CN116011561A - Information extrapolation method, device, equipment and storage medium based on neural network - Google Patents
Information extrapolation method, device, equipment and storage medium based on neural network Download PDFInfo
- Publication number
- CN116011561A CN116011561A CN202310310433.6A CN202310310433A CN116011561A CN 116011561 A CN116011561 A CN 116011561A CN 202310310433 A CN202310310433 A CN 202310310433A CN 116011561 A CN116011561 A CN 116011561A
- Authority
- CN
- China
- Prior art keywords
- data
- information
- neural network
- training
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The invention discloses an information extrapolation method, device, equipment and storage medium based on a neural network, wherein the method is characterized in that an atmospheric information history sequence is obtained, and missing time point data in the atmospheric information history sequence is supplemented by using a K approach algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the speed and the efficiency of information extrapolation based on a neural network are improved.
Description
Technical Field
The present invention relates to the field of high-precision positioning technologies, and in particular, to an information extrapolation method, apparatus, device and storage medium based on a neural network.
Background
In the current stage, in the process of single-frequency precise single-point positioning (Precise Point Positioning, PPP), whether an atmospheric delay correction parameter is accurate has a critical influence on positioning accuracy and convergence time, one high-accuracy atmospheric delay correction can improve the positioning accuracy from decimeter level to centimeter level, and shorten the convergence time to minute level, in the current stage, the weather delay processing method in the single-frequency PPP mainly utilizes a physical or mathematical model to estimate or directly uses the data of the last epoch as an initial value of the current epoch to calculate, but the atmospheric change is rapid and complex, the change condition of the atmospheric delay processing method cannot be accurately described by utilizing the simple physical or mathematical model, a large amount of calculation load is increased by the complex model, positioning timeliness is influenced, and the influence of the atmospheric change is ignored when the last epoch data is directly used for calculation, so that the atmospheric delay initial value meeting the positioning accuracy cannot be obtained.
Disclosure of Invention
The invention mainly aims to provide an information extrapolation method, device, equipment and storage medium based on a neural network, and aims to solve the technical problems that in the prior art, under the condition of complex atmospheric change, the change condition cannot be accurately described by using a simple physical or mathematical model, a complex model can increase a large amount of calculation load, and the positioning timeliness and the positioning precision are poor.
In a first aspect, the present invention provides a neural network-based information extrapolation method, including the steps of:
acquiring an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result;
historical weather information is acquired, the historical weather information is combined with target data received through broadcasting to obtain historical forecast information, the historical forecast information is input into the preset neural network model, output information is obtained, and extrapolation is conducted on the output information.
Optionally, the acquiring the atmospheric information history sequence and using a K-nearest neighbor algorithm to complement missing time point data in the atmospheric information history sequence includes:
acquiring an atmospheric information history sequence, and acquiring historical atmospheric delay data around a point to be interpolated from the atmospheric information history sequence;
and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
Optionally, calculating interpolation data corresponding to the historical atmospheric delay data by using a K-nearest neighbor algorithm, and filling missing time point data in the atmospheric information historical sequence according to the interpolation data, including:
acquiring known point data around each point to be interpolated in the historical atmospheric delay data;
calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated;
and supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
Optionally, the supplementing missing time point data in the atmospheric information history sequence according to the interpolation data includes:
and interpolating the missing values near the existing data according to the interpolation data, and recursively filling the missing whole area and time point data of the time period in the atmospheric information history sequence after interpolation is completed.
Optionally, the performing variance weighting on the aligned sequence to obtain variance data, inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result, including:
Acquiring a historical weather information estimation result, and carrying out variance weighting on the sequence according to the time and the resolution progress according to the historical weather information estimation result;
removing the data wild value in the fixed weight sequence to obtain variance data;
and inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
Optionally, the removing the data wild value in the weighted sequence to obtain variance data includes:
calculating a wild value judgment threshold according to the preset confidence and the preset degree of freedom, determining a data wild value in the weight sequence according to the wild value judgment threshold, and correspondingly, obtaining the data wild value through the following formula:
wherein G is an outlier estimate, Y i In order to calculate the result of the calculation,
data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
Optionally, the inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result, including:
Extracting a period term coefficient of troposphere delay in the variance data, and obtaining a residual term corresponding to the period term coefficient by using a least square method;
forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model;
removing periodic item data in ionosphere delay in the variance data to obtain residual error data;
inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model;
and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
In a second aspect, to achieve the above object, the present invention further provides an information extrapolation apparatus based on a neural network, including:
the sequence acquisition module is used for acquiring an atmospheric information history sequence and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
the training broadcast module is used for carrying out variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training, obtaining a training result and broadcasting the training result;
The extrapolation module is used for acquiring historical weather information, combining the historical weather information with target data received by broadcasting to acquire historical forecast information, inputting the historical forecast information into the preset neural network model to acquire output information, and extrapolating the output information.
In a third aspect, to achieve the above object, the present invention also proposes an information extrapolation apparatus based on a neural network, including: a memory, a processor, and a neural network-based information extrapolation program stored on the memory and executable on the processor, the neural network-based information extrapolation program configured to implement the steps of the neural network-based information extrapolation method as described above.
In a fourth aspect, to achieve the above object, the present invention also proposes a storage medium having stored thereon a neural network based information extrapolation program, which when executed by a processor, implements the steps of the neural network based information extrapolation method as described above.
According to the information extrapolation method based on the neural network, the atmospheric information history sequence is obtained, and missing time point data in the atmospheric information history sequence is supplemented by using a K-nearest neighbor algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the speed and the efficiency of information extrapolation based on a neural network are improved.
Drawings
FIG. 1 is a schematic diagram of a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a flowchart of a first embodiment of a neural network based information extrapolation method according to the present invention;
FIG. 3 is a flowchart of a second embodiment of a neural network based information extrapolation method according to the present invention;
FIG. 4 is a flowchart of a third embodiment of a neural network based information extrapolation method according to the present invention;
FIG. 5 is a flowchart of a fourth embodiment of a neural network based information extrapolation method according to the present invention;
fig. 6 is a functional block diagram of a first embodiment of the neural network-based information extrapolation apparatus of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Description of the embodiments
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The solution of the embodiment of the invention mainly comprises the following steps: the method comprises the steps of obtaining an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, the information extrapolating speed and efficiency based on the neural network are improved, and the technical problems that in the prior art, under the condition of complex atmospheric change, the change condition of the information extrapolating speed and efficiency based on the neural network cannot be accurately described by using a simple physical or mathematical model, a large amount of calculation load is increased by using the complex model, the positioning timeliness is affected, and the positioning accuracy is poor are solved.
Referring to fig. 1, fig. 1 is a schematic device structure diagram of a hardware running environment according to an embodiment of the present invention.
As shown in fig. 1, the apparatus may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display, an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., wi-Fi interface). The Memory 1005 may be a high-speed RAM Memory or a stable Memory (Non-Volatile Memory), such as a disk Memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
It will be appreciated by those skilled in the art that the apparatus structure shown in fig. 1 is not limiting of the apparatus and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components.
As shown in fig. 1, an operation device, a network communication module, a user interface module, and a neural network-based information extrapolation program may be included in the memory 1005 as one storage medium.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 by the processor 1001 and performs the following operations:
acquiring an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result;
historical weather information is acquired, the historical weather information is combined with target data received through broadcasting to obtain historical forecast information, the historical forecast information is input into the preset neural network model, output information is obtained, and extrapolation is conducted on the output information.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
acquiring an atmospheric information history sequence, and acquiring historical atmospheric delay data around a point to be interpolated from the atmospheric information history sequence;
and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
acquiring known point data around each point to be interpolated in the historical atmospheric delay data;
calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated;
and supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
and interpolating the missing values near the existing data according to the interpolation data, and recursively filling the missing whole area and time point data of the time period in the atmospheric information history sequence after interpolation is completed.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
acquiring a historical weather information estimation result, and carrying out variance weighting on the sequence according to the time and the resolution progress according to the historical weather information estimation result;
Removing the data wild value in the fixed weight sequence to obtain variance data;
and inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
calculating a wild value judgment threshold according to the preset confidence and the preset degree of freedom, determining a data wild value in the weight sequence according to the wild value judgment threshold, and correspondingly, obtaining the data wild value through the following formula:
wherein G is an outlier estimate, Y i In order to calculate the result of the calculation,
data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
The apparatus of the present invention calls the neural network-based information extrapolation program stored in the memory 1005 through the processor 1001, and performs the following operations:
extracting a period term coefficient of troposphere delay in the variance data, and obtaining a residual term corresponding to the period term coefficient by using a least square method;
Forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model;
removing periodic item data in ionosphere delay in the variance data to obtain residual error data;
inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model;
and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
According to the scheme, the atmospheric information history sequence is obtained, and the missing time point data in the atmospheric information history sequence is supplemented by using a K-nearest neighbor algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the speed and the efficiency of information extrapolation based on a neural network are improved.
Based on the hardware structure, the embodiment of the information extrapolation method based on the neural network is provided.
Referring to fig. 2, fig. 2 is a flowchart of a first embodiment of the information extrapolation method based on the neural network according to the present invention.
In a first embodiment, the neural network-based information extrapolation method includes the steps of:
and S10, acquiring an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K-nearest neighbor algorithm.
After the atmospheric information history sequence is obtained, the K-nearest neighbor method may be used to supplement the vacant time point data, that is, the K-nearest neighbor algorithm may be used to supplement the missing time point data in the atmospheric information history sequence, and in a specific implementation, the atmospheric information history sequences of a plurality of base stations may be collected by the calculation center.
And S20, carrying out variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
It can be understood that the variance weight is given to the sequence which is aligned, variance data can be obtained, the variance data is input into a preset neural network model for training, a training result is obtained, the training result is broadcasted, in actual operation, the variance weight can be given according to time and precision, after the data is obtained, the neural network model is added for training, and the training result is broadcasted through a communication means.
Step S30, acquiring historical weather information, combining the historical weather information with target data received by broadcasting to acquire historical forecast information, inputting the historical forecast information into the preset neural network model to acquire output information, and extrapolating the output information.
It should be understood that, the historical weather information is weather information in a period of time before the current moment, after the historical weather information is obtained, the historical weather information and the target data received by broadcasting can be combined, the historical forecast information can be obtained, the historical forecast information is further input into the preset neural network model, output information is obtained, and the output information is extrapolated.
In a specific implementation, after the user receives the atmospheric information model issued by the resolving center, the user combines the historical data and the model to obtain a prediction result suitable for the atmospheric condition of the location of the user, and the embodiment can provide extrapolation prediction of troposphere information and extrapolation prediction of ionosphere information and broadcast the extrapolation prediction to the user equipment by the server.
According to the scheme, the atmospheric information history sequence is obtained, and the missing time point data in the atmospheric information history sequence is supplemented by using a K-nearest neighbor algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the speed and the efficiency of information extrapolation based on a neural network are improved.
Further, fig. 3 is a schematic flow chart of a second embodiment of the information extrapolation method based on a neural network according to the present invention, and as shown in fig. 3, the second embodiment of the information extrapolation method based on a neural network according to the present invention is proposed based on the first embodiment, and in this embodiment, the step S10 specifically includes the following steps:
and S11, acquiring an atmosphere information history sequence, and acquiring historical atmosphere delay data around a point to be interpolated from the atmosphere information history sequence.
It should be noted that, before the data is filled, the historical atmospheric delay data around the point to be interpolated needs to be collected, and after the atmospheric information history sequence is obtained, the historical atmospheric delay data around the point to be interpolated may be obtained from the atmospheric information history sequence.
And step S12, calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
It can be understood that the interpolation data corresponding to the historical atmospheric delay data can be calculated by using a K-nearest neighbor algorithm, and then missing time point data in the atmospheric information historical sequence is supplemented according to the interpolation data.
According to the scheme, the historical atmospheric delay data around the point to be interpolated is obtained from the atmospheric information history sequence by obtaining the atmospheric information history sequence; and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data, so that the positioning accuracy and the convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the information extrapolation speed and efficiency based on a neural network are improved.
Further, fig. 4 is a schematic flow chart of a third embodiment of the information extrapolation method based on a neural network according to the present invention, and as shown in fig. 4, the third embodiment of the information extrapolation method based on a neural network according to the present invention is proposed based on the first embodiment, and in this embodiment, the step S12 specifically includes the following steps:
step S121, acquiring known point data around each point to be interpolated in the historical atmospheric delay data.
It should be noted that, the K-nearest neighbor algorithm may first obtain known point data around each point to be interpolated in the historical atmospheric delay data by using the characteristic that spatial data changes continuously in time and space and has strong correlation.
And step S122, calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated.
It can be understood that the euclidean distance and inverse distance weighting is calculated by using the known point data around the point to be interpolated to estimate the data of the point to be interpolated, that is, the euclidean distance and inverse distance of each known point are calculated by using the K-nearest neighbor algorithm according to the known point data, and the euclidean distance and inverse distance are weighted and estimated to obtain the interpolation data corresponding to each point to be interpolated.
And step 123, supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
It will be appreciated that the K-nearest neighbor algorithm is used to complement missing time point data in the data sequence, i.e. to complement missing time point data in the atmospheric information history sequence according to the interpolation data.
Further, the step S123 specifically includes the following steps:
and interpolating the missing values near the existing data according to the interpolation data, and recursively filling the missing whole area and time point data of the time period in the atmospheric information history sequence after interpolation is completed.
It can be understood that, because the intervals of the missing data windows are unequal and the occurrence time is irregular, firstly, interpolation is performed on the missing values near the existing data, and data completion of the whole area and the time period is completed recursively, namely, interpolation is performed on the missing values near the existing data according to the interpolation data, and after the interpolation is completed, the missing time point data of the whole area and the time period in the atmosphere information history sequence is completed recursively.
According to the scheme, the known point data around each point to be interpolated in the historical atmospheric delay data are obtained; calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated; according to the interpolation data, missing time point data in the atmospheric information history sequence can be supplemented accurately, positioning accuracy is further improved, and information extrapolation speed and efficiency based on a neural network are improved.
Further, fig. 5 is a schematic flow chart of a fourth embodiment of the information extrapolation method based on a neural network according to the present invention, as shown in fig. 5, and the fourth embodiment of the information extrapolation method based on a neural network according to the present invention is proposed based on the first embodiment, in which the step S20 specifically includes the following steps:
And S21, acquiring a historical weather information estimation result, and carrying out variance weighting on the sequence according to the time and the resolution progress.
After the weather information estimation result before the current epoch is collected, the time and the resolving precision can be subjected to variance weighting, namely, after the historical weather information estimation result is obtained, the time and the resolving progress can be used for carrying out variance weighting on the sequence which is aligned according to the historical weather information estimation result.
And S22, eliminating the data wild value in the fixed weight sequence to obtain variance data.
It can be understood that the corresponding variance data can be obtained after the data wild value in the weighted sequence is eliminated less. Further, the step S22 specifically includes the following steps:
calculating a wild value judgment threshold according to the preset confidence and the preset degree of freedom, determining a data wild value in the weight sequence according to the wild value judgment threshold, and correspondingly, obtaining the data wild value through the following formula:
wherein G is an outlier estimate, Y i To calculate the result，
Data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
In specific implementation, grubbs criterion can be used for eliminating wild value of data, whether the wild value exists in the data can be judged through the formula, G value of the data is calculated, then a threshold value is calculated through T test with alpha/2N as confidence and N-2 as freedom degree, and whether the wild value exists under the conditions of the confidence and the freedom degree is judged; and (5) carrying out subsequent processing after the data are complemented and the gross errors are removed by the K nearest neighbor interpolation method and Grubbs criterion.
And S23, inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
It should be understood that variance weighting is performed according to time and precision, after data is obtained, a neural network model is added for training, and a training result is broadcast through a communication means, namely the variance data can be input into a preset neural network model for training, a training result is obtained, and then the training result is broadcast.
Further, the step S23 specifically includes the following steps:
Extracting a period term coefficient of troposphere delay in the variance data, and obtaining a residual term corresponding to the period term coefficient by using a least square method;
forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model;
removing periodic item data in ionosphere delay in the variance data to obtain residual error data;
inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model;
and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
It can be understood that the tropospheric delay and the ionospheric delay are processed separately, the tropospheric delay is mainly represented by Zenith Tropospheric Delay (ZTD) in the positioning process, and after the historical tropospheric ZTD correction information is collected, the cycle term is extracted first, where the cycle term is as follows:
wherein A is 0 For a fixed trend term coefficient, A 1 And B is connected with 1 For the annual change coefficient, A 2 And B is connected with 2 The half-year period change coefficient is represented by a year, year is represented by year of data epoch, doy is year of data epoch, hour is represented by 24 hours system.
It should be understood that the residual term is extracted after the coefficient of the period term is obtained by using a least square method. And carrying out sliding window smoothing on the residual error item to remove noise points in the residual error item, and then forecasting the residual error item by utilizing an LSTM NN model, and combining the residual error item with an original periodic signal to obtain a ZTD forecasting model.
In a specific implementation, the ionosphere delay is mainly represented by using a vertical ionosphere (VTEC) in a zenith direction in a positioning process, after the historical ionosphere delay VTEC is collected, the solar activity condition and the geomagnetic activity condition corresponding to the historical time are considered, and for the period in which a magnetic storm is likely to occur and the period in which the ionosphere is calm, the ionosphere occupies less time, and therefore, a residual error can be obtained by a method of extracting and removing a periodic item, the residual error is input to an LSTM NN for training, and an ionosphere forecast model is provided by combining the periodic item data after training.
According to the scheme, by acquiring the historical weather information estimation result, variance weighting is carried out on the sequence which is aligned according to time and the resolution progress according to the historical weather information estimation result; removing the data wild value in the fixed weight sequence to obtain variance data; the variance data is input into a preset neural network model for training, a training result is obtained, and the training result is broadcasted, so that the positioning accuracy and the convergence time can be greatly improved, the calculation load can be reduced, and the positioning timeliness is improved.
Correspondingly, the invention further provides an information extrapolation device based on the neural network.
Referring to fig. 6, fig. 6 is a functional block diagram of a first embodiment of the neural network-based information extrapolation apparatus of the present invention.
In a first embodiment of the neural network-based information extrapolation apparatus of the present invention, the neural network-based information extrapolation apparatus includes:
the sequence acquisition module 10 is configured to acquire an atmospheric information history sequence, and supplement missing time point data in the atmospheric information history sequence by using a K-nearest neighbor algorithm.
The training broadcast module 20 is configured to perform variance weighting on the aligned sequences, obtain variance data, input the variance data into a preset neural network model for training, obtain a training result, and broadcast the training result.
The extrapolation module 30 is configured to obtain historical weather information, combine the historical weather information with target data received by broadcasting to obtain historical forecast information, input the historical forecast information into the preset neural network model, obtain output information, and extrapolate the output information.
The sequence acquisition module 10 is further configured to acquire an atmospheric information history sequence, and obtain historical atmospheric delay data around a point to be interpolated from the atmospheric information history sequence; and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
The sequence acquisition module 10 is further configured to acquire known point data around each point to be interpolated in the historical atmospheric delay data; calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated; and supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
The sequence obtaining module 10 is further configured to interpolate a missing value near existing data according to the interpolation data, and recursively fill in the missing time point data of the whole area and the time period in the atmospheric information history sequence after interpolation is completed.
The training broadcast module 20 is further configured to obtain a historical weather information estimation result, and perform variance weighting on the sequence according to the time and the resolution progress according to the historical weather information estimation result; removing the data wild value in the fixed weight sequence to obtain variance data; and inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
The training broadcast module 20 is further configured to calculate a wild value judgment threshold according to a preset confidence coefficient and a preset degree of freedom, determine a data wild value in the fixed weight sequence according to the wild value judgment threshold, and correspondingly obtain the data wild value according to the following formula:
wherein G is an outlier estimate, Y i In order to calculate the result of the calculation,
data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
The extrapolation module 30 is further configured to extract a periodic term coefficient of troposphere delay in the variance data, and obtain a residual term corresponding to the periodic term coefficient by using a least square method; forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model; removing periodic item data in ionosphere delay in the variance data to obtain residual error data; inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model; and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
The steps implemented by each functional module of the information extrapolation apparatus based on the neural network may refer to each embodiment of the information extrapolation method based on the neural network of the present invention, which is not described herein.
In addition, the embodiment of the invention also provides a storage medium, wherein the storage medium stores an information extrapolation program based on a neural network, and the information extrapolation program based on the neural network realizes the following operations when being executed by a processor:
acquiring an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result;
historical weather information is acquired, the historical weather information is combined with target data received through broadcasting to obtain historical forecast information, the historical forecast information is input into the preset neural network model, output information is obtained, and extrapolation is conducted on the output information.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
Acquiring an atmospheric information history sequence, and acquiring historical atmospheric delay data around a point to be interpolated from the atmospheric information history sequence;
and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
acquiring known point data around each point to be interpolated in the historical atmospheric delay data;
calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated;
and supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
and interpolating the missing values near the existing data according to the interpolation data, and recursively filling the missing whole area and time point data of the time period in the atmospheric information history sequence after interpolation is completed.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
acquiring a historical weather information estimation result, and carrying out variance weighting on the sequence according to the time and the resolution progress according to the historical weather information estimation result;
removing the data wild value in the fixed weight sequence to obtain variance data;
and inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
calculating a wild value judgment threshold according to the preset confidence and the preset degree of freedom, determining a data wild value in the weight sequence according to the wild value judgment threshold, and correspondingly, obtaining the data wild value through the following formula:
wherein G is an outlier estimate, Y i In order to calculate the result of the calculation,
data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
Further, the neural network-based information extrapolation program, when executed by the processor, further performs the following operations:
Extracting a period term coefficient of troposphere delay in the variance data, and obtaining a residual term corresponding to the period term coefficient by using a least square method;
forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model;
removing periodic item data in ionosphere delay in the variance data to obtain residual error data;
inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model;
and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
According to the scheme, the atmospheric information history sequence is obtained, and the missing time point data in the atmospheric information history sequence is supplemented by using a K-nearest neighbor algorithm; performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result; the method comprises the steps of obtaining historical weather information, combining the historical weather information with target data received by broadcasting to obtain historical forecast information, inputting the historical forecast information into a preset neural network model to obtain output information, and extrapolating the output information, so that the positioning accuracy and convergence time can be greatly improved, the calculation load can be reduced, the positioning timeliness is improved, the positioning accuracy is improved, and the speed and the efficiency of information extrapolation based on a neural network are improved.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (10)
1. A neural network-based information extrapolation method, comprising:
Acquiring an atmospheric information history sequence, and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
performing variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training to obtain a training result, and broadcasting the training result;
historical weather information is acquired, the historical weather information is combined with target data received through broadcasting to obtain historical forecast information, the historical forecast information is input into the preset neural network model, output information is obtained, and extrapolation is conducted on the output information.
2. The neural network-based information extrapolation method of claim 1, wherein the acquiring the historical sequence of atmospheric information and using a K-nearest neighbor algorithm to patch missing time point data in the historical sequence of atmospheric information comprises:
acquiring an atmospheric information history sequence, and acquiring historical atmospheric delay data around a point to be interpolated from the atmospheric information history sequence;
and calculating interpolation data corresponding to the historical atmospheric delay data by using a K approach algorithm, and supplementing missing time point data in the atmospheric information historical sequence according to the interpolation data.
3. The method for extrapolating information based on a neural network according to claim 2, wherein calculating interpolation data corresponding to the historical atmospheric delay data using a K-nearest neighbor algorithm, and supplementing missing time point data in the historical sequence of atmospheric information according to the interpolation data, comprises:
acquiring known point data around each point to be interpolated in the historical atmospheric delay data;
calculating Euclidean distance and inverse distance of each known point by using a K-nearest algorithm according to the known point data, and carrying out weighted estimation on the Euclidean distance and the inverse distance to obtain interpolation data corresponding to each point to be interpolated;
and supplementing missing time point data in the atmosphere information history sequence according to the interpolation data.
4. A neural network based information extrapolation method as claimed in claim 3, wherein the supplementing missing time point data in the atmospheric information history sequence based on the interpolation data comprises:
and interpolating the missing values near the existing data according to the interpolation data, and recursively filling the missing whole area and time point data of the time period in the atmospheric information history sequence after interpolation is completed.
5. The neural network-based information extrapolation method of claim 1, wherein the performing variance weighting on the aligned sequences to obtain variance data, inputting the variance data into a preset neural network model for training to obtain training results, and broadcasting the training results comprises:
Acquiring a historical weather information estimation result, and carrying out variance weighting on the sequence according to the time and the resolution progress according to the historical weather information estimation result;
removing the data wild value in the fixed weight sequence to obtain variance data;
and inputting the variance data into a preset neural network model for training, obtaining a training result, and broadcasting the training result.
6. The method for extrapolation of information based on neural network as claimed in claim 5, wherein the culling the data outliers in the weighted sequence to obtain variance data includes:
calculating a wild value judgment threshold according to the preset confidence and the preset degree of freedom, determining a data wild value in the weight sequence according to the wild value judgment threshold, and correspondingly, obtaining the data wild value through the following formula:
wherein G is an outlier estimate, Y i In order to calculate the result of the calculation,
data mean, N is data number, s is data variance, T is wild value confidence, N is preset degree of freedom, T α/(2N)，N-2 Presetting confidence;
and eliminating the data wild value from the weight sequence to obtain variance data.
7. The neural network-based information extrapolation method of claim 5, wherein inputting the variance data into a predetermined neural network model for training, obtaining training results, and broadcasting the training results, comprises:
Extracting a period term coefficient of troposphere delay in the variance data, and obtaining a residual term corresponding to the period term coefficient by using a least square method;
forecasting the residual error item by using a preset neural network model, and combining forecast data with an original periodic signal to obtain a troposphere forecast model;
removing periodic item data in ionosphere delay in the variance data to obtain residual error data;
inputting the residual data into the preset neural network model for training to obtain ionosphere training data, and combining the ionosphere training data with the periodic item data to obtain an ionosphere forecast model;
and taking the troposphere forecast model and the ionosphere forecast model as training results, and broadcasting the training results.
8. A neural network-based information extrapolation apparatus, comprising:
the sequence acquisition module is used for acquiring an atmospheric information history sequence and supplementing missing time point data in the atmospheric information history sequence by using a K approach algorithm;
the training broadcast module is used for carrying out variance weighting on the complemented sequence to obtain variance data, inputting the variance data into a preset neural network model for training, obtaining a training result and broadcasting the training result;
The extrapolation module is used for acquiring historical weather information, combining the historical weather information with target data received by broadcasting to acquire historical forecast information, inputting the historical forecast information into the preset neural network model to acquire output information, and extrapolating the output information.
9. A neural network-based information extrapolation apparatus, comprising: a memory, a processor, and a neural network-based information extrapolation program stored on the memory and executable on the processor, the neural network-based information extrapolation program configured to implement the steps of the neural network-based information extrapolation method of any one of claims 1-7.
10. A storage medium having stored thereon a neural network based information extrapolation program, which when executed by a processor implements the steps of the neural network based information extrapolation method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310310433.6A CN116011561B (en) | 2023-03-28 | 2023-03-28 | Information extrapolation method, device, equipment and storage medium based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310310433.6A CN116011561B (en) | 2023-03-28 | 2023-03-28 | Information extrapolation method, device, equipment and storage medium based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116011561A true CN116011561A (en) | 2023-04-25 |
| CN116011561B CN116011561B (en) | 2023-06-30 |
Family
ID=86032186
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310310433.6A Active CN116011561B (en) | 2023-03-28 | 2023-03-28 | Information extrapolation method, device, equipment and storage medium based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116011561B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6405132B1 (en) * | 1997-10-22 | 2002-06-11 | Intelligent Technologies International, Inc. | Accident avoidance system |
| US20190331832A1 (en) * | 2018-04-25 | 2019-10-31 | Microsoft Technology Licensing, Llc | Predicting microclimate |
| CN111210082A (en) * | 2020-01-13 | 2020-05-29 | 东南大学 | Optimized BP neural network algorithm-based precipitation prediction method |
| CN112446419A (en) * | 2020-10-29 | 2021-03-05 | 中山大学 | Time-space neural network radar echo extrapolation forecasting method based on attention mechanism |
| EP3867841A1 (en) * | 2018-10-15 | 2021-08-25 | Laser Navigation S.r.l. | System for controlling and managing a process within an environment using artificial intelligence techniques and relative method |
| US20220003894A1 (en) * | 2018-09-26 | 2022-01-06 | Sofar Ocean Technologies, Inc. | Ocean weather forecasting system |
| CN113971477A (en) * | 2021-09-26 | 2022-01-25 | 浙江大学 | PM based on time series and deep learning framework2.5Estimation method |
| CN114675311A (en) * | 2022-05-30 | 2022-06-28 | 长沙金维信息技术有限公司 | High-performance PVT (geometric velocity vector test) calculation method and navigation method |
| CN115526298A (en) * | 2022-10-18 | 2022-12-27 | 安徽工业大学 | High-robustness comprehensive prediction method for concentration of atmospheric pollutants |
-
2023
- 2023-03-28 CN CN202310310433.6A patent/CN116011561B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6405132B1 (en) * | 1997-10-22 | 2002-06-11 | Intelligent Technologies International, Inc. | Accident avoidance system |
| US20190331832A1 (en) * | 2018-04-25 | 2019-10-31 | Microsoft Technology Licensing, Llc | Predicting microclimate |
| US20220003894A1 (en) * | 2018-09-26 | 2022-01-06 | Sofar Ocean Technologies, Inc. | Ocean weather forecasting system |
| EP3867841A1 (en) * | 2018-10-15 | 2021-08-25 | Laser Navigation S.r.l. | System for controlling and managing a process within an environment using artificial intelligence techniques and relative method |
| CN111210082A (en) * | 2020-01-13 | 2020-05-29 | 东南大学 | Optimized BP neural network algorithm-based precipitation prediction method |
| CN112446419A (en) * | 2020-10-29 | 2021-03-05 | 中山大学 | Time-space neural network radar echo extrapolation forecasting method based on attention mechanism |
| CN113971477A (en) * | 2021-09-26 | 2022-01-25 | 浙江大学 | PM based on time series and deep learning framework2.5Estimation method |
| CN114675311A (en) * | 2022-05-30 | 2022-06-28 | 长沙金维信息技术有限公司 | High-performance PVT (geometric velocity vector test) calculation method and navigation method |
| CN115526298A (en) * | 2022-10-18 | 2022-12-27 | 安徽工业大学 | High-robustness comprehensive prediction method for concentration of atmospheric pollutants |
Non-Patent Citations (3)
| Title |
|---|
| HUAN ZHANG 等: "A Tropospheric Zenith Delay Forecasting Model Based on a Long Short-Term Memory Neural Network and Its Impact on Precise Point Positioning", 《REMOTE SENSING》 * |
| LEI LI 等: "A Regional NWP Tropospheric Delay Inversion Method Based on a General Regression Neural Network Model", 《SENSORS》 * |
| 孔岩 等: "基于方差分析周期叠加外推法的电离层 ＴＥＣ短期预报研究", 《大地测量与地区动力学》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116011561B (en) | 2023-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108761574B (en) | Rainfall estimation method based on multi-source information fusion | |
| CN112965146B (en) | Quantitative precipitation estimation method combining meteorological radar and rainfall barrel observation data | |
| CN111025346A (en) | Method for rapidly estimating clock error of GNSS precision satellite and storage medium | |
| CN111401602B (en) | Assimilation method for satellite and ground rainfall measurement values based on neural network | |
| CN111539453B (en) | Global ionized layer electron total content prediction method based on deep cycle neural network | |
| CN104180781B (en) | Deformation monitoring data processing method for single frequency and double frequency GPS hybrid network | |
| CN111077550A (en) | Gross error detection method and system applied to RTD positioning of intelligent terminal | |
| CN113094653B (en) | Method for reconstructing atmospheric temperature profile | |
| CN105068097A (en) | Self-adaptive filtering method for carrier smoothed code pseudorange | |
| KR101423278B1 (en) | System for calculating rainrate using Local Gauge Correction and method thereof | |
| CN116011561B (en) | Information extrapolation method, device, equipment and storage medium based on neural network | |
| CN116106948A (en) | Network RTK ionosphere interference resistant positioning method and related equipment | |
| CN111597692B (en) | Method, system, electronic equipment and storage medium for estimating surface net radiation | |
| JP3854111B2 (en) | Weather forecasting system, weather forecasting method, and weather forecasting program | |
| Sayin et al. | Optimum temporal update periods for regional ionosphere monitoring | |
| JP2006220444A (en) | Weather prediction system and its assimilation processing method | |
| CN115173925B (en) | Inter-satellite single difference ionosphere delay determination method and system | |
| CN111814855B (en) | Global ionospheric total electron content prediction method based on residual seq2seq neural network | |
| CN114155167B (en) | Automatic rapid correction method based on-line remote sensing satellite image | |
| CN111539433B (en) | Semantic segmentation based global ionosphere total electron content prediction method | |
| CN111539455B (en) | Global ionosphere electron total content prediction method based on image primary difference | |
| CN114004426B (en) | Dynamic adjustment method of short-time rainstorm forecast release model | |
| CN113177702B (en) | Meteorological input data matching method and system | |
| CN113791434B (en) | GNSS troposphere wet delay estimation method, system, equipment and medium | |
| CN115826017B (en) | Constraint ambiguity positioning method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |