weight of each forecast source based on the detection result. It can simply and effectively avoid the problems of data missing and low credibility, make full use of the data of each forecast source, obtain high-quality ensemble forecast data, and provide effective guarantee for the reliability and accuracy of the ensemble forecast.

[0006] In order to achieve the above objectives, it is necessary to provide a data-missing ensemble forecasting method, system, computer device and storage medium in response to the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides an ensemble forecasting method for data missing, the method comprising the following steps:

[0008] Acquire multiple forecast source data, and initialize the forecast weight of each forecast source data to obtain an ensemble forecast basic data set; the forecast source data includes a start time, a forecast time, and forecast element values;

[0009] Performing data missing detection on each forecast source data in the ensemble forecast basic data set respectively, and updating the forecast weight of each forecast source data according to the corresponding missing detection result, so as to obtain a first ensemble forecast data set;

[0010] Performing credibility detection on each forecast source data in the first ensemble forecast data set respectively, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result, so as to obtain a second ensemble forecast data set;

[0011] Ensemble forecasting is performed according to the second ensemble forecast data set to obtain a forecast result.

[0012] Furthermore, the step of initializing the forecast weights of each forecast source data includes:

[0013] The historical forecast accuracy of each forecast source data is obtained, and the forecast weight corresponding to each forecast time of each forecast source data is initialized according to each historical forecast accuracy.

[0014] Furthermore, the step of performing data missing detection on each forecast source data in the ensemble forecast basic data set and updating the forecast weight of each forecast source data according to the corresponding missing detection result comprises:

[0015] According to the order of forecast time, traverse each forecast source data, detect whether there is a forecast element value missing in the forecast source data corresponding to each forecast time, and give up the forecast weight of the forecast source data with missing forecast element values ​​to the weight giving pool;

[0016] The total weights given up in the given up pool weights corresponding to each forecast time are counted, and the total weights given up are redistributed to each forecast source data, and the forecast weights of each forecast source data corresponding to the forecast time are updated.

[0017] Furthermore, the step of reallocating the released total weight to each forecast source data comprises:

[0018] The total weight to be given up is distributed to each forecast source data according to the forecast weight ratio of each forecast source data after the weight giving up is completed.

[0019] Further, the steps of respectively performing credibility detection on each forecast source data in the first ensemble forecast data set and updating the forecast weight of each forecast source data according to the corresponding credibility detection result include:

[0020] Calculating the forecast time lag of each forecast time of each forecast source data, and performing credibility detection according to the forecast time lag, determining the weight transfer ratio of each forecast source data, and transferring the forecast weight of each forecast source data to the weight transfer pool according to the corresponding weight transfer ratio;

[0021] The total transfer weight in the weight transfer pool corresponding to each forecast time is counted, and the total transfer weight is redistributed to each forecast source data, and the forecast weight of each forecast source data corresponding to the forecast time is updated.

[0022] Furthermore, the step of performing credibility detection according to the forecast time lag and determining the weight transfer ratio of each forecast source data includes:

[0023] According to the forecast time lag, the data credibility is obtained, and according to the data credibility, the weight transfer ratio is determined; the weight transfer ratio is expressed as:

[0024] Trans_R=1-Conf_L

[0025] In the formula,

[0026]

[0027] Among them, Trans_R, Conf_L and delay represent the weight transfer ratio, data credibility and forecast lag respectively; delay_1 and delay_2 represent the first lag threshold and the second lag threshold respectively.

[0028] Furthermore, the step of redistributing the total transfer weight to each forecast source data includes:

[0029] The total transfer weight is distributed to each forecast source data according to the forecast weight ratio of each forecast source data after the weight transfer is completed.

[0030] In a second aspect, an embodiment of the present invention provides an ensemble forecasting system for data missing, the system comprising:

[0031] A data acquisition module is used to acquire a variety of forecast source data and initialize the forecast weight of each forecast source data to obtain an ensemble forecast basic data set; the forecast source data includes a start time, a forecast time and forecast element values;

[0032] A missing detection module is used to perform data missing detection on each forecast source data in the ensemble forecast basic data set, and update the forecast weight of each forecast source data according to the corresponding missing detection result to obtain a first ensemble forecast data set;

[0033] a credible detection module, configured to perform credibility detection on each forecast source data in the first ensemble forecast data set, and update the forecast weight of each forecast source data according to the corresponding credibility detection result, so as to obtain a second ensemble forecast data set;

[0034] The ensemble forecasting module is used to perform ensemble forecasting according to the second ensemble forecasting data set to obtain a forecasting result.

[0035] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0036] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0037] The present application provides a data-missing ensemble forecasting method, system, computer device and storage medium. Through the method, a plurality of forecast source data are obtained, and the forecast weights of each forecast source data are initialized to obtain the ensemble forecast basic data set, and then the data missing detection is performed on each forecast source data in the ensemble forecast basic data set, and the forecast weights of each forecast source data are updated according to the corresponding missing detection results to obtain the first ensemble forecast data set, and then the credibility detection is performed on each forecast source data in the first ensemble forecast data set, and the forecast weights of each forecast source data are updated according to the corresponding credibility detection results to obtain the second ensemble forecast data set, and the ensemble forecast is performed according to the second ensemble forecast data set to obtain the technical solution of the forecast result. Compared with the prior art, the data-missing ensemble forecasting method reasonably adjusts and allocates the forecast weights of each forecast source based on the data missing situation and the effective detection of the credibility of each forecast source, which can not only simply and effectively avoid the problems of data missing and low credibility, but also make full use of the data of each forecast source to provide effective guarantee for the reliability and accuracy of the ensemble forecast. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Schematic diagram of an application scenario of an ensemble forecasting method for data missing according to an embodiment of the present invention;

[0039] Figure 2is a flow chart of an ensemble forecasting method for data missing in an embodiment of the present invention;

[0040] Figure 3 is a schematic diagram of the structure of an ensemble forecasting system for data missing according to an embodiment of the present invention;

[0041] Figure 4 It is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and beneficial effects of the present application clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the embodiments described below are part of the embodiments of the present invention and are only used to illustrate the present invention, but are not used to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] The data-missing ensemble forecasting method provided by the present invention takes into account the problem that the processing methods of missing data and low-reliability data in existing ensemble forecasting will reduce the quality of ensemble forecasting data, thereby causing inaccurate ensemble forecasting results. The ensemble forecasting method provided by the present invention reasonably adjusts and allocates the forecast weights of each forecasting source based on the effective detection of the data missing situation and reliability of each forecasting source, thereby obtaining a more scientific and reasonable ensemble forecasting method for combining the weights of multiple source data. The ensemble forecasting method can be applied to Figure 1 On the terminal or server shown. Among them, the terminal can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server can be implemented with an independent server or a server cluster composed of multiple servers. The server or terminal can reasonably redistribute the forecast weights of each forecast source data through the weight giving mechanism according to the ensemble forecast method for data missing of the present invention, so as to ensure that the original forecast data is fully utilized while avoiding the unreliable and inaccurate ensemble forecast caused by data missing, and solve the problem of low credibility of the forecast source through the weight transfer mechanism, enhance the role of high credibility data, weaken the adverse effects of low credibility, and obtain a more scientific and reasonable weight combination of multi-source data, so as to ensure that high-quality multi-source ensemble forecast data is finally obtained, thereby simply and effectively improving the reliability, stability and accuracy of multi-source ensemble forecast data; the following embodiments will explain the ensemble forecast method for data missing of the present invention in detail.

[0044] In one embodiment, Figure 2 As shown, a data-missing ensemble forecasting method is provided, comprising the following steps:

[0045] S11. Acquire multiple forecast source data, and initialize the forecast weights of each forecast source data to obtain an ensemble forecast basic data set; the forecast source data include a start time, a forecast time, and forecast element values; wherein, multiple forecast source data can be understood as a data sequence group from multiple sources that is used for the same forecasting business and solves the same problem, each type of forecast source data includes data such as a start time, a forecast time, element values, etc., and each data source has the same forecast time interval and the same or different rolling update cycles; correspondingly, the start time refers to the time when the current forecast is initiated in the forecast source data, and the forecast time refers to each forecast time in the forecast sequence.

[0046] The forecast weight can be understood as the weighted average coefficient of each forecast source data participating in the ensemble prediction. The forecast weight initialization here can be set arbitrarily according to the requirements in principle. However, in order to ensure the reliability and accuracy of the ensemble forecast result, this embodiment preferably performs a reasonable initialization configuration according to the historical forecast accuracy of each forecast source data. Specifically, the step of initializing the forecast weight of each forecast source data includes:

[0047] The historical forecast accuracy of each forecast source data is obtained, and the forecast weight corresponding to each forecast time of each forecast source data is initialized according to each historical forecast accuracy; that is, according to the historical forecast accuracy statistics of each forecast source data, the forecast weight value is allocated according to the principle that the higher the accuracy, the higher the forecast weight value is, and it is ensured that the sum of the forecast weight values ​​of each forecast source data is equal to 1.

[0048] The forecast source data actually obtained can be roughly divided into three categories: 1) Normal forecast data, the specified forecast time period has the latest start time, for example, the start time of a forecast source data is 2022-08-08 20:00:00, the forecast time period is 2022-08-09 00:15:00 to 2022-08-19 00:00:00, and the wind speed data (forecast element value) is [3.61, 2.95, 4.21, …, 8.56]; 2) Element missing data, there is no forecast element value in the specified forecast time period, for example, the start time of a forecast source data is 2022-08-08 20:00:00, and the forecast time period is 2022-08-09 00:15:00 to 2022-08-19 There is no data at all within 00:00:00; 3) Low credibility data, there is no latest start time in the specified forecast time period, but there are historical start times, that is, the latest one or several data are missing, and there are only data reported in the past few days. For example, the start time is 2022-08-05 20:00:00, but the forecast time period is 2022-08-09 00:15:00 to 2022-08-16 00:00:00.

[0049] When using multiple forecast source data including the above three types of data as basic data and performing ensemble forecasting according to the corresponding initial forecast weights, in principle only normal data can be used for effective ensemble forecasting. The participation of element missing data and low-credibility data will reduce the quality of the ensemble forecast. However, in order to make full use of each forecast source data as much as possible and ensure that enough basic data participate in the ensemble forecast, thereby ensuring the accuracy and universality of the ensemble forecast, this embodiment takes into account the application defects of the abnormal data processing method of directly abandoning or filling in the existing ensemble forecast, and preferably adopts the following method steps, using the weight giving mechanism and the weight transfer mechanism to adjust the forecast weights of each forecast source data to obtain high-quality multi-source ensemble forecast data.

[0050] S12, performing data missing detection on each forecast source data in the ensemble forecast basic data set respectively, and updating the forecast weight of each forecast source data according to the corresponding missing detection result, to obtain a first ensemble forecast data set;

[0051] Among them, data missing detection can be understood as detecting whether each forecast source data has a corresponding forecast element value at each forecast time within a specified forecast time period. If the forecast element value corresponding to a forecast time of a certain forecast source data exists, it means that the data of the forecast time of the forecast source data is not missing, and it can normally participate in the ensemble forecast of the forecast time. Otherwise, it means that the data of the forecast time of the forecast source data is missing, and it cannot normally participate in the ensemble forecast of the forecast time. It is necessary to abandon the forecast weight of the forecast source data at the forecast time, and the abandoned forecast weight value is reallocated to other forecast source data corresponding to the forecast time that do not have data missing situations;

[0052] Specifically, the step of performing data missing detection on each forecast source data in the ensemble forecast basic data set and updating the forecast weight of each forecast source data according to the corresponding missing detection result includes:

[0053] In the order of forecast time, traverse each forecast source data, detect whether there is a missing forecast element value in the forecast source data corresponding to each forecast time, and give up the forecast weight of the forecast source data with missing forecast element values ​​to the weight giving pool; it should be noted that after the forecast weight of the forecast source data with missing forecast element values ​​is given up to the weight giving pool, the forecast weight of the forecast source data corresponding to the forecast time will change from the initialization value to 0. For example, there are 3 forecast source data A, B, and C, and the corresponding initialization set forecast weights are 0.3, 0.2, and 0.5 respectively. When performing the data missing detection in this step, it is found that only the forecast source data A has a missing forecast element value at a certain forecast time, then the forecast weight of 0.3 of the forecast source data A is given up to the weight giving pool, and the forecast weight value of the forecast source data A at the forecast time is updated to 0, and the forecast weights of other forecast source data B and C are still the corresponding initialization values;

[0054] Counting the total weights given up in the given-up pool weights corresponding to each forecast time, and redistributing the total weights given up to each forecast source data, and updating the forecast weights of each forecast source data corresponding to the forecast time;

[0055] Among them, the total weight given up in the giving up pool weight can be understood as the sum of all initial forecast weights given up due to missing forecast element values ​​when executing the data missing detection step above. For example, if there are three forecast source data A, B, and C, the corresponding initialization set forecast weights are 0.3, 0.2, and 0.5 respectively. At a certain forecast time, only forecast source data A has missing forecast element values, then the total weight given up is 0.3. If forecast source data A and B both have missing forecast element values, then the corresponding total weight given up is 0.3+0.2=0.5.

[0056] After the total weights given up in the given up pool weights corresponding to each forecast time are obtained by the above method, the given up total weights are redistributed to the forecast source data corresponding to the forecast time that still retains the initialized forecast weights; specifically, the step of redistributing the given up total weights to each forecast source data includes:

[0057] The total weight given up is distributed to each forecast source data according to the proportion of the forecast weight of each forecast source data after the weight giving up is completed; wherein, the distribution of the proportion of the forecast weight of each forecast source data after the weight giving up is completed can be understood as the proportion of the forecast weight of each forecast source data corresponding to the forecast time in the sum of the weights not given up after the data missing detection step is executed. For example, in the above example, there are 3 forecast source data A, B, and C, and the corresponding initialization set forecast weights are 0.3, 0.2, and 0.5, respectively. At a certain forecast time, only the forecast source data A has a missing forecast element value, then the total weight given up is is 0.3, then after the corresponding weight giving up is completed, the forecast weight proportions of each forecast source data are: 0 / (0+0.2+0.5), 0.2 / (0+0.2+0.5) and 0.5 / (0+0.2+0.5). When the total weight of 0.3 is given up according to the forecast weight proportion, the forecast weights of forecast source data A, forecast source data B and forecast source data C are updated to: 0.3*0 / (0+0.2+0.5), 0.3*0.2 / (0+0.2+0.5) and 0.3*0.5 / (0+0.2+0.5) respectively;

[0058] Similarly, the forecast weights of each forecast source data at each forecast time can be updated and adjusted after data missing detection to obtain the first ensemble forecast data set with updated forecast weights and the original data of the forecast source data remaining unchanged. This can ensure that when the first ensemble forecast data set is used for ensemble forecasting, the role of the original forecast data can be fully utilized and the unreliable and inaccurate ensemble forecast caused by data missing can be avoided.

[0059] S13, respectively performing credibility detection on each forecast source data in the first ensemble forecast data set, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result, to obtain a second ensemble forecast data set; wherein the credibility detection can be understood as detecting the time reliability of the forecast element value corresponding to each forecast time of each forecast source data within the specified forecast time period according to a preset credibility detection principle, for example, if the difference between the forecast time and the start time in a certain forecast source data of a certain forecast time exceeds a first time lag threshold, the data is not completely feasible, and if it exceeds a second time lag threshold, the data is completely unreliable;

[0060] After each forecast source data at each forecast time has completed the corresponding credibility test according to the above steps, it is necessary to transfer the forecast weight of the unreliable proportion according to the credibility test result, so as to enhance the effect of high-credibility data and weaken the adverse effects of low-credibility data to ensure the scientificity and rationality of the ensemble forecast; specifically, the steps of performing credibility test on each forecast source data in the first ensemble forecast data set respectively and updating the forecast weight of each forecast source data according to the corresponding credibility test result include:

[0061] Calculating the forecast time lag of each forecast time of each forecast source data, and performing credibility detection according to the forecast time lag, determining the weight transfer ratio of each forecast source data, and transferring the forecast weight of each forecast source data to the weight transfer pool according to the corresponding weight transfer ratio;

[0062] Among them, as mentioned above, the forecast lag can be understood as the difference between the forecast time and the start time. For example, if the start time is 2022-08-05 20:00:00, the forecast lag corresponding to the forecast time of 2022-08-06 15:00:00 is 19 hours, the forecast lag corresponding to the forecast time of 2022-08-06 20:00:00 is 1 day, and the forecast lag corresponding to the forecast time of 2022-08-10 20:00:00 is 5 days; the shorter the forecast lag in the ensemble forecast, the higher the corresponding data credibility, and vice versa, the credibility gradually decreases with the increase of the forecast lag. The specific credibility evaluation criteria used can be reasonably formulated in combination with the application scenario, and the proportion of unreliable data that needs to be transferred out of the corresponding forecast weight is determined according to the actual assessed credibility. The credibility evaluation criteria are not specifically limited here;

[0063] Specifically, the step of performing credibility detection according to the forecast time lag and determining the weight transfer ratio of each forecast source data includes:

[0064] According to the forecast time lag, the data credibility is obtained, and according to the data credibility, the weight transfer ratio is determined; the weight transfer ratio is expressed as:

[0065] Trans_R=1-Conf_L

[0066] In the formula,

[0067]

[0068] Among them, Trans_R, Conf_L and delay represent the weight transfer ratio, data credibility and forecast lag respectively; delay_1 and delay_2 represent the first lag threshold and the second lag threshold respectively, which can be determined according to actual application requirements. For example, according to the characteristics of meteorological data, the forecast for the next day is credible, and the forecast within 4 days is also relatively reliable. If the forecast time is longer, it will be inaccurate. In the actual meteorological ensemble forecast application, the first lag threshold and the second lag threshold can be set to 1 day and 4 days respectively; if the forecast weights of the three forecast source data A, B and C of the first ensemble forecast data set at a certain forecast time are 0.3, 0.2 and 0.5 respectively when performing credibility detection, and the obtained credibility is 0.8, 1 and 1 respectively, it means that only the forecast source data A needs to move part of the weight to the weight transfer pool, and the corresponding weight transfer ratio is 1-0.8=0.2; it should be noted that the settings of the first lag threshold, the second lag threshold, the credibility and the weight transfer ratio here are only exemplary descriptions, and no specific restrictions are imposed on the protection scope.

[0069] The total weight transferred in the weight transfer pool corresponding to each forecast time is counted, and the total weight transferred is redistributed to each forecast source data, and the forecast weight of each forecast source data of the corresponding forecast time is updated; wherein, the total weight transferred in the weight transfer pool can be understood as the sum of all initial forecast weights removed due to forecast time lag when performing the credibility detection in the above step. For example, the forecast weights of the three forecast source data A, B and C of the first set forecast data set at a certain forecast time are 0.3, 0.2 and 0.5 respectively. If the corresponding credibility is 0.8, 1 and 1 respectively, then only the forecast source data A needs to transfer the weight 0.2*0.3, and the corresponding total weight transferred is 0.2*0.3. If the corresponding credibility is 0.8, 0.6 and 1 respectively, then the corresponding total weight given up is (1-0.8)*0.3+(1-0.6)*0.2=0.14;

[0070] After the total transfer weight in the weight transfer pool corresponding to each forecast time is obtained by the above method, the total transfer weight is redistributed to each forecast source data; wherein the step of redistributing the total transfer weight to each forecast source data includes:

[0071] The total transfer weight is distributed to each forecast source data according to the forecast weight ratio of each forecast source data after the weight transfer is completed; wherein, the forecast weight ratio distribution of each forecast source data after the weight transfer is completed can be understood as the ratio of the untransferred forecast weight of each forecast source data corresponding to the forecast time to the sum of the untransferred weights after the credibility detection step is performed. For example, in the above example, the three forecast source data A, B, and C have corresponding untransferred forecast weights of 0.8*0.3, 0.2, and 0.5 at a certain forecast time, respectively. Then, the forecast weight ratios of each forecast source data after the corresponding weight transfer are: 0.8*0.3 / (0.8*0.3+0.2+0.5), 0.2 / (0.8*0.3+0.2+ 0.5) and 0.5 / (0.8*0.3+0.2+0.5). When the total transfer weight is 0.2*0.3=0.06 according to the forecast weight ratio, the forecast weights of forecast source data A, B and C are updated as: (0.2*0.3)*(0.8*0.3) / (0.8*0.3+0.2+0.5), (0.2*0.3)*0.2 / (0.8*0.3+0.2+0.5) and (0.2*0.3)*0.5 / (0.8*0.3+0.2+0.5). For the case where there are multiple forecast source data with forecast weight transfer at the same time, the total weight calculation of the weight transfer pool and the corresponding allocation ratio calculation can be obtained by analogy.

[0072] Similarly, the forecast weights of each forecast source data at each forecast time can be updated and adjusted after credibility detection to obtain a second ensemble forecast data set with updated forecast weights and the original data of the forecast source data remaining unchanged. This can ensure that when the second ensemble forecast data set is used for ensemble forecasting, the role of high-credibility data is enhanced and the adverse effects of low-credibility data are weakened, thereby obtaining a more scientific and reasonable weight combination of multi-source data and ultimately obtaining high-quality multi-source ensemble forecast data.

[0073] S14. Perform ensemble forecasting according to the second ensemble forecast data set to obtain a forecast result; wherein the forecast result can be understood as weighted averaging of each forecast source data at each forecast time in the second ensemble forecast data set according to the corresponding forecast weight to obtain the corresponding forecast element prediction value.

[0074] The embodiment of the present application utilizes the weight giving mechanism to reasonably reallocate the weights of each forecast source, give full play to the role of the original forecast data, avoid the unreliability and inaccuracy of the ensemble forecast caused by data missing, and utilize the weight transfer mechanism to solve the problem of low credibility of the forecast source, enhance the role of high-credibility data, and weaken the adverse effects of low credibility, so as to obtain a more scientific and reasonable weight combination of multi-source data, and finally obtain a universal method for high-quality multi-source ensemble forecast data, which simply and effectively improves the reliability, stability and accuracy of the multi-source ensemble forecast data, thereby providing an effective guarantee for the reliability and accuracy of the ensemble forecast.

[0075] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0076] In one embodiment, Figure 3 As shown, a data-missing ensemble forecasting system is provided, the system comprising:

[0077] Data acquisition module 1 is used to acquire multiple forecast source data and initialize the forecast weight of each forecast source data to obtain the ensemble forecast basic data set; the forecast source data includes the start time, forecast time and forecast element value;

[0078] A missing detection module 2 is used to perform data missing detection on each forecast source data in the ensemble forecast basic data set, and update the forecast weight of each forecast source data according to the corresponding missing detection result to obtain a first ensemble forecast data set;

[0079] Credibility detection module 3, used for respectively performing credibility detection on each forecast source data in the first ensemble forecast data set, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result, so as to obtain a second ensemble forecast data set;

[0080] The ensemble forecasting module 4 is configured to perform ensemble forecasting according to the second ensemble forecasting data set to obtain a forecasting result.

[0081] For the specific definition of an ensemble forecasting system for data missing, please refer to the definition of an ensemble forecasting method for data missing mentioned above, which will not be repeated here. Each module in the above-mentioned ensemble forecasting system for data missing can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0082] Figure 4 FIG. 1 shows an internal structure diagram of a computer device in an embodiment, and the computer device may specifically be a terminal or a server. Figure 4 As shown, the computer device includes a processor, a memory, a network interface, a display and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a data-missing ensemble forecasting method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0083] It can be understood by those skilled in the art that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0084] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the above method are implemented when the processor executes the computer program.

[0085] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0086] In summary, the embodiments of the present invention provide a data-missing ensemble forecasting method, system, computer device and storage medium. The data-missing ensemble forecasting method realizes the acquisition of multiple forecast source data, and after initializing the forecast weights of each forecast source data to obtain the ensemble forecast basic data set, data missing detection is performed on each forecast source data in the ensemble forecast basic data set, and according to the corresponding missing detection results, the forecast weights of each forecast source data are updated to obtain a first ensemble forecast data set, and then credibility detection is performed on each forecast source data in the first ensemble forecast data set, and according to the corresponding credibility detection results, the forecast weights of each forecast source data are updated to obtain a second ensemble forecast data set, and ensemble forecasting is performed according to the second ensemble forecast data set to obtain a forecast result. The technical solution, based on the data missing situation of each forecast source and the effective detection of credibility, the method reasonably adjusts and allocates the forecast weights of each forecast source, which can not only simply and effectively avoid the problems of data missing and low credibility, but also make full use of the data of each forecast source to provide effective guarantee for the reliability and accuracy of the ensemble forecast.

[0087] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A data-missing ensemble forecasting method, characterized in that: The method comprises the following steps: Acquire multiple forecast source data, and initialize the forecast weight of each forecast source data to obtain an ensemble forecast basic data set; the forecast source data includes a start time, a forecast time, and forecast element values; Performing data missing detection on each forecast source data in the ensemble forecast basic data set respectively, and updating the forecast weight of each forecast source data according to the corresponding missing detection result, so as to obtain a first ensemble forecast data set; Performing credibility detection on each forecast source data in the first ensemble forecast data set respectively, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result, so as to obtain a second ensemble forecast data set; performing ensemble forecasting according to the second ensemble forecast data set to obtain a forecast result; The step of performing data missing detection on each forecast source data in the ensemble forecast basic data set and updating the forecast weight of each forecast source data according to the corresponding missing detection result comprises: According to the order of forecast time, traverse each forecast source data, detect whether there is a forecast element value missing in the forecast source data corresponding to each forecast time, and give up the forecast weight of the forecast source data with missing forecast element values ​​to the weight giving pool; Counting the total weights given up in the given-up pool weights corresponding to each forecast time, and redistributing the total weights given up to each forecast source data, and updating the forecast weights of each forecast source data corresponding to the forecast time; The step of performing credibility detection on each forecast source data in the first ensemble forecast data set respectively and updating the forecast weight of each forecast source data according to the corresponding credibility detection result comprises: Calculating the forecast time lag of each forecast time of each forecast source data, and performing credibility detection according to the forecast time lag, determining the weight transfer ratio of each forecast source data, and transferring the forecast weight of each forecast source data to the weight transfer pool according to the corresponding weight transfer ratio; Counting the total weight transferred in the weight transfer pool corresponding to each forecast time, and redistributing the total weight transferred to each forecast source data, and updating the forecast weight of each forecast source data corresponding to the forecast time; The step of performing credibility detection according to the forecast time lag to determine the weight transfer ratio of each forecast source data comprises: According to the forecast time lag, the data credibility is obtained, and according to the data credibility, the weight transfer ratio is determined; the weight transfer ratio is expressed as: In the formula, in, , and They represent weight transfer ratio, data credibility and forecast lag respectively; and represent the first hysteresis threshold and the second hysteresis threshold respectively.

2. The data-missing ensemble forecasting method according to claim 1, characterized in that: The step of initializing the forecast weights of each forecast source data comprises: The historical forecast accuracy of each forecast source data is obtained, and the forecast weight corresponding to each forecast time of each forecast source data is initialized according to each historical forecast accuracy.

3. The data-missing ensemble forecasting method according to claim 1, characterized in that: The step of reallocating the given up total weight to each forecast source data comprises: The total weight to be given up is distributed to each forecast source data according to the forecast weight ratio of each forecast source data after the weight giving up is completed.

4. The data-missing ensemble forecasting method according to claim 1, characterized in that: The step of reallocating the total transfer weight to each forecast source data comprises: The total transfer weight is distributed to each forecast source data according to the forecast weight ratio of each forecast source data after the weight transfer is completed.

5. A data-missing ensemble forecasting system, characterized in that: The system comprises: A data acquisition module is used to acquire a variety of forecast source data and initialize the forecast weight of each forecast source data to obtain an ensemble forecast basic data set; the forecast source data includes a start time, a forecast time and forecast element values; A missing detection module is used to perform data missing detection on each forecast source data in the ensemble forecast basic data set, and update the forecast weight of each forecast source data according to the corresponding missing detection result to obtain a first ensemble forecast data set; a credible detection module, configured to perform credibility detection on each forecast source data in the first ensemble forecast data set, and update the forecast weight of each forecast source data according to the corresponding credibility detection result, so as to obtain a second ensemble forecast data set; an ensemble forecasting module, configured to perform ensemble forecasting according to the second ensemble forecasting data set to obtain a forecasting result; The step of performing data missing detection on each forecast source data in the ensemble forecast basic data set respectively and updating the forecast weight of each forecast source data according to the corresponding missing detection result comprises: According to the order of forecast time, traverse each forecast source data, detect whether there is a forecast element value missing in the forecast source data corresponding to each forecast time, and give up the forecast weight of the forecast source data with missing forecast element values ​​to the weight giving pool; Counting the total weights given up in the given-up pool weights corresponding to each forecast time, and redistributing the total weights given up to each forecast source data, and updating the forecast weights of each forecast source data corresponding to the forecast time; The performing credibility detection on each forecast source data in the first ensemble forecast data set respectively, and updating the forecast weight of each forecast source data according to the corresponding credibility detection result, comprises: Calculating the forecast time lag of each forecast time of each forecast source data, and performing credibility detection according to the forecast time lag, determining the weight transfer ratio of each forecast source data, and transferring the forecast weight of each forecast source data to the weight transfer pool according to the corresponding weight transfer ratio; Counting the total weight transferred in the weight transfer pool corresponding to each forecast time, and redistributing the total weight transferred to each forecast source data, and updating the forecast weight of each forecast source data corresponding to the forecast time; The step of performing credibility detection according to the forecast time lag to determine the weight transfer ratio of each forecast source data comprises: According to the forecast time lag, the data credibility is obtained, and according to the data credibility, the weight transfer ratio is determined; the weight transfer ratio is expressed as: In the formula, in, , and represent weight transfer ratio, data credibility and forecast lag respectively; and represent the first hysteresis threshold and the second hysteresis threshold respectively.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images