CN115129706B - Soil moisture observation data quality evaluation method considering periodic characteristics - Google Patents
Soil moisture observation data quality evaluation method considering periodic characteristics Download PDFInfo
- Publication number
- CN115129706B CN115129706B CN202211068209.2A CN202211068209A CN115129706B CN 115129706 B CN115129706 B CN 115129706B CN 202211068209 A CN202211068209 A CN 202211068209A CN 115129706 B CN115129706 B CN 115129706B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- time
- calculation formula
- tcn
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of meteorological analysis, and provides a soil moisture observation data quality assessment method considering periodic characteristics. The method comprises the steps of obtaining observation data, eliminating abnormal data to serve as training data, improving a cause-effect convolution structure to form a new TCN model by taking a year as a data period and taking the structure of the TCN model as a basis, extracting features, setting a loss function according to the period of the training data, setting a training method, calculating to obtain model parameters, extracting the features of sample data, judging whether the distance of the sample data is compared with the features of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging whether the distance is generally normal or not, otherwise, sending an error positioning command, and determining an error position according to the model parameters and the TCN model of the cause-effect convolution structure. According to the scheme, the time sequence and periodicity characteristics of soil observation moisture data are fully and comprehensively utilized, and the data of an observation system are subjected to quality evaluation through a TCN model of a causal convolution structure.
Description
Technical Field
The invention relates to the technical field of meteorological analysis, in particular to a soil moisture observation data quality evaluation method considering periodic characteristics.
Background
The soil moisture is an important index for researching soil erosion, crop drought monitoring, yield prediction and the like, is also one of key parameters of a climate model, a hydrological model, an ecological model and a land process model, and has important significance for agriculture, drought, climate and the like. The method for acquiring the soil moisture through the ground automatic soil moisture observation system is an important data acquisition means at present. Currently, the observation of the soil moisture on the ground is generally carried out by a contact sensor. After long-term operation, the sensors, data transmission components, transmission networks, wireless networks, etc. inevitably have losses or faults, and these losses and faults may cause errors in the data measured, and the network faults may cause errors in the data received by the server, and these losses and faults are inevitable, which results in errors in the data received by the observation system. How to fully and comprehensively utilize the time sequence and periodicity characteristics of soil observation moisture data, the quality evaluation method is provided, and the method has important significance for improving the construction level of a ground soil moisture observation system.
The meteorological elements are formed by the revolution and rotation of the earth, the external forces of the sun, the moon and the like, the self structure of the earth and the like, and the meteorological elements are characterized in that on one hand: in a short period, the meteorological elements have the characteristic of changing along with the change of time, namely, the meteorological elements have time sequence; on the other hand, the soil moisture element takes the unit of year as a longer period, and the soil moisture element has stronger periodicity. How to fully and comprehensively utilize the time sequence and periodicity characteristics of soil observation moisture data, the quality evaluation method is provided, and the method has important significance for improving the construction level of an observation system. However, before the technology of the present invention, the TCN model is a feature extraction model with strong advantages appearing in recent years, and has certain advantages in terms of processing timing problems. But the model does not take into account the periodic nature of the time series data.
Disclosure of Invention
In view of the above problems, the invention provides a soil moisture observation data quality evaluation method considering periodic characteristics, and provides a quality evaluation method based on a deep learning technology on the basis of a TCN model on the basis of fully and comprehensively utilizing the time-sequence and periodic characteristics of soil moisture observation data, so as to evaluate the quality of data acquired by an observation system.
According to the first aspect of the embodiment of the invention, a soil moisture observation data quality evaluation method considering periodic characteristics is provided.
In one or more embodiments, preferably, the soil moisture observation data quality evaluation method considering the periodic characteristics includes:
acquiring all observation data, and removing abnormal data in the observation data to be used as training data;
updating the TCN model for extracting features on the basis of the structure of the TCN model by taking the year as a data period and the convolutional structure of the reason results;
setting a loss function of a TCN model of a causal convolution structure according to the period of the training data;
setting a training method, and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
extracting the characteristics of the sample data, judging the distance comparison with the characteristics of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging whether the distance is generally normal or not, and otherwise, sending an error positioning command;
and after receiving an error positioning command, determining the error position according to the model parameters and the TCN model of the causal convolution structure.
In one or more embodiments, preferably, the acquiring all observation data and rejecting abnormal data therein as training data specifically includes:
acquiring all soil humidity data, judging whether a sensor has a fault in the soil humidity data generation process, and if so, taking the corresponding soil humidity observation data as elimination error data;
and according to the removed error data, soil humidity data corresponding to the abnormal climate time period in the data are lifted and used as the training data.
In one or more embodiments, preferably, the new TCN model formed by improving the cause-effect convolution structure based on the structure of the TCN model and taking the year as a data cycle is used to extract features, and specifically includes:
dividing the input into values at a plurality of time points on the basis of the structure of the TCN model;
when calculating the characteristics of a certain node in the structure of the TCN model, further extracting new characteristics based on the characteristics generated by the previous data, and storing the result as the TCN model of the causal convolution structure;
and on the convolution layer set by the TCN model of the causal convolution structure, the original data covered by the last layer of the TCN model of the causal convolution structure is a complete period.
In one or more embodiments, preferably, the setting a loss function of the TCN model of the causal convolution structure according to the period of the training data specifically includes:
setting a periodic statistical characteristic for counting the characteristic in a certain period;
setting comprehensive characteristics for the characteristics of the last layer of the network;
setting a periodicity constraint characteristic for solving the minimum sum of squares of the difference between any two periods;
summing the intra-period statistical features, the comprehensive features and the periodic constraint features to obtain periodic losses;
setting a loss function of a TCN model of a causal convolution structure;
the loss function of the TCN model of the causal convolution structure is:
Loss=T 1 +T 2
wherein, T 1 As a loss of the TCN modelLoss function, T 2 For periodic losses, loss is the TCN model of the causal convolution structure.
In one or more embodiments, preferably, the setting a training method, and calculating to obtain a model parameter according to the training data, the loss function of the TCN model of the causal convolution structure, and the TCN model of the causal convolution structure, specifically include:
setting a training method to be gradient descent according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
and under the condition that the loss function of the TCN model of the causal convolution structure is minimum, calculating to obtain a set of model parameters.
In one or more embodiments, preferably, the extracting the features of the sample data, determining to compare the distances with the features of each period, and if the distance is smaller than a preset normal threshold, determining that the sample data is overall normal, otherwise, issuing an error positioning command specifically includes:
extracting the characteristics of the sample data, wherein the characteristics are one of mean value, variance, standard deviation, homogeneity or entropy and serve as basic statistical characteristics;
performing corresponding basic statistical characteristic analysis on all sample data in a period to obtain basic characteristics of the period;
setting a preset normal threshold value;
calculating the absolute value of the difference between the periodic basic feature and the basic statistical feature to be used as a comparison difference value;
and when the comparison difference is larger than the preset normal threshold value, sending the error positioning command, otherwise, judging that the overall system is normal.
In one or more embodiments, preferably, after receiving the error location command, determining the error location according to the model parameter and the TCN model of the causal convolution structure includes:
after the error positioning command is received, calculating a predicted value corresponding to each time and position according to the model parameters and the TCN model of the causal convolution structure;
calculating the number of time vertexes by using a first calculation formula, and calculating the number of space vertexes by using a second calculation formula;
fitting the corresponding predicted value of each time and positionxThe predicted value calculation function corresponding to the position is a third calculation formula;
fitting the corresponding predicted value of each time and positiontA predicted value calculation function corresponding to time is a fourth calculation formula;
judging the starting time and the ending time of the area meeting the fifth calculation formula according to the recorded value and each point meeting the third calculation formula;
judging the start position coordinates and the end position coordinates of the area meeting the sixth calculation formula according to the recorded values and each point meeting the fourth calculation formula;
calculating the coordinate position of the target by using a seventh calculation formula, and calculating the target time by using an eighth calculation formula;
determining the error position at a target time and a target coordinate position;
the first calculation formula is:
wherein the content of the first and second substances,yin order to predict the value of the target,tas a matter of time, the time is,n t is the number of time vertices;
the second calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,xin order to be a position of the user,n x the number of spatial vertices;
the third calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,y x is composed ofxThe predicted value corresponding to the position is obtained,A i is as followsiThe time coefficient of each of the plurality of time coefficients,A 0 is a time shift index;
the fourth calculation formula is:
wherein the content of the first and second substances,y t is composed oftThe predicted value corresponding to the position is obtained,B j is as followsjThe coefficient of the position,B 0 is a position deviation index;
the fifth calculation formula is:
wherein the content of the first and second substances,y c is the recorded value;
the sixth calculation formula is:
the seventh calculation formula is:
wherein the content of the first and second substances,X m in order to be able to determine the target coordinate position,X 1 is the coordinates of the position of the starting point,X 2 is a terminal position coordinate;
the eighth calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,t m for the purpose of the target time, the time is,t 1 is the starting time of the motor vehicle,t 2 is the termination time.
According to a second aspect of the embodiments of the present invention, there is provided a soil moisture observation data quality evaluation system considering a periodic characteristic.
In one or more embodiments, preferably, the soil moisture observation data quality evaluation system considering the periodic characteristics includes:
the training data organization module is used for acquiring all observation data, and eliminating abnormal data in the observation data to be used as training data;
the model design module is used for improving a TCN model formed by a reason result convolution structure on the basis of taking the year as a data cycle and taking the structure of the TCN model as a basis, and is used for extracting features;
the loss function design module is used for setting a loss function of a TCN model of a causal convolution structure according to the period of the training data;
the training setting module is used for setting a training method and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
the quality evaluation module is used for extracting the characteristics of the sample data, judging the distance comparison with the characteristics of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging whether the distance is generally normal or not, and otherwise, sending an error positioning command;
and the error positioning module is used for determining the error position according to the model parameters and the TCN model of the causal convolution structure after receiving an error positioning command.
According to a third aspect of embodiments of the present invention, there is provided a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method according to any one of the first aspect of embodiments of the present invention.
According to a fourth aspect of embodiments of the present invention, there is provided an electronic device, comprising a memory and a processor, the memory being configured to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method of any one of the first aspect of embodiments of the present invention.
The technical scheme provided by the embodiment of the invention can have the following beneficial effects:
the characteristics of soil humidity data are fully analyzed in the scheme of the invention, and firstly, the soil humidity data have certain periodicity by taking years as units; secondly, the soil humidity has regionality, namely the soil humidity of a certain region has a similar change mode; and thirdly, the time sequence is provided, namely, a certain time state is only related to the states at the previous times, namely, the certain time state is influenced by the previous states. On the basis of fully analyzing the characteristics, the invention provides a quality evaluation method based on a deep learning technology.
In the scheme of the invention, time sequence data to be detected are divided into two types: one is sequence data determined to be normal, and the other is sequence data determined to possibly contain error data, and for the sequence data possibly containing error data, the position where the error data exists is further indicated.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a soil moisture observation data quality evaluation method considering a periodic characteristic according to an embodiment of the present invention.
Fig. 2 is a flowchart of acquiring all observation data and removing abnormal data therein as training data in the soil moisture observation data quality assessment method considering the periodic characteristics according to an embodiment of the present invention.
Fig. 3 is a flowchart of a new TCN model formed by improving a cause-effect convolution structure based on the structure of the TCN model and taking an year as a data period in the soil moisture observation data quality assessment method considering periodic characteristics according to an embodiment of the present invention, for extracting characteristics.
Fig. 4 is a flowchart of setting a penalty function of a TCN model of a causal convolution structure according to a period of training data in a soil moisture observation data quality assessment method considering a periodic feature according to an embodiment of the present invention.
Fig. 5 is a flowchart of a setup training method in the soil moisture observation data quality assessment method considering the periodic characteristics, and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure, and the TCN model of the causal convolution structure, according to an embodiment of the present invention.
Fig. 6 is a flowchart of performing feature extraction on sample data, comparing the distance with the feature of each period, and determining whether the distance is smaller than a preset normal threshold value, if so, the sample data is generally normal, otherwise, an error positioning command is issued in the soil moisture observation data quality assessment method considering the periodic features according to an embodiment of the present invention.
FIG. 7 is a flowchart of determining the location of an error according to the model parameters and the TCN model of the causal convolution structure after an error location command is received in the soil moisture observation data quality assessment method considering the periodic characteristics according to the embodiment of the invention.
Fig. 8 is a block diagram of a soil moisture observation data quality evaluation system in consideration of a periodic characteristic according to an embodiment of the present invention.
FIG. 9 is a diagram of a conventional causal convolution.
FIG. 10 is a diagram of the logic relationships between three nodes in a conventional causal convolution structure.
FIG. 11 is a logical relationship between three nodes in a structure diagram of a TCN model of a causal convolution structure in an embodiment of the present invention.
Fig. 12 is a block diagram of an electronic device in one embodiment of the invention.
Detailed Description
In some flows described in the present specification and claims and above figures, a number of operations are included that occur in a particular order, but it should be clearly understood that these operations may be performed out of order or in parallel as they occur herein, with the order of the operations being given as 101, 102, etc. merely to distinguish between various operations, and the order of the operations itself does not represent any order of performance. Additionally, the flows may include more or fewer operations, and the operations may be performed sequentially or in parallel. It should be noted that, the descriptions of "first", "second", etc. in this document are used for distinguishing different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit the types of "first" and "second".
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Meteorological element ground data is an important basic data source, and soil moisture is an important meteorological element and has an important influence on weather. The acquisition of meteorological element data through a ground observation system is an important acquisition means at present. Currently, the observation of soil moisture in a ground observation system is generally performed by a contact sensor. After long-term operation, the sensors, data transmission components, transmission networks, wireless networks, etc. inevitably have losses or faults, and these losses and faults may cause errors in the data measured, and the network faults may cause errors in the data received by the server, and these losses and faults are inevitable, which results in errors in the data received by the observation system. How to fully and comprehensively utilize the time sequence and periodicity characteristics of soil observation moisture data, the quality evaluation method is provided, and the method has important significance for improving the construction level of a ground soil moisture observation system.
The meteorological elements are formed by the revolution and rotation of the earth, the external force of the sun, the moon and the like, the self structure of the earth and the like, and the meteorological elements have the outstanding characteristics that, on one hand: in a short period, the meteorological elements have the characteristic of changing along with the change of time, namely, the meteorological elements have time sequence; on the other hand, in a long-term time, meteorological elements take years as units and have strong periodicity. How to fully and comprehensively utilize the time sequence and periodicity characteristics of soil observation moisture data, the quality evaluation method is provided, and the method has important significance for improving the construction level of an observation system. However, before the technology of the present invention, the TCN model is a feature extraction model with strong advantages appearing in recent years, and has certain advantages in terms of processing timing problems. But the model does not take into account the periodic nature of the time series data.
The embodiment of the invention provides a soil moisture observation data quality evaluation method considering periodic characteristics. According to the technical scheme, on the basis of fully and comprehensively utilizing the time sequence and periodicity characteristics of soil observation moisture data, a quality evaluation method based on a deep learning technology is provided on the basis of a TCN (transmission control network) model, and quality evaluation is carried out on data acquired by an observation system.
According to the first aspect of the embodiment of the invention, the soil moisture observation data quality evaluation method considering the periodic characteristics is provided.
Fig. 1 is a flowchart of a soil moisture observation data quality evaluation method considering a periodic characteristic according to an embodiment of the present invention.
In one or more embodiments, preferably, the soil moisture observation data quality evaluation method considering the periodic characteristics includes:
s101, acquiring all observation data, and removing abnormal data in the observation data to serve as training data;
s102, taking the year as a data cycle, and improving a cause result convolution structure to form a new TCN model for extracting features on the basis of the structure of the TCN model;
s103, setting a loss function of a TCN model of a causal convolution structure according to the period of the training data;
s104, setting a training method, and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
s105, extracting features of the sample data, judging the distance comparison with the features of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging whether the distance is generally normal, and otherwise, sending an error positioning command;
and S106, after receiving the error positioning command, determining the error position according to the model parameters and the TCN model of the causal convolution structure.
In the embodiment of the invention, soil humidity data (meteorological element ground data) is an important basic data source, but because of the inevitable loss of environment and equipment in long-term operation, some error data inevitably appear in data observed by a ground observation system in the long-term operation process, how to detect and correct the data to ensure the reliability of a basic database is a problem that must be solved by the meteorological observation system.
Fig. 2 is a flowchart of acquiring all observation data and removing abnormal data therein as training data in the soil moisture observation data quality evaluation method considering the periodic characteristics according to an embodiment of the present invention.
As shown in fig. 2, in one or more embodiments, preferably, the acquiring all observation data and rejecting abnormal data therein as training data specifically includes:
s201, acquiring all observation data of soil humidity, judging whether a sensor has a fault in the process of generating the observation data of the soil humidity, and if so, deleting the corresponding observation data of the soil humidity to serve as elimination error data;
s202, according to the error eliminating data, extracting observation data of soil humidity corresponding to the climate abnormal time period to serve as the training data.
In the embodiment of the invention, the verified data is used as the training data, and the requirement on the data is two aspects: firstly, the data are verified manually, so that no error data caused by faults of sensors and the like exist, and secondly, the data are abnormal due to abnormal weather, and the abnormal weather is not representative.
Fig. 3 is a flowchart of a new TCN model formed by improving a cause-effect convolution structure based on the structure of the TCN model and taking an year as a data period in the soil moisture observation data quality assessment method considering periodic characteristics according to an embodiment of the present invention, for extracting characteristics.
As shown in fig. 3, in one or more embodiments, preferably, the new TCN model formed by improving the convolution structure of the reason and effect based on the structure of the TCN model is used to extract features, and specifically includes:
s301, dividing input into values of a plurality of time points on the basis of the structure of the TCN model;
s302, when calculating the characteristics of a certain node in the structure of the TCN model, further extracting new characteristics based on the characteristics generated by the previous data, and storing the result as the TCN model of a causal convolution structure;
and S303, on the convolution layer set by the TCN model of the causal convolution structure, covering the original data by the last layer of the TCN model of the causal convolution structure by a complete period.
In the embodiment of the present invention, the TCN model is a time domain convolutional network, which is a video-based motion segmentation time convolutional network, and is generally used for sound event localization and detection and probability prediction. In the traditional TCN model, the causal convolution is selected as a basic component for feature extraction, and then the causal convolution is stacked through an expansion convolution mode, so that the expression of one state can be influenced by a plurality of previous states. Using a residual structure as a connection structure of multiple layers of feature extraction layers, in the conventional TCN model, each original data (e.g. x 0) needs to provide input for multiple upper layers, which, although providing consistent operation results, has two drawbacks: (1) the amount of computation is too large in computation. (2) Logically, although the influence between non-adjacent timings is indirect, the influence is not direct.
FIG. 4 is a flowchart of setting a loss function of a TCN model of a causal convolution structure according to a period of training data in a soil moisture observation data quality assessment method considering a periodic characteristic according to an embodiment of the present invention.
As shown in fig. 4, in one or more embodiments, preferably, the setting a loss function of the TCN model of the causal convolution structure according to the period of the training data specifically includes:
s401, setting a periodic statistic feature for counting the feature in a certain period;
s402, setting comprehensive characteristics for characteristics of the last layer of the network;
s403, setting a periodicity constraint characteristic for solving the minimum sum of squares of differences between any two periods;
s404, adding the statistical characteristics in the period, the comprehensive characteristics and the periodic constraint characteristics to obtain periodic loss;
s405, setting a loss function of a TCN model of a causal convolution structure;
the loss function of the TCN model of the causal convolution structure is:
Loss=T 1 +T 2
wherein, T 1 Being a loss function of the TCN model, T 2 For periodic losses, loss is the TCN model of the causal convolution structure.
In the embodiment of the invention, the loss function is designed to meet the requirements of two aspects. First, the timing requirement, that is, the preceding data, has an effect on the following data; second, periodic characteristics. Since the penalty function of TCN itself already expresses the timing requirements, the use of periodic characteristics is emphasized below. The periodic characteristics include the following aspects: (1) statistical characteristics: in other words, for a certain period, the characteristics of the bottom layer are counted, and the following statistics are used: the following are used: mean (Mean), variance (Variance), standard deviation (Std), homogeneity (Homogeneity), entropy (Entropy), establish the basic statistics. (2) comprehensive periodic characteristics: i.e. the characteristics of the last layer of the network. And (3) constraint of periodic characteristics. The sum of the squares of the differences between any two cycles is minimal (indicative of the phenological characteristics). Based on the characteristics, the Loss function is proposed to be less = T1+ T2.
Fig. 5 is a flowchart of a setup training method in the soil moisture observation data quality assessment method considering the periodic characteristics, and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure, and the TCN model of the causal convolution structure, according to an embodiment of the present invention.
As shown in fig. 5, in one or more embodiments, preferably, the setting a training method, and calculating to obtain a model parameter according to the training data, the loss function of the TCN model of the causal convolution structure, and the TCN model of the causal convolution structure, specifically includes:
s501, setting a training method to be gradient descent according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
s502, under the condition that the loss function of the TCN model of the causal convolution structure is minimum, calculating to obtain a group of model parameters.
In the embodiment of the invention, gradient descent is used as a training method, and a group of model parameters is obtained after model training is finished.
Fig. 6 is a flowchart of performing feature extraction on sample data, comparing the distance with the feature of each period, and determining whether the distance is smaller than a preset normal threshold value, if so, the sample data is generally normal, otherwise, an error positioning command is issued in the soil moisture observation data quality assessment method considering the periodic features according to an embodiment of the present invention.
As shown in fig. 6, in one or more embodiments, preferably, the extracting features of the sample data, comparing the distances of the sample data with the features of each period, and if the distance is smaller than a preset normal threshold, the result is that the sample data is overall normal, otherwise, an error location command is issued, specifically including:
s601, extracting characteristics of the sample data, wherein the characteristics are one of mean value, variance, standard deviation, homogeneity or entropy and serve as basic statistical characteristics;
s602, performing corresponding basic statistical characteristic analysis on all sample data in a period to obtain basic characteristics of the period;
s603, setting a preset normal threshold value;
s604, calculating an absolute value of the difference between the periodic basic feature and the basic statistical feature to serve as a comparison difference value;
s605, when the comparison difference value is larger than the preset normal threshold value, sending the error positioning command, otherwise, judging that the overall system is normal.
In the embodiment of the invention, the evaluation of whether the whole is normal or not is carried out by combining the cycle basic characteristics, so that the judgment result of the whole is sent out, and if the judgment result is abnormal, the further analysis is carried out.
FIG. 7 is a flowchart of determining the location of an error according to the model parameters and the TCN model of the causal convolution structure after an error location command is received in the soil moisture observation data quality assessment method considering the periodic characteristics according to the embodiment of the invention.
As shown in fig. 7, in one or more embodiments, preferably, after receiving the error location command, determining the error location according to the model parameter and the TCN model of the causal convolution structure includes:
s701, after the error positioning command is received, calculating a predicted value corresponding to each time and each position according to the model parameters and the TCN model of the causal convolution structure;
s702, calculating the number of time vertexes by using a first calculation formula, and calculating the number of space vertexes by using a second calculation formula;
s703, fitting the predicted value corresponding to each time and position with the predicted value calculation function corresponding to the x position to form a third calculation formula;
s704, fitting a predicted value corresponding to each time and position to a predicted value calculation function corresponding to t time to form a fourth calculation formula;
s705, judging the starting time and the ending time of an area meeting a fifth calculation formula according to the recorded values and each point meeting the third calculation formula;
s706, judging the start point position coordinates and the end point position coordinates of the area meeting the sixth calculation formula according to the recorded values and each point meeting the fourth calculation formula;
s707, calculating a target coordinate position by using a seventh calculation formula, and calculating target time by using an eighth calculation formula;
s708, determining the error position at the target time and the target coordinate position;
the first calculation formula is:
wherein the content of the first and second substances,yin order to predict the value of the target,tas a matter of time, the time is,n t is the number of time vertices;
the second calculation formula is:
wherein the content of the first and second substances,xin order to be a position of the user,n x the number of spatial vertices;
the third calculation formula is:
wherein the content of the first and second substances,y x is composed ofxThe predicted value corresponding to the position is obtained,A i is a firstiThe time coefficient of each of the plurality of time coefficients,A 0 is a time shift index;
the fourth calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,y t is composed oftThe predicted value corresponding to the position is calculated,B j is a firstjThe coefficient of the position is,B 0 is a position deviation index;
the fifth calculation formula is:
wherein the content of the first and second substances,y c is the recorded value;
the sixth calculation formula is:
the seventh calculation formula is:
wherein the content of the first and second substances,X m in order to be able to determine the target coordinate position,X 1 is the coordinates of the position of the starting point,X 2 is a terminal position coordinate;
the eighth calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,t m in order to be the target time of the time,t 1 as a result of the start-up time,t 2 is the termination time.
In the embodiment of the invention, for time sequence data possibly having errors, the extracted features are firstly used for preliminarily judging the error position and confirming the time prediction function, and the specific steps are as follows: determining a period in which errors may exist according to the periodic characteristics; and performing predictive calculation on the data in the period one by using the successfully trained model, comparing the predicted value with the recorded value, and obtaining the error position when the predicted value and the recorded value reach the target time and the target coordinate position which are correspondingly obtained by the first calculation formula, the second calculation formula, the third calculation formula, the fourth calculation formula, the fifth calculation formula, the sixth calculation formula, the seventh calculation formula and the eighth calculation formula.
According to a second aspect of the embodiments of the present invention, there is provided a soil moisture observation data quality evaluation system considering a periodic characteristic.
Fig. 8 is a block diagram of a soil moisture observation data quality evaluation system in consideration of a periodic characteristic according to an embodiment of the present invention.
In one or more embodiments, preferably, the soil moisture observation data quality evaluation system considering the periodic characteristics includes:
a training data organization module 801, configured to acquire all observation data, and remove abnormal data therein as training data;
a model design module 802, configured to improve a cause-effect convolution structure to form a new TCN model for feature extraction based on a TCN model structure with a year as a data cycle;
a loss function design module 803, configured to set a loss function of the TCN model of the causal convolution structure according to the period of the training data;
a training setting module 804, configured to set a training method, and calculate to obtain a model parameter according to the training data, the loss function of the TCN model of the causal convolution structure, and the TCN model of the causal convolution structure;
the quality evaluation module 805 is configured to perform feature extraction on the sample data, determine a distance comparison with features of each period, determine whether the distance is smaller than a preset normal threshold, determine that the sample data is generally normal, and otherwise, send an error positioning command;
an error localization module 806, configured to determine, after receiving the error localization command, an error location according to the model parameter and the TCN model of the causal convolution structure.
In the embodiment of the invention, a modular design structure considering periodic characteristics is provided, so that the quality evaluation of soil moisture observation data is realized.
Specifically, for example, a conventional causal convolution structure is shown in FIG. 9. In fig. 9, each original data (e.g. x 0) is to be provided as input for multiple upper layers, which, although providing consistent operation results, has two disadvantages: (1) the amount of computation is too large in computation. (2) Logically, although the influence between non-adjacent timings is indirect, the influence is not direct. Specifically, as shown in fig. 10.
In fig. 10, it is illustrated that, although the value at time point No. 1 affects the value at time point No. 2 and the value at time point No. 3, the effect on the value at time point No. 3 is indirect, and node C in fig. 10 indicates that features are extracted using 3 nodes. Thus, logically, node C uses 1,2,3 as a parallel quantity and does not express this precedence relationship.
Fig. 11 is a diagram illustrating a structure of a novel causal convolution according to an embodiment of the present invention. The novel causal convolution structure is as follows: 1,2,3 is the original sequence. The node B needs to extract the feature of the node A3, but the input used by the node B is the feature of the node A and the node B3, because the meaning of the expression of the node A is the comprehensive action of the nodes, the mode expresses the influence of the nodes before the node B3 on the node B3, expresses the influence of the nodes before the node B3 through the node B2 and is indirect influence. When the characteristics of a certain node are calculated, the novel causal convolution structure designed by the invention is not extracted from original data again as before, but is further extracted based on the generated characteristics of the previous data, so that on one hand, the calculated amount can be reduced, and on the other hand, the novel causal convolution structure better accords with the characteristics of time sequence data: a point is affected by its previous node.
According to a third aspect of embodiments of the present invention, there is provided a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method according to any one of the first aspect of embodiments of the present invention.
According to a fourth aspect of the embodiments of the present invention, there is provided an electronic apparatus. Fig. 12 is a block diagram of an electronic device in one embodiment of the invention. The electronic device shown in fig. 12 is a soil moisture observation data quality evaluation device that generally considers periodic characteristics. The electronic device can be a smart phone, a tablet computer and the like. As shown, electronic device 1200 includes a processor 1201 and a memory 1202. The processor 1201 is electrically connected to the memory 1202. The processor 1201 is a control center of the terminal 1200, connects various parts of the entire terminal using various interfaces and lines, and performs various functions of the terminal and processes data by running or calling a computer program stored in the memory 1202 and calling data stored in the memory 1202, thereby performing overall monitoring of the terminal.
In this embodiment, the processor 1201 in the electronic device 1200 loads instructions corresponding to one or more computer program processes into the memory 1202 according to the following steps, and the processor 1201 runs the computer program stored in the memory 1202, thereby implementing various functions: the method comprises the steps of acquiring all observation data, removing abnormal data to serve as training data, taking years as data periods, improving a TCN model with a cause-effect convolution structure to form a new TCN model used for extracting features on the basis of the structure of the TCN model, setting a loss function of the TCN model with a cause-effect convolution structure according to the period of the training data, setting a training method, calculating to obtain model parameters according to the training data, the loss function of the TCN model with the cause-effect convolution structure and the TCN model with the cause-effect convolution structure, extracting features of sample data, judging whether the distance is smaller than a preset normal threshold value or not by comparing the distance with the features of each period, sending an error positioning command if the distance is smaller than the preset normal threshold value, and determining an error position according to the model parameters and the TCN model with the cause-effect convolution structure after the error positioning command is received.
Memory 1202 may be used for storing computer programs and data. Memory 1202 stores computer programs containing instructions executable in the processor. The computer program may constitute various functional modules. The processor 1201 executes various functional applications and data processing by calling a computer program stored in the memory 1202.
The technical scheme provided by the embodiment of the invention can have the following beneficial effects:
the characteristics of soil humidity data are fully analyzed in the scheme of the invention, and firstly, the soil humidity data have certain periodicity by taking years as units; secondly, the soil humidity has regionality, namely the soil humidity of a certain region has a similar change mode; and thirdly, the time sequence is provided, namely, a certain time state is only related to the states at the previous times, namely, the certain time state is influenced by the previous states. On the basis of fully analyzing the characteristics, the invention provides a quality evaluation method based on a deep learning technology.
In the scheme of the invention, the time sequence data to be detected are divided into two types: one is sequence data determined to be normal, and the other is sequence data determined to possibly contain error data, and for the sequence data possibly containing error data, the position where the error data exists is further indicated.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (7)
1. A soil moisture observation data quality assessment method considering periodic characteristics is characterized by comprising the following steps:
acquiring all observation data, and removing abnormal data in the observation data to be used as training data;
taking years as a data period, and on the basis of the structure of the TCN model, improving the structure of the cause-effect convolution to form a new TCN model for extracting features;
setting a loss function of a TCN model of a causal convolution structure according to the period of the training data;
setting a training method, and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
extracting the characteristics of the sample data, comparing the distance with the characteristics of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging that the sample data is generally normal, and otherwise, sending an error positioning command;
after an error positioning command is received, determining an error position according to the model parameters and the TCN model of the causal convolution structure;
the new TCN model formed by improving the cause-effect convolution structure is used to extract features, and the new TCN model is formed based on the structure of the TCN model and using the year as a data cycle, and specifically includes:
dividing the input into values of a plurality of time points based on the structure of the TCN model;
when calculating the characteristics of a certain node in the structure of the TCN model, further extracting new characteristics based on the characteristics generated by the previous data, and storing the result as the TCN model of a causal convolution structure;
on the TCN model setting convolution layer of the causal convolution structure, the original data covered by the last layer of the TCN model of the causal convolution structure is a complete period;
wherein, according to the period of the training data, the setting of the loss function of the TCN model of the causal convolution structure specifically includes:
setting a periodic statistical characteristic for counting the characteristic in a certain period;
setting comprehensive characteristics for the characteristics of the last layer of the network;
setting a periodicity constraint characteristic for solving the minimum sum of squares of the difference between any two periods;
summing the intra-period statistical features, the comprehensive features and the periodic constraint features to obtain periodic losses;
setting a loss function of a TCN model of a causal convolution structure;
the loss function of the TCN model of the causal convolution structure is:
Loss=T 1 +T 2
wherein, T 1 Being a loss function of the TCN model, T 2 For periodic losses, loss is the TCN model of the causal convolution structure;
after receiving the error positioning command, determining an error position according to the model parameter and the TCN model of the causal convolution structure, specifically including:
after the error positioning command is received, calculating a predicted value corresponding to each time and position according to the model parameters and the TCN model of the causal convolution structure;
calculating the number of time vertexes by using a first calculation formula, and calculating the number of space vertexes by using a second calculation formula;
fitting the corresponding predicted value of each time and positionxThe predicted value calculation function corresponding to the position is a third calculation formula;
fitting the corresponding predicted value of each time and positiontThe predicted value calculation function corresponding to the time is a fourth calculation formula;
judging the starting time and the ending time of the area meeting the fifth calculation formula according to the recorded value and each point meeting the third calculation formula;
judging the start position coordinates and the end position coordinates of the area meeting the sixth calculation formula according to the recorded values and each point meeting the fourth calculation formula;
calculating the target coordinate position by using a seventh calculation formula, and calculating the target time by using an eighth calculation formula;
determining the error position at a target time and a target coordinate position;
the first calculation formula is:
wherein the content of the first and second substances,yin order to predict the value of the target,tas a matter of time, the time is,n t for the number of time vertices, count () is a counting function;
the second calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,xin order to be a position of the user,n x the number of spatial vertices;
the third calculation formula is:
wherein the content of the first and second substances,y x is composed ofxThe predicted value corresponding to the position is obtained,A i is as followsiThe time coefficient of each of the plurality of time coefficients,A 0 in order to be an index of the time shift,ithe number of the time coefficient;
the fourth calculation formula is:
wherein the content of the first and second substances,y t is composed oftThe predicted value corresponding to the position is obtained,B j is a firstjThe coefficient of the position is,B 0 in order to be an index of the position deviation,jthe number of the position coefficient;
the fifth calculation formula is:
wherein the content of the first and second substances,y c is the recorded value;
the sixth calculation formula is:
the seventh calculation formula is:
wherein the content of the first and second substances,X m for the purpose of the target coordinate position,X 1 is the coordinates of the position of the starting point,X 2 is a terminal position coordinate;
the eighth calculation formula is:
wherein the content of the first and second substances,t m for the purpose of the target time, the time is,t 1 as a result of the start-up time,t 2 is the termination time.
2. The method for evaluating the quality of soil moisture observation data considering periodic characteristics according to claim 1, wherein the acquiring all observation data and removing abnormal data therein as training data specifically comprises:
acquiring all observation data of soil humidity, judging whether a sensor of the soil humidity data has a fault in the observation process, and if the sensor has the fault, deleting the observation data of the corresponding soil humidity data to be taken as reject error data;
and extracting observation data of soil humidity corresponding to the climate abnormal time period according to the removed error data to serve as the training data.
3. The method for evaluating the quality of soil moisture observation data considering periodic characteristics as claimed in claim 1, wherein the setting of a training method and the calculation of model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure comprise:
setting a training method to be gradient descent according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
and under the condition that the loss function of the TCN model of the causal convolution structure is minimum, calculating to obtain a group of model parameters.
4. The method according to claim 1, wherein the step of extracting the features of the sample data, comparing the distances of the extracted sample data with the features of each period, judging whether the distances are smaller than a preset normal threshold value, if not, determining that the samples are generally normal, otherwise, sending an error positioning command includes:
extracting characteristics of the sample data, wherein the characteristics are one of mean value, variance, standard deviation, homogeneity or entropy and serve as basic statistical characteristics;
performing corresponding basic statistical characteristic analysis on all sample data in a period to obtain basic characteristics of the period;
setting a preset normal threshold value;
calculating the absolute value of the difference between the periodic basic feature and the basic statistical feature to be used as a comparison difference value;
and when the comparison difference is larger than the preset normal threshold value, sending the error positioning command, otherwise, judging that the overall system is normal.
5. A soil moisture observation data quality evaluation system considering a periodic characteristic, the system comprising:
the training data organization module is used for acquiring all observation data, and eliminating abnormal data in the observation data to be used as training data;
the model design module is used for improving a TCN model formed by a reason result convolution structure on the basis of taking the year as a data cycle and taking the structure of the TCN model as a basis, and is used for extracting features;
the loss function design module is used for setting a loss function of a TCN model of a causal convolution structure according to the period of the training data;
the training setting module is used for setting a training method and calculating to obtain model parameters according to the training data, the loss function of the TCN model of the causal convolution structure and the TCN model of the causal convolution structure;
the quality evaluation module is used for extracting the characteristics of the sample data, comparing the distance with the characteristics of each period, judging whether the distance is smaller than a preset normal threshold value or not, judging that the sample data is overall normal, and otherwise, sending an error positioning command;
the error positioning module is used for determining the error position according to the model parameters and the TCN model of the causal convolution structure after receiving an error positioning command;
the new TCN model formed by improving the cause-effect convolution structure based on the year as a data cycle and the TCN model structure is used for extracting features, and the method specifically comprises the following steps:
dividing the input into values of a plurality of time points based on the structure of the TCN model;
when calculating the characteristics of a certain node in the structure of the TCN model, further extracting new characteristics based on the characteristics generated by the previous data, and storing the result as the TCN model of a causal convolution structure;
on the TCN model setting convolution layer of the causal convolution structure, the original data covered by the last layer of the TCN model of the causal convolution structure is a complete period;
wherein, according to the period of the training data, the setting of the loss function of the TCN model of the causal convolution structure specifically includes:
setting a periodic statistical characteristic for counting the characteristic in a certain period;
setting comprehensive characteristics for the characteristics of the last layer of the network;
setting a periodicity constraint characteristic for solving the minimum sum of squares of the difference between any two periods;
summing the intra-period statistical features, the comprehensive features and the periodic constraint features to obtain periodic losses;
setting a loss function of a TCN model of a causal convolution structure;
the loss function of the TCN model of the causal convolution structure is:
Loss=T 1 +T 2
wherein, T 1 Being a loss function of the TCN model, T 2 For periodic losses, loss is the TCN model of the causal convolution structure;
after receiving the error positioning command, determining an error position according to the model parameter and the TCN model of the causal convolution structure, specifically including:
after the error positioning command is received, calculating a predicted value corresponding to each time and position according to the model parameters and the TCN model of the causal convolution structure;
calculating the number of time vertexes by using a first calculation formula, and calculating the number of space vertexes by using a second calculation formula;
fitting the corresponding predicted value of each time and positionxThe predicted value calculation function corresponding to the position is a third calculation formula;
fitting the corresponding predicted value of each time and positiontThe predicted value calculation function corresponding to the time is a fourth calculation formula;
judging the starting time and the ending time of the area meeting the fifth calculation formula according to the recorded value and each point meeting the third calculation formula;
judging the start position coordinates and the end position coordinates of the area meeting the sixth calculation formula according to the recorded values and each point meeting the fourth calculation formula;
calculating the coordinate position of the target by using a seventh calculation formula, and calculating the target time by using an eighth calculation formula;
determining the error position at a target time and a target coordinate position;
the first calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,yin order to predict the value of the target,tin the form of a time, the time,n t for the number of time vertices, count () is a counting function;
the second calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,xin order to be a position of the user,n x the number of spatial vertices;
the third calculation formula is:
wherein the content of the first and second substances,y x is composed ofxThe predicted value corresponding to the position is obtained,A i is as followsiThe time coefficient of each of the plurality of time coefficients,A 0 in order to be an index of the time shift,ithe number of the time coefficient;
the fourth calculation formula is:
wherein the content of the first and second substances,y t is composed oftThe predicted value corresponding to the position is obtained,B j is as followsjThe coefficient of the position,B 0 in order to be an index of the position deviation,jthe number of the position coefficient;
the fifth calculation formula is:
wherein the content of the first and second substances,y c is the recorded value;
the sixth calculation formula is:
the seventh calculation formula is:
wherein the content of the first and second substances,X m for the purpose of the target coordinate position,X 1 is the coordinates of the position of the starting point,X 2 is a terminal position coordinate;
the eighth calculation formula is:
wherein, the first and the second end of the pipe are connected with each other,t m for the purpose of the target time, the time is,t 1 is the starting time of the motor vehicle,t 2 is the termination time.
6. A computer-readable storage medium on which computer program instructions are stored, which, when executed by a processor, implement the method of any one of claims 1-4.
7. An electronic device comprising a memory and a processor, wherein the memory is configured to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211068209.2A CN115129706B (en) | 2022-09-02 | 2022-09-02 | Soil moisture observation data quality evaluation method considering periodic characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211068209.2A CN115129706B (en) | 2022-09-02 | 2022-09-02 | Soil moisture observation data quality evaluation method considering periodic characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115129706A CN115129706A (en) | 2022-09-30 |
| CN115129706B true CN115129706B (en) | 2022-11-25 |
Family
ID=83387147
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211068209.2A Active CN115129706B (en) | 2022-09-02 | 2022-09-02 | Soil moisture observation data quality evaluation method considering periodic characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115129706B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115587629B (en) * | 2022-12-07 | 2023-04-07 | 中国科学院上海高等研究院 | Covariance expansion coefficient estimation method, model training method and storage medium terminal |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210012767A1 (en) * | 2020-09-25 | 2021-01-14 | Intel Corporation | Real-time dynamic noise reduction using convolutional networks |
| AU2020104000A4 (en) * | 2020-12-10 | 2021-02-18 | Guangxi University | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model |
| CN112668775A (en) * | 2020-12-25 | 2021-04-16 | 西安翔迅科技有限责任公司 | Air quality prediction method based on time sequence convolution network algorithm |
| CN114740388A (en) * | 2022-04-01 | 2022-07-12 | 浙江大学 | Lithium battery residual life state evaluation method based on improved TCN |
| CN114861533A (en) * | 2022-04-26 | 2022-08-05 | 东南大学 | Wind power ultra-short-term prediction method based on time convolution network |
-
2022
- 2022-09-02 CN CN202211068209.2A patent/CN115129706B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115129706A (en) | 2022-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108563548B (en) | Abnormality detection method and apparatus | |
| CN109359698B (en) | Leakage identification method based on long-time memory neural network model | |
| CN112987675B (en) | Method, device, computer equipment and medium for anomaly detection | |
| CN111339129B (en) | Remote meter reading abnormity monitoring method and device, gas meter system and cloud server | |
| CN111027686A (en) | Landslide displacement prediction method, device and equipment | |
| CN112735094A (en) | Geological disaster prediction method and device based on machine learning and electronic equipment | |
| CN115129706B (en) | Soil moisture observation data quality evaluation method considering periodic characteristics | |
| CN110555477A (en) | municipal facility fault prediction method and device | |
| CN109063885A (en) | A kind of substation's exception metric data prediction technique | |
| CN112860676B (en) | Data cleaning method applied to big data mining and business analysis and cloud server | |
| CN108960329A (en) | A kind of chemical process fault detection method comprising missing data | |
| CN115935139A (en) | Space field interpolation method for ocean observation data | |
| CN114911788B (en) | Data interpolation method and device and storage medium | |
| CN110309947A (en) | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium | |
| CN117891644B (en) | Data acquisition system and method based on digital twin technology | |
| CN114330120B (en) | 24-Hour PM prediction based on deep neural network2.5Concentration method | |
| CN115793590A (en) | Data processing method and platform suitable for system safety operation and maintenance | |
| CN117708625B (en) | Dam monitoring historical data filling method under spent data background | |
| CN115858606A (en) | Method, device and equipment for detecting abnormity of time series data and storage medium | |
| Jian et al. | Anomaly detection and classification in water distribution networks integrated with hourly nodal water demand forecasting models and feature extraction technique | |
| CN116091874A (en) | Image verification method, training method, device, medium, equipment and program product | |
| CN109739840A (en) | Data processing empty value method, apparatus and terminal device | |
| CN114580101B (en) | Method and system for predicting residual service life of rotary machine | |
| CN117114523B (en) | Runoff forecasting model construction and runoff forecasting method based on condition mutual information | |
| CN114386814B (en) | Method and device for acquiring service radius of public service facility |
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 |