Visibility prediction method and device
By collecting and analyzing meteorological parameter data, determining the correlation and training prediction models, the problem of large visibility prediction error in the prior art is solved, and higher prediction accuracy is achieved.
Patent Information
- Application Number
- CN202411256216.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-09
- Publication Date
- 2024-12-24
AI Technical Summary
When the existing visibility prediction methods superimpose meteorological parameter prediction and visibility prediction, the prediction errors are large, making it difficult to improve the accuracy of the prediction results.
By collecting meteorological parameter data for continuous time periods, the correlation between visibility and meteorological parameters is determined, the training data set is generated, and the single-element and multi-element prediction models are trained. The multi-element model is used to predict visibility to reduce errors.
The accuracy of visibility prediction is improved, and the accuracy of input parameters is enhanced through the joint input of multi-time meteorological parameters and visibility, and the prediction error is reduced.
Smart Images
Figure CN119179123A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of visibility prediction, and in particular to a visibility prediction method and device. Background Art
[0002] At present, the research on visibility prediction methods at home and abroad mainly takes meteorological parameters as input and visibility as output. First, the future trend of meteorological parameters is predicted, and then visibility is predicted by the predicted values of each meteorological parameter. The superposition of the two predictions will increase the visibility prediction error. Therefore, how to improve the accuracy of visibility prediction is an important problem that the industry needs to solve. Summary of the invention
[0003] The present invention provides a visibility prediction method and device to reduce visibility prediction error and improve the accuracy of prediction results.
[0004] To this end, the present invention provides the following technical solutions:
[0005] A visibility prediction method, the method comprising:
[0006] Collect meteorological parameter data for a certain period of time;
[0007] Determining the correlation between visibility and meteorological parameters, and determining visibility influencing factors based on the correlation;
[0008] Selecting the meteorological parameter data according to the visibility influencing factors to generate a training data set, each data sample in the training data set is a set of meteorological parameter data at the same time;
[0009] Training a single factor prediction model according to each meteorological parameter at different times, and using the single factor prediction model to predict each meteorological parameter;
[0010] Training a multi-factor visibility prediction model using the data samples and corresponding visibility;
[0011] Visibility prediction is performed using the multi-factor visibility prediction model.
[0012] Optionally, the meteorological parameters include visibility and any one or more of the following meteorological parameters corresponding to the visibility: temperature, humidity, dew, wind speed, wind direction, air pressure, and cloud cover.
[0013] Optionally, the method further comprises:
[0014] Filling in the missing data of the meteorological parameters;
[0015] The meteorological parameter data is normalized.
[0016] Optionally, determining visibility influencing factors includes:
[0017] Divide visibility levels and uniformly code each visibility level;
[0018] Determine the correlation between various meteorological parameters and visibility;
[0019] The visibility influencing factor is determined according to the correlation.
[0020] Optionally, determining the correlation between each meteorological parameter and visibility includes: determining the correlation between each meteorological parameter and visibility according to a change trend of each meteorological parameter corresponding to different visibility levels.
[0021] Optionally, determining the visibility influencing factor according to the correlation includes: taking a parameter type with a high correlation as a visibility influencing factor.
[0022] Optionally, the single factor prediction model includes an input layer, a hidden layer, and an output layer, taking visibility, one of the meteorological parameters, as an example;
[0023] The input layer has k nodes, and the input of the k nodes is the visibility at k time points; the output layer has t nodes, and the output of the t nodes is the visibility prediction value at t moments; k≥1, t≥1.
[0024] Optionally, the multi-factor visibility prediction model includes an input layer, a hidden layer, and an output layer;
[0025] The input of the input layer includes data samples from time i to i+j-1 and the visibility corresponding to the data samples; the output of the output layer includes the visibility at each moment from time i+k-1 to i+j+t-2; i≥1, k≥1, t≥1.
[0026] A visibility prediction device, comprising:
[0027] A data collection module is used to collect meteorological parameter data within a certain continuous period of time;
[0028] An influencing factor determination module is used to determine the correlation between visibility and meteorological parameters, and determine visibility influencing factors according to the correlation;
[0029] A sample generation module, used for selecting the meteorological parameter data to generate a training data set according to the visibility influencing factors, wherein each data sample in the training data set is a set of meteorological parameter data at the same time;
[0030] The first model training module is used to train a single factor prediction model according to various meteorological parameters at different times, and use the single factor prediction model to predict various meteorological parameters;
[0031] A second model training module, used for training a multi-factor visibility prediction model using the data samples and corresponding visibility;
[0032] The visibility prediction module is used to perform visibility prediction using the multi-factor visibility prediction model.
[0033] A computer-readable storage medium stores a computer program, which executes the steps of the visibility prediction method when the computer program is executed by a processor.
[0034] The visibility prediction method and device provided by the present invention collects meteorological parameter data within a certain continuous time period, determines the correlation between visibility and meteorological parameters, and determines visibility influencing factors based on the correlation, so that meteorological parameter data with strong correlation with visibility can be selected to participate in model training, thereby improving model performance. In addition, a single-factor prediction model is trained based on each meteorological parameter at different times, and each meteorological parameter is predicted using the single-factor prediction model, thereby filling in missing data or predicting each meteorological parameter at the next moment as one of the inputs of a multi-factor visibility prediction model, and then the data sample and its corresponding visibility are used to train a multi-factor visibility prediction model, and visibility prediction is performed using the multi-factor visibility prediction model, and meteorological parameter data at multiple moments are combined with visibility as input to predict visibility at a certain moment, thereby improving the accuracy of the input parameters and thus improving the accuracy of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0036] Figure 1 is a flow chart of the visibility prediction method provided by the present invention;
[0037] Figure 2 is a schematic diagram of the correlation between three different meteorological parameter data and visibility in an embodiment of the present invention;
[0038] Figure 3 It is a structural schematic diagram of a single factor prediction model in an embodiment of the present invention;
[0039] Figure 4 is a structural schematic diagram of a multi-factor visibility prediction model in an embodiment of the present invention;
[0040] Figure 5 is a schematic diagram of a cyclic prediction process in an embodiment of the present invention;
[0041] Figure 6 It is a structural schematic diagram of the visibility prediction device provided by the present invention. DETAILED DESCRIPTION
[0042] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.
[0043] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0044] like Figure 1 FIG. 1 is a flow chart of a visibility prediction method provided by the present invention, comprising the following steps:
[0045] Step 101, collecting meteorological parameter data within a certain continuous period of time.
[0046] Since the weather in different regions is greatly affected by the environment and geographical location, when collecting data, meteorological parameter data within a specific area can be collected so that the final trained prediction model can better adapt to the area and improve the accuracy of the prediction results.
[0047] The meteorological parameter data may be obtained from some websites or a database of a meteorological department, which is not limited in the embodiment of the present invention.
[0048] The meteorological parameter data includes visibility and any one or more of the following meteorological parameters corresponding to the visibility: temperature, humidity, dew, wind speed, wind direction, air pressure, and cloud cover. In this way, meteorological parameter data corresponding to different time points can be obtained.
[0049] Furthermore, the collected data may be missing to varying degrees. If the default data is directly deleted, the continuity of the time series will be affected. Therefore, the default data needs to be filled to ensure the integrity of the data of different meteorological parameters at each moment. For example, multi-source data such as FY4A satellite fog products can be used for filling, or the missing data can be filled based on a single factor prediction model, etc., which is not limited in the embodiments of the present invention.
[0050] Furthermore, since different types of meteorological parameters have different measurement units, in order to better compare the correlation between different meteorological parameters and visibility, the collected meteorological parameter data can also be normalized, and each meteorological parameter data can be normalized to data between (0, 1). The normalization method can adopt a conventional calculation method, which is not limited in the embodiment of the present invention.
[0051] Step 102, determining the correlation between visibility and meteorological parameters, and determining visibility influencing factors according to the correlation.
[0052] For example, the visibility influencing factors can be determined as follows:
[0053] (1) Divide visibility levels and assign uniform codes to each visibility level.
[0054] For example, the visibility is divided into four levels: less than 1 km, 1-2 km, 2-4 km, and greater than 4 km. The above four levels are uniformly coded.
[0055] For example, the visibility codes corresponding to different levels are shown in Table 1 below.
[0056] Table 1
[0057] visibility Visibility Coding Less than 1km 0 1~2km 1 2~4km 2 More than 4km 3
[0058] (2) Determine the correlation between each meteorological parameter data and visibility.
[0059] For example, the correlation between each meteorological parameter and visibility is determined according to the changing trend of each meteorological parameter corresponding to different visibility levels.
[0060] Figure 2 (a), (b), and (c) show schematic diagrams of the correlation between three different meteorological observation data and visibility.
[0061] like Figure 2 As shown in (a), when the visibility is less than 1 km, the temperature changes in the range of 14 to 25 °C; when the visibility is greater than 4 km, the temperature changes in the range of 5 to 35 °C. It can be seen that a certain temperature has a certain influence on low visibility, and the correlation is good.
[0062] like Figure 2 As shown in (b), when the visibility is less than 1 km, the relative humidity varies from 86 to 99°C; when the visibility is greater than 4 km, the temperature varies from 35 to 99°C, indicating that relative humidity has a certain impact on low visibility.
[0063] like Figure 2As shown in (c), when the visibility is less than 1 km, the wind direction changes in the range of 50 to 290°C; when the visibility is greater than 4 km, the wind direction changes in the range of 5 to 350°C, indicating that the wind direction has little effect on low visibility.
[0064] (3) Determine visibility influencing factors based on the correlation.
[0065] In some embodiments, a type of meteorological parameter with a high correlation may be selected as a visibility influencing factor.
[0066] In other embodiments, considering that too many influencing factors will increase the time cost of model training, n (n≥1) meteorological parameter types with high correlation can be selected as visibility influencing factors, defined as influencing factor 1, influencing factor 2, ... influencing factor n.
[0067] Step 103, selecting the meteorological parameter data according to the visibility influencing factors to generate a training data set, each data sample in the training data set is a group of meteorological parameter data at the same time.
[0068] For example, the data sample at time i can be expressed as X 1 ,X 2 ,…,X n .
[0069] Step 104: training a single factor prediction model according to each meteorological parameter at different times, and using the single factor prediction model to predict each meteorological parameter.
[0070] For example, in a non-limiting embodiment, the single factor prediction model uses a BP (Back Propagation) neural network, which can be specifically used Figure 3 The single-factor prediction model structure shown takes visibility, one of the meteorological parameters, as an example.
[0071] Figure 3 The single factor prediction model shown includes an input layer, a hidden layer, and an output layer. Among them:
[0072] There are k (k ≥ 1) nodes in the input layer, and the input of the k nodes is the visibility at k time points;
[0073] There are t (t≥1) nodes in the output layer, and the outputs of the t nodes are visibility prediction values at t moments.
[0074] The process of building a single factor prediction model is as follows:
[0075] (1) Setting parameters
[0076] Set the number of input data groups to k;
[0077] Set t prediction moments;
[0078] The visibility at the first moment of the forecast can be obtained through other forecast products or single-factor forecast models, and the visibility at other moments (such as 1-hour intervals) is set to 0;
[0079] Set the maximum number of training cycles;
[0080] Set performance target values;
[0081] Set the learning coefficient.
[0082] (2) Determine the prediction model input layer
[0083] The visibility from the i-th to i+k-1th time is recorded as Y, then the input layer is Y i ,Y i+1 ,…,Y i+k-2 ,Y i+k-1 , where i=1,2,…,m-k+1.
[0084] The number of input layer nodes can be set to k.
[0085] (3) Determine the output layer of the prediction model
[0086] Y i+k-1 ,…,Y i+k+t-2 is the visibility from time i+k-1 to time i+k+t-2, which serves as the output layer of the prediction model.
[0087] The number of nodes in the output layer can be set to t, representing t moments.
[0088] (4) Determine the hidden layer of the prediction model
[0089] The number of hidden layer nodes can be determined using the following empirical formula:
[0090]
[0091] Where p is the number of hidden layer nodes, α is an integer whose value is [1,10]. α is tested from 1 to 10 in sequence to determine the optimal number of hidden layer nodes, that is, the optimal value of α.
[0092] (5) Establishing a single factor prediction model
[0093] The single factor prediction model can be expressed as follows:
[0094] (Y i+k-1 ,…,Y i+k+t-2 )=(Y i ,Y i+1 ,…,Y i+k-1 )
[0095] Among them, the output Y is added to the input, the input is continuously updated, multiple rounds of iterative training are implemented, and the model parameters are obtained.
[0096] The prediction range can be set to j, and the visibility at j+t-1 moments can be predicted using this single factor prediction model.
[0097] The corresponding relationship between the input parameters and output parameters of the single factor prediction model is shown in Table 2 below.
[0098] Table 2
[0099]
[0100] Using this single factor prediction model, various meteorological parameters can be predicted.
[0101] Step 105: training a multi-factor visibility prediction model using the data samples and corresponding visibility.
[0102] In a non-limiting embodiment, the multi-factor visibility prediction model can also use a BP neural network, for example, Figure 4 The structure of the multi-factor visibility prediction model is shown.
[0103] The multi-factor visibility prediction model includes an input layer, a hidden layer, and an output layer. Among them:
[0104] The input of the input layer includes data samples from time i to i+k-1 and the visibility corresponding to the data samples;
[0105] The output of the output layer includes the visibility at each time from i+k-1 to i+k+t-2, i≥1, t≥1.
[0106] The process of constructing a multi-factor visibility prediction model is as follows:
[0107] (1) Setting parameters
[0108] Set the number of input data groups to k;
[0109] Set t prediction moments;
[0110] Set the total number of samples to m;
[0111] The visibility at the first moment of the forecast is obtained through other forecast products or single-factor forecast models, and the visibility at other moments (1 hour interval) is set to 0, that is, the visibility of the m+1th sample is the prediction result of other forecast products or single-factor forecast models, and the visibility of the m+1th to m+tth samples is 0, and the total number of samples is m+t;
[0112] Set the maximum number of training cycles;
[0113] Set performance target values;
[0114] Set the learning coefficient.
[0115] (2) Determine the prediction model input layer
[0116] Multiple characteristic parameters from the i-th to i+k-1th time: influencing factor 1, influencing factor 2, ..., influencing factor n, visibility, are denoted as X 1 ,X 2 ,…,X n ,Y, then the input layer is (X 1 ,X 2 ,…,X n ,Y) i ,(X 1 ,X 2 ,…,X n ,Y) i+1 ,……,(X 1 ,X 2 ,…,X n ,Y) i+k-2 ,(X 1 ,X 2 ,…,X n ,Y) i+k-1 .
[0117] in:
[0118] (X 1 ,X 2 ,…,X n ,Y) i are influencing factor 1, influencing factor 2, ..., influencing factor n and visibility at time i;
[0119] (X 1 ,X 2 ,…,X n ,Y) i+1 Influencing factor 1, influencing factor 2, ..., influencing factor n, visibility at time i+1;
[0120] …
[0121] (X 1 ,X 2 ,…,X n ,Y) i+k-2 Influencing factor 1, influencing factor 2, ..., influencing factor n, visibility at time i+k-2;
[0122] (X 1 ,X 2 ,…,X n ,Y) i+k-1Influencing factor 1, influencing factor 2, ..., influencing factor n, visibility at time i+k-1;
[0123] i=1,2,…,m-k+1.
[0124] The number of input layer nodes can be set to k.
[0125] (3) Determine the output layer of the prediction model
[0126] Y i+k-1 ,…,Y i+k+t-2 is the visibility at the i+k-1,…,i+k+t-2 moments, which serves as the output layer of the prediction model.
[0127] The number of output layer nodes can be set to t.
[0128] (4) Determine the hidden layer of the prediction model
[0129] The number of hidden layer nodes can be calculated using the empirical formula:
[0130]
[0131] Where p is the number of hidden layer nodes, α is an integer whose value is [1,10]. α is tested from 1 to 10 in sequence to determine the optimal number of hidden layer nodes, that is, the optimal value of α.
[0132] (5) Determine the multi-factor visibility prediction model
[0133] The multi-factor visibility prediction model can be expressed as follows:
[0134]
[0135] Y=f(X,Y)
[0136] Add the output Y to the input, continuously update the input, implement multiple iterative training, and obtain the model parameters.
[0137] The prediction range can be set to j. Using this multidimensional visibility prediction model, the visibility at j+t-1 moments can be predicted each time.
[0138] The corresponding relationship between the input parameters and output parameters of the multi-factor visibility prediction model is shown in Table 3 below.
[0139] Table 3
[0140]
[0141] Step 106: Use the multi-factor visibility prediction model to perform visibility prediction.
[0142] When using the multi-factor visibility prediction model to predict visibility, the corresponding input parameters can be determined according to the number of input nodes to obtain visibility prediction values at multiple moments. Moreover, the multi-factor visibility prediction model can be used to realize cyclic prediction of visibility. The process of cyclic prediction is as follows: Figure 5 shown.
[0143] The visibility prediction method provided by the embodiment of the present invention constructs a multi-factor visibility prediction model, combines multi-time meteorological parameter data with visibility as input, predicts the visibility at a certain time, improves the accuracy of the input parameters, and further improves the accuracy of the prediction results.
[0144] Accordingly, an embodiment of the present invention further provides a visibility prediction device, such as Figure 6 Shown is a structural schematic diagram of the device.
[0145] The visibility prediction device 600 includes the following modules:
[0146] The data collection module 601 is used to collect meteorological parameter data within a certain continuous period of time;
[0147] An influencing factor determination module 602 is used to determine the correlation between visibility and meteorological parameters, and determine visibility influencing factors according to the correlation;
[0148] A sample generation module 603 is used to select the meteorological parameter data according to the visibility influencing factors to generate a training data set, where each data sample in the training data set is a set of meteorological parameter data at the same time;
[0149] The first model training module 604 is used to train a single factor prediction model according to each meteorological parameter at different times, and use the single factor prediction model to predict each meteorological parameter;
[0150] A second model training module 605 is used to train a multi-factor visibility prediction model using the data samples and corresponding visibility;
[0151] The visibility prediction module 606 is used to perform visibility prediction using the multi-factor visibility prediction model.
[0152] The specific implementation of each of the above modules can refer to the description in the previous embodiment of the method of the present invention, which will not be repeated here.
[0153] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.
[0154] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0155] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways.
[0156] The present invention also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed when it is run. Figure 1 The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The storage medium may also include a non-volatile memory (non-volatile) or a non-transitory memory, etc.
[0157] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless means.
[0158] The embodiments of the present invention are described in detail above. The present invention is described in detail using specific implementation methods herein. The description of the above embodiments is only used to help understand the method and system of the present invention. It is only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should belong to the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A visibility prediction method, characterized in that: The method comprises: Collect meteorological parameter data for a certain period of time; Determining the correlation between visibility and meteorological parameters, and determining visibility influencing factors based on the correlation; Selecting the meteorological parameter data according to the visibility influencing factors to generate a training data set, each data sample in the training data set is a set of meteorological parameter data at the same time; Training a single factor prediction model according to each meteorological parameter at different times, and using the single factor prediction model to predict each meteorological parameter; Training a multi-factor visibility prediction model using the data samples and corresponding visibility; Visibility prediction is performed using the multi-factor visibility prediction model.
2. The visibility prediction method according to claim 1, characterized in that: The meteorological parameters include visibility and any one or more of the following meteorological parameters corresponding to the visibility: temperature, humidity, dew, wind speed, wind direction, air pressure, and cloud cover.
3. The visibility prediction method according to claim 1, characterized in that: The method further comprises: Filling in the missing data of the meteorological parameters; The meteorological parameter data is normalized.
4. The visibility prediction method according to claim 1, characterized in that: The factors affecting visibility include: Divide visibility levels and uniformly code each visibility level; Determine the correlation between various meteorological parameters and visibility; The visibility influencing factor is determined according to the correlation.
5. The visibility prediction method according to claim 4, characterized in that: Determining the correlation between each meteorological parameter and visibility includes: The correlation between each meteorological parameter and visibility is determined according to the changing trend of each meteorological parameter corresponding to different visibility levels.
6. The visibility prediction method according to claim 1, characterized in that: Determining the visibility influencing factors according to the correlation includes: Parameter types with high correlation are selected as visibility influencing factors.
7. The visibility prediction method according to claim 1, characterized in that: The single factor prediction model includes an input layer, a hidden layer, and an output layer, and takes visibility, one of the meteorological parameters, as an example; The input layer has k nodes, and the input of the k nodes is the visibility at k time points; the output layer has t nodes, and the output of the t nodes is the visibility prediction value at t time points; k≥1, t≥1.
8. The visibility prediction method according to claim 1, characterized in that: The multi-factor visibility prediction model includes an input layer, a hidden layer, and an output layer; The input of the input layer includes data samples from time i to i+k-1 and the visibility corresponding to the data samples; the output of the output layer includes the visibility at each time from time i+k-1 to i+k+t-2; i≥1, k≥1, t≥1.
9. A visibility prediction device, characterized in that: The device comprises: A data collection module is used to collect meteorological parameter data within a certain continuous period of time; An influencing factor determination module is used to determine the correlation between visibility and meteorological parameters, and determine visibility influencing factors according to the correlation; A sample generation module, used for selecting the meteorological parameter data to generate a training data set according to the visibility influencing factors, wherein each data sample in the training data set is a set of meteorological parameter data at the same time; The first model training module is used to train a single factor prediction model according to various meteorological parameters at different times, and use the single factor prediction model to predict various meteorological parameters; A second model training module, used for training a multi-factor visibility prediction model using the data samples and corresponding visibility; The visibility prediction module is used to perform visibility prediction using the multi-factor visibility prediction model.
10. 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 visibility prediction method according to any one of claims 1 to 8 are performed.