An intelligent and dynamically calibrated multi-dimensional air quality monitoring data optimization system

Through the intelligent dynamic calibration system, the matching of related factor, error prediction and dynamic calibration modules are used to solve the problems of high cost and low efficiency of traditional air quality monitoring data, and the efficient and accurate optimization of air quality monitoring data is achieved.

CN118569445BActive Publication Date: 2024-12-06SHANDONG SHENGUANG AEROSPACE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411034538.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-12-06
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

The traditional air quality monitoring data calibration methods are costly and inefficient, and cannot reflect real-time air quality in a timely and accurate manner, and cannot adapt to changes in complex environmental factors.

Method used

Design an intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system, including user side, environmental status monitor, sensor calibration components, dynamic prediction through associated element matching module, error prediction network training module, dynamic prediction module and dynamic calibration module, dynamic prediction module and calibrate sensor errors to generate more accurate air quality monitoring data.

Benefits of technology

It realizes intelligent and automated optimization of air quality monitoring data, improves the accuracy and practicality of data, reduces calibration costs, and adapts to modern environmental monitoring needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118569445B_ABST
    Figure CN118569445B_ABST
Patent Text Reader

Abstract

The present application provides a multi-dimensional air quality monitoring data optimization system with intelligent dynamic calibration, which relates to the field of multi-dimensional data processing technology. The system includes: a user end for receiving data optimization requests; an environmental status monitor for monitoring characteristic values ​​of environmental elements; a sensor calibration component, including: an associated element matching module for obtaining associated environmental elements; an error prediction network training module for training an error prediction network; a dynamic prediction module for generating system prediction errors and random prediction errors; a dynamic calibration module for obtaining air quality update monitoring data; and a data interaction component for transmitting back to the user end. The present application can solve the technical problem that the traditional calibration method is based on multiple device collections or collections from multiple devices, resulting in high calibration costs and low efficiency, and achieve the technical effect of reducing collection costs and improving the efficiency and accuracy of data calibration by dynamically predicting and calibrating sensor errors.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of multidimensional data processing, and in particular to a multidimensional air quality monitoring data optimization system with intelligent dynamic calibration. Background Art

[0002] With the continuous advancement of urbanization and industrialization, air quality monitoring technology has been widely used in the fields of environmental protection and public health. However, with the increase in monitoring needs and the complexity of the monitoring environment, traditional air quality monitoring data calibration methods are facing more and more challenges. Traditional calibration methods usually rely on multiple equipment acquisitions or calibration of data collected by multiple devices. This method is not only costly and inefficient, but also easily affected by environmental changes and equipment performance fluctuations, resulting in low calibration accuracy. In addition, traditional methods are unable to cope with complex and changeable environmental factors and cannot reflect the real-time air quality situation in a timely and accurate manner, thus affecting the credibility and practicality of monitoring data. Therefore, there is an urgent need for a new type of air quality monitoring data optimization system that can dynamically predict and calibrate sensor errors, improve the accuracy and reliability of monitoring data, and meet the needs of modern environmental monitoring. Summary of the invention

[0003] The purpose of this application is to provide a multi-dimensional air quality monitoring data optimization system with intelligent dynamic calibration, so as to solve the technical problems that the traditional calibration method is based on multiple equipment collection or collection of multiple equipment, resulting in high calibration cost and low efficiency.

[0004] In view of the above problems, the present application provides an intelligent dynamically calibrated multi-dimensional air quality monitoring data optimization system.

[0005] The present application provides a multi-dimensional air quality monitoring data optimization system with intelligent dynamic calibration, the system comprising:

[0006] A user end, the user end is used to receive an air quality monitoring data optimization request, wherein the air quality monitoring data optimization request includes air quality monitoring data and a sensor model;

[0007] Environmental status monitor, used to monitor the characteristic values ​​of environmental elements in the target area;

[0008] Sensor calibration kit, including:

[0009] A correlation element matching module, used to perform correlation analysis on the environmental element attribute set based on the sensor model to obtain the correlation environmental element;

[0010] An error prediction network training module, used for receiving the associated environmental elements, the sensor model and the quality monitoring data attributes from the associated element matching module, and training the error prediction network;

[0011] A dynamic prediction module, used for receiving the characteristic value of the environmental element from the environmental status monitor, performing error analysis on the sensor model in combination with the error prediction network, and generating a system prediction error and a random prediction error;

[0012] A dynamic calibration module, used for optimizing the air quality monitoring data according to the system prediction error and the random prediction error to obtain updated air quality monitoring data;

[0013] The data interaction component is used to receive the air quality update monitoring data and transmit it back to the user end.

[0014] Furthermore, the associated element matching module executes the steps including:

[0015] Using the environmental factor attribute set as a variable, the sensor model as a quantitative value, and the quality monitoring data attribute as a follow-up variable to perform sample mining to obtain a retrieval sample data set, wherein any retrieval sample data includes environmental factor feature record data, sensor control parameter record data, and quality monitoring data;

[0016] Based on the sensor control parameter record data, the retrieval sample data set is grouped to obtain multiple groups of retrieval sample data;

[0017] Traversing the multiple groups of search sample data to perform correlation analysis and obtain multiple groups of correlation analysis results;

[0018] The environmental element attribute sets are sorted according to the multiple groups of association analysis results to obtain the associated environmental elements.

[0019] Furthermore, the correlation element matching module executes the steps of traversing the multiple groups of search sample data to perform correlation analysis and obtain multiple groups of correlation analysis results, including:

[0020] Extracting a first set of retrieval sample data according to the multiple sets of retrieval sample data, wherein the first set of retrieval sample data includes an environmental element feature record data set and a quality monitoring data set;

[0021] De-dimensionalizing the quality monitoring data set to construct a parent data sequence;

[0022] Wherein, the environmental factor feature record data set includes a first attribute environmental factor feature record data set, a second attribute environmental factor feature record data set, and up to an Nth attribute environmental factor feature record data set;

[0023] Traversing the first attribute environment factor feature record data set, the second attribute environment factor feature record data set, and the Nth attribute environment factor feature record data set to perform dedimensionalization processing, and constructing a first attribute subsequence, a second attribute subsequence, and the Nth attribute subsequence;

[0024] Grey relational analysis is performed on the parent data sequence, the first attribute subsequence, the second attribute subsequence, and up to the Nth attribute subsequence to generate a first group of relational analysis results, which are added to the multiple groups of relational analysis results.

[0025] Furthermore, the associated element matching module performs the step of sorting the environmental element attribute sets according to the multiple groups of association degree analysis results to obtain the associated environmental elements, including:

[0026] Extracting a first attribute association degree set, a second attribute association degree set, and finally an Nth attribute association degree set according to the plurality of sets of association degree analysis results;

[0027] Traversing the first attribute association degree set, the second attribute association degree set, and the Nth attribute association degree set to perform feature fusion, and obtaining the first attribute feature association degree, the second attribute feature association degree, and the Nth attribute feature association degree, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis, and central tendency analysis;

[0028] According to the first attribute feature association degree, the second attribute feature association degree, and up to the Nth attribute feature association degree, the associated environmental elements greater than or equal to the association degree threshold are sorted from the environmental element attribute set.

[0029] Furthermore, the error prediction network training module execution steps include:

[0030] Taking the sensor model and the quality monitoring data attributes as static restrictions, taking the associated environmental factors as input variables, taking the systematic error and the random error as output variables, collecting an associated environmental factor characteristic value record data set, a first systematic error record data set, and a first random error record data set, wherein the associated environmental factor characteristic value record data set, the first systematic error record data set, and the first random error record data set correspond one to one;

[0031] Traversing the associated environmental factor characteristic value record data set, taking the sensor model and the quality monitoring data attribute as static restrictions, assigning a preset number of values ​​to the system error and the random error, and obtaining a second system error record data set and a second random error record data set, wherein the associated environmental factor characteristic value record data set is in a one-to-many relationship with the second system error record data set and the second random error record data set;

[0032] According to the second system error record data set and the second random error record data set, combined with the first system error record data set and the first random error record data set, data fusion is performed to generate a system error identification data set and a random error identification data set, wherein the system error identification data is a vector and the random error identification data is a modulus;

[0033] According to the system error identification data set, the associated environmental element feature value record data set is supervised to train a system error prediction subnetwork;

[0034] According to the random error identification data set, the associated environmental element feature value record data set is supervised to train a random error prediction subnetwork;

[0035] The input layers of the system error prediction subnetwork and the random error prediction subnetwork are fused to generate the error prediction network.

[0036] Further, the error prediction network training module performs the steps of: performing data fusion according to the second system error record data set and the second random error record data set, combining the first system error record data set and the first random error record data set, to generate a system error identification data set and a random error identification data set, including:

[0037] Extracting a plurality of system error record data of first associated environmental element characteristic value record data according to the second system error record data set and the first system error record data set;

[0038] Performing feature fusion on the plurality of system error record data to generate first system error identification data, and adding the first system error identification data to the system error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis;

[0039] Extracting a plurality of random error record data of the first associated environmental element characteristic value record data according to the second random error record data set and the first random error record data set;

[0040] Perform feature fusion on the multiple random error record data to generate first random error identification data, and add the first random error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis.

[0041] Further, the dynamic calibration module executes the steps including:

[0042] Determine whether the random prediction error is greater than or equal to a random error threshold, wherein the random error threshold is custom configured by the user end;

[0043] When the random prediction error is greater than or equal to the random error threshold, the air quality monitoring data is marked as abnormal and abnormal identification data is generated.

[0044] Further, the dynamic calibration module executes the steps including:

[0045] When the random prediction error is less than the random error threshold, the air quality monitoring data is calibrated according to the system prediction error to obtain the air quality updated monitoring data.

[0046] The technical solution provided in this application has at least the following technical effects or advantages:

[0047] The above-mentioned intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system, when the system receives an optimization request for air quality monitoring data, monitors the characteristic values ​​of environmental elements in the target area, such as temperature, humidity, wind speed, etc. Subsequently, the attribute set of environmental elements related to the sensor model is searched, and a correlation analysis is performed to determine which environmental elements are directly related to the data monitored by the current sensor. This step is to identify external factors that may affect the accuracy of the sensor. After that, the error prediction network is trained using these associated environmental elements, sensor models, and attributes of air quality monitoring data. The function of this network is to predict the error that the sensor may produce under specific environmental conditions. Then, based on the characteristic values ​​of environmental elements monitored in real time, combined with the error prediction network, an error analysis is performed on the air quality monitoring data under the current sensor model. This analysis generates a systematic prediction error and a random prediction error. The systematic prediction error is a predictable error based on environmental conditions and sensor characteristics, while the random prediction error is an error caused by uncertain factors. After obtaining the error data, the original air quality monitoring data is optimized and adjusted to eliminate or reduce the impact of these errors, thereby obtaining more accurate and reliable air quality update monitoring data. Finally, the optimized air quality monitoring data is transmitted back to the user end so that the user can understand the more accurate air quality status in real time. The whole process realizes the intelligent and automated optimization of air quality monitoring data, and improves the accuracy and practicality of the data.

[0048] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0050] Figure 1 A schematic diagram of the structure of a multi-dimensional air quality monitoring data optimization system for intelligent dynamic calibration in this application;

[0051] Figure 2 A flowchart of the execution steps of a correlation factor matching module of a multi-dimensional air quality monitoring data optimization system for intelligent dynamic calibration in this application.

[0052] Description of reference numerals:

[0053] User end 10, environmental status monitor 20, sensor calibration component 30, associated element matching module 31, error prediction network training module 32, dynamic prediction module 33, dynamic calibration module 34, data interaction component 40. DETAILED DESCRIPTION

[0054] The present application provides a multi-dimensional air quality monitoring data optimization system with intelligent dynamic calibration, which solves the technical problems of high calibration cost and low efficiency caused by traditional calibration methods because they are based on multiple device collections or collections from multiple devices. The application achieves the technical effect of reducing collection costs and improving the efficiency and accuracy of data calibration by dynamically predicting and calibrating sensor errors.

[0055] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0056] For examples, please see the attached Figure 1 The present application provides an intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system, the system specifically includes the following contents:

[0057] The user terminal 10 is used to receive an air quality monitoring data optimization request, wherein the air quality monitoring data optimization request includes air quality monitoring data and a sensor model.

[0058] Specifically, the user terminal 10 is an interactive interface of the multi-dimensional air quality monitoring data optimization system, which allows the user to submit an optimization request for air quality monitoring data. In this request, the user needs to provide some key information, including the optimized air quality monitoring data itself, and which type of sensor these data are collected by. In this way, the system terminal can perform subsequent optimization processing based on this specific information, thereby improving the accuracy and reliability of the data.

[0059] The environmental status monitor 20 is used to monitor the characteristic values ​​of environmental elements in the target area.

[0060] Specifically, the environmental status monitor 20 is a device connected to the multi-dimensional air quality monitoring data optimization system, which is used to continuously observe and record the characteristic values ​​of various environmental elements in the target area. This target area refers to the area corresponding to the air quality monitoring data. These environmental elements include temperature, humidity, noise level, light intensity, etc. The environmental status monitor 20 can collect these data in real time, obtain the characteristic values ​​of environmental elements, and provide important reference information for subsequent correlation analysis.

[0061] The sensor calibration component 30 includes: a correlation element matching module 31, which is used to perform correlation analysis on the environmental element attribute set based on the sensor model to obtain the correlated environmental element.

[0062] Specifically, the sensor calibration component 30 includes multiple modules, namely, an associated element matching module 31, an error prediction network training module 32, a dynamic prediction module 33, a dynamic calibration module 34, etc. These modules work together to complete the calibration optimization of multi-dimensional air quality monitoring data. In the associated element matching module 31, the system terminal generates an environmental element attribute set by statistically analyzing the environmental element characteristic values ​​collected by the environmental state monitor 20. This environmental element attribute set contains all environmental element categories in the environmental element characteristic values. According to the sensor model, the sensor model is analyzed to determine which environmental elements are closely associated with each other, that is, a correlation analysis is performed. This correlation analysis is based on sample mining of environmental element characteristic recording data, sensor control parameter recording data and quality monitoring data, and gray correlation calculation. The purpose is to find out environmental elements that can significantly affect the accuracy of sensor measurement, such as temperature, humidity, etc. Through correlation analysis, the system terminal can determine which environmental elements need special attention, and add these environmental elements to the associated environmental elements so that they can be considered in subsequent error predictions, thereby obtaining more accurate monitoring results.

[0063] Further, if Figure 2 As shown, the present application provides that the associated element matching module 31 performs the following steps:

[0064] Sample mining is performed with the environmental element attribute set as a variable, the sensor model as a quantitative variable, and the quality monitoring data attribute as a follow-up variable to obtain a retrieval sample data set, wherein any retrieval sample data includes environmental element feature record data, sensor control parameter record data and quality monitoring data; based on the sensor control parameter record data, the retrieval sample data set is grouped to obtain multiple groups of retrieval sample data.

[0065] In a preferred embodiment, the system terminal uses the environmental factor attribute set as a variable, which has different values ​​under different conditions. At the same time, the sensor model is used as a quantitative, that is, the sensor model is fixed and will not change, and the quality monitoring data is used as a follow-up variable, because the accuracy of the quality monitoring data is affected by the environmental factors and the sensor model. Subsequently, data mining technology is used to find samples related to these variables and quantitative data from a large amount of data. These samples include characteristic record data of environmental factors, control parameter record data of sensors when collecting data, and corresponding quality monitoring data. The environmental factor characteristic record data, sensor control parameter record data and quality monitoring data in each retrieval sample data present a corresponding relationship. Afterwards, the system terminal combines these retrieval sample data to form a retrieval sample data set. After obtaining the retrieval sample data set, the system terminal groups this data set according to the sensor control parameter record data. Because different sensor control parameters may affect the accuracy and stability of the monitoring data, the system terminal stores the retrieval sample data with the same sensor control parameters together through grouping to form multiple groups of retrieval sample data, which is helpful for subsequent correlation analysis.

[0066] The plurality of groups of search sample data are traversed to perform correlation analysis to obtain a plurality of groups of correlation analysis results; the environmental element attribute sets are sorted according to the plurality of groups of correlation analysis results to obtain the associated environmental elements.

[0067] In a preferred embodiment, after obtaining multiple groups of retrieval sample data, the system terminal will traverse the multiple groups of retrieval sample data, de-dimensionalize the environmental factor feature record data set and the quality monitoring data set in each group of retrieval sample data, and analyze the de-dimensionalization results in combination with the gray correlation degree to calculate the correlation analysis results of each group. After the traversal is completed, the system terminal obtains multiple groups of correlation analysis results, and each group of correlation analysis results reflects the degree of correlation between environmental factors and quality monitoring data under the corresponding sensor control parameters. Subsequently, the system terminal performs feature fusion on the N attribute correlation sets in each group of correlation analysis results to obtain N attribute feature correlations. Afterwards, the N attribute feature correlations are sorted according to the correlation threshold, and the corresponding associated environmental factors are extracted from the environmental factor attribute set according to the sorting results, providing a basis for subsequent error prediction.

[0068] Furthermore, the present application provides the steps of executing the correlation element matching module 31: traversing the multiple groups of search sample data to perform correlation analysis and obtain multiple groups of correlation analysis results, including:

[0069] According to the multiple groups of retrieval sample data, a first group of retrieval sample data is extracted, wherein the first group of retrieval sample data includes an environmental element feature record data set and a quality monitoring data set; the quality monitoring data set is dedimensionalized to construct a parent data sequence; wherein the environmental element feature record data set includes a first attribute environmental element feature record data set, a second attribute environmental element feature record data set, and up to an Nth attribute environmental element feature record data set.

[0070] In an optional embodiment, the system terminal sequentially traverses the first group of retrieval sample data from multiple groups of retrieval sample data as the analysis object. This group of retrieval sample data includes an environmental factor feature record data set and a quality monitoring data set. These two sets correspond to the same sensor control parameter record data. Among them, the environmental factor feature record data set records the specific characteristic values ​​of various factors affecting the environment, and the quality monitoring data set provides direct measurement data on environmental quality. Subsequently, for the convenience and accuracy of subsequent correlation analysis, the system terminal performs de-dimensionalization on the quality monitoring data set. De-dimensionalization is a data preprocessing technology that aims to eliminate the influence of different dimensions or magnitudes on data analysis, so that data from different sources or different properties can be compared and analyzed on the same scale. It is achieved by calculating the ratio of the difference between the current data and the minimum data in the quality monitoring data set to the difference between the maximum data and the minimum data. After this step, the system terminal constructs a parent data sequence, which is the basis for subsequent analysis. The environmental factor characteristic record data set includes the first attribute environmental factor characteristic record data set, the second attribute environmental factor characteristic record data set, and so on to the Nth attribute environmental factor characteristic record data set, and each attribute environmental factor characteristic record data set corresponds to a specific environmental factor attribute. This subdivision of environmental factors enables the system terminal to perform correlation analysis on each specific environmental factor attribute, so as to more accurately understand their impact on environmental quality.

[0071] The first attribute environmental factor feature record data set, the second attribute environmental factor feature record data set, and the Nth attribute environmental factor feature record data set are traversed for dedimensionalization processing to construct a first attribute subsequence, a second attribute subsequence, and the Nth attribute subsequence; grey correlation analysis is performed based on the parent data sequence, the first attribute subsequence, the second attribute subsequence, and the Nth attribute subsequence to generate a first group of correlation analysis results, which are added to the multiple groups of correlation analysis results.

[0072] In an optional implementation, after obtaining the environmental factor feature record data set, the system terminal uses the same method as the aforementioned dedimensionalization of the quality monitoring data set to dedimensionalize the first attribute environmental factor feature record data set, the second attribute environmental factor feature record data set, and the Nth attribute environmental factor feature record data set, and constructs the first attribute subsequence, the second attribute subsequence, and the Nth attribute subsequence, ensuring that data of different dimensions or magnitudes can be compared and analyzed on the same scale. Subsequently, the constructed parent data sequence and these attribute subsequences are used to perform gray correlation analysis. For the first attribute subsequence, the system terminal calculates the correlation coefficient of each data point in the parent data sequence with the data point at the same position in the first attribute subsequence. The correlation coefficient is obtained by obtaining the maximum absolute difference and the minimum absolute difference of all the same data points, summing the minimum absolute difference and the product of the resolution coefficient and the maximum absolute difference, summing the absolute difference of the current data point and the product of the resolution coefficient and the maximum absolute difference, and then calculating the ratio of the first summation result to the second summation result. Among them, the resolution coefficient is a constant between 0 and 1, which is used to control the sensitivity of the correlation coefficient to the difference. A smaller resolution coefficient will make the correlation coefficient more sensitive to the difference, and it is generally taken as 0.5. After that, the calculated correlation coefficients of each data point are summed up, and then the ratio is calculated with the number of correlation coefficients to obtain the correlation analysis results of the first attribute subsequence and the parent data sequence. Repeat the above process, the system terminal calculates the correlation analysis results of the remaining N-1 attribute subsequences, and organizes these correlation analysis results into the first group of correlation analysis results. The first group of correlation analysis results reveals the degree of influence of each environmental factor attribute in the first group of environmental factor feature record data set on the quality monitoring results. Then, the system terminal adds the first group of correlation analysis results to the set of multiple groups of correlation analysis results for subsequent sorting of environmental factor attribute sets.

[0073] Furthermore, the present application provides the steps of executing the correlation element matching module 31: sorting the environmental element attribute sets according to the multiple groups of correlation analysis results to obtain the correlated environmental elements, including:

[0074] According to the multiple groups of correlation analysis results, extract the first attribute correlation set, the second attribute correlation set, and the Nth attribute correlation set; traverse the first attribute correlation set, the second attribute correlation set, and the Nth attribute correlation set to perform feature fusion to obtain the first attribute feature correlation, the second attribute feature correlation, and the Nth attribute feature correlation, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis, and central tendency analysis; according to the first attribute feature correlation, the second attribute feature correlation, and the Nth attribute feature correlation, sort the associated environmental elements that are greater than or equal to the correlation threshold from the environmental element attribute set.

[0075] In an optional implementation, after obtaining multiple groups of correlation analysis results, the system terminal extracts the correlations of different attributes based on the results of the multiple groups of correlation analysis to form multiple correlation sets. These sets are divided according to the attributes, namely, the first attribute correlation set, the second attribute correlation set, and the Nth attribute correlation set. The elements in each set represent the degree of association between the attribute and the parent data sequence. Subsequently, feature fusion processing is performed on the correlation set of each attribute to obtain the first attribute feature correlation, the second attribute feature correlation, and the Nth attribute feature correlation. The feature fusion method includes one of mean analysis (calculating the average value), mode evaluation (finding the value with the most occurrences), median analysis (finding the median), or central tendency analysis (analyzing the overall distribution trend of the data). When the data distribution is relatively uniform, mean analysis is used; when the data distribution has an obvious peak, mode evaluation is used; when the data has extreme values, median analysis is used; when the overall characteristics of the data need to be described, central tendency analysis is used. The purpose of this step is to combine multiple correlation values ​​into a more representative feature correlation value to simplify the subsequent analysis process. Afterwards, the system terminal selects environmental elements whose attribute feature correlation is greater than or equal to the preset correlation threshold from the environmental element attribute set according to the feature correlation obtained after feature fusion of each attribute. The purpose of this step is to identify the environmental elements that are most relevant or important to the parent data sequence in order to perform subsequent error prediction.

[0076] The error prediction network training module 32 is used to receive the associated environmental elements, the sensor model and the quality monitoring data attributes from the associated element matching module 31, and train the error prediction network.

[0077] Specifically, after the system terminal completes the correlation analysis in the correlation element matching module 31, the correlation element matching module 31 will output the correlation environmental elements, sensor models and corresponding quality monitoring data attributes, and pass them to the error prediction network training module 32. In the error prediction network training module 32, the system terminal collects data from the historical error log according to the sensor model and quality monitoring data attributes to obtain the correlation environmental element characteristic value record data set, the system error identification data set and the random error identification data set. Subsequently, the error prediction network is trained with the correlation environmental element characteristic value record data set as input data and the system error identification data set and the random error identification data set as output data until the error prediction network reaches the training accuracy threshold.

[0078] Furthermore, the present application provides that the error prediction network training module 32 performs the following steps:

[0079] Taking the sensor model and the quality monitoring data attributes as static restrictions, taking the associated environmental elements as input variables, taking the systematic error and the random error as output variables, collecting the associated environmental element characteristic value record data set, the first systematic error record data set and the first random error record data set, wherein the associated environmental element characteristic value record data set, the first systematic error record data set and the first random error record data set correspond one to one; traversing the associated environmental element characteristic value record data set, taking the sensor model and the quality monitoring data attributes as static restrictions, assigning the systematic error and the random error a preset number of times, and obtaining the second systematic error record data set and the second random error record data set, wherein the associated environmental element characteristic value record data set is in a one-to-many relationship with the second systematic error record data set and the second random error record data set; performing data fusion according to the second systematic error record data set and the second random error record data set, in combination with the first systematic error record data set and the first random error record data set, generating a systematic error identification data set and a random error identification data set, wherein the systematic error identification data is a vector, and the random error identification data is a modulus.

[0080] In a preferred embodiment, the system terminal uses the sensor model and quality monitoring data attributes as static constraints, the associated environmental factors as input variables, and the system error and random error as predicted output variables. After the static constraints, input variables, and output variables are defined, the system terminal retrieves the historical error log according to the static constraints and collects the first historical error data that is the same as the static constraints. Subsequently, the historical error data is extracted according to the input variables to obtain the associated environmental factor characteristic value record data set, and then the first system error record data set and the first random error record data set that correspond to the associated environmental factor characteristic value record data set are extracted from the first historical error data. Afterwards, in order to enhance the generalization ability and prediction accuracy of the error prediction network, the system terminal traverses the associated environmental factor characteristic value record data set. During the traversal process, the sensor model and quality monitoring data attributes are continued to be the unchanged static constraints, historical error data are collected, and the second historical error data is extracted according to the traversal results. Then, the system error and the random error are assigned a preset number of times, that is, according to the business accuracy requirements, the number of secondary extractions is set to obtain a one-to-many relationship between the associated environmental element characteristic value record data set and the second system error record data set and the second random error record data set, that is, the same environmental element characteristic value corresponds to multiple error values. The preset number of assignments is to simulate the error changes that may be caused by multiple measurements under the same environmental element. After obtaining the first system error record data set, the first random error record data set, the second system error record data set, and the second random error record data set, the system terminal uses the aforementioned similar feature fusion method to fuse the first system error record data set and the second system error record data set. Since the system error is eliminable and has a clear source and directionality, before data fusion, the system terminal configures the corresponding positive and negative signs for each system error according to whether the system error makes the quality monitoring data larger or smaller. The system terminal constructs a vector for each system error based on the direction and size information of the error. The direction of the vector corresponds to the direction of the error impact, and the modulus of the vector corresponds to the size of the error. These vectors are then fused using the same feature fusion method to obtain a system error identification data set. The first random error record data set and the second random error record data set are fused. Unlike systematic errors, random errors fluctuate and are usually difficult to eliminate. Therefore, during data fusion, random errors are expressed in the form of modulus. The modulus only indicates the size of the error, without involving the direction, which reflects the disorder and unpredictability of random errors.

[0081] According to the system error identification data set, the associated environmental element characteristic value record data set is supervised to train the system error prediction subnetwork; according to the random error identification data set, the associated environmental element characteristic value record data set is supervised to train the random error prediction subnetwork; the input layers of the system error prediction subnetwork and the random error prediction subnetwork are fused to generate the error prediction network.

[0082] In a preferred embodiment, after obtaining the system error identification data set and the random error identification data set, the system terminal determines the number of layers of the neural network, the number of neurons in each layer, the activation function and other parameters according to the nature of the problem and the characteristics of the data. Subsequently, 70% of the associated environmental factor feature value record data set and the system error identification data set are used as training sets, and the remaining 30% are used as validation sets, and the weights and biases of the constructed neural network are initialized according to random numbers. After that, the neural network training parameters are set, including the learning rate, the maximum number of iterations, the batch size, the loss function, etc. The learning rate controls the update step size of the neural network parameters during the optimization process. A learning rate that is too small will result in a slow training process, and a learning rate that is too large will result in oscillation or even divergence near the optimal solution. The maximum number of iterations is the number of times the entire training set is traversed and learned. The batch size is the number of samples used to update the neural network parameters in each iteration. A smaller batch size can improve the generalization ability of the model, but will increase the training time; a larger batch size can speed up the training process, but may sacrifice a certain generalization ability. The loss function uses the mean square error (MSE) to calculate the average of the squares of the difference between the predicted value and the true value. MSE is sensitive to outliers and can better reflect the deviation between the predicted value and the actual value. Then, the system terminal inputs the training set into the neural network for forward propagation, and calculates through each layer of the neural network to obtain the predicted value. The mean square error loss function is used to calculate the loss between the predicted value and the corresponding system error identification data. According to the gradient information of the loss function, the parameters of the neural network are updated using the back propagation algorithm to minimize the loss. After each iteration, the system terminal uses the validation set to evaluate the current neural network and calculate the accuracy of the neural network. If the calculated accuracy meets the training accuracy threshold, the system terminal stops the iteration and uses the current neural network as the system error prediction subnetwork. Otherwise, the next round of iterative training is performed. If the number of iterations reaches the maximum number of iterations and still does not meet the requirements of the training accuracy threshold, the system terminal adjusts the neural network, for example, optimizes the network structure, adjusts the training parameters, etc., to improve the prediction accuracy and generalization ability of the model.

[0083] The system terminal uses the same training method to train the random error prediction subnetwork based on the random error identification data set and the associated environmental element characteristic value record data set. Subsequently, the input layers of the obtained system error prediction subnetwork and random error prediction subnetwork are fused, that is, they share the same input. In the fused neural network, the two subnetworks are responsible for predicting the system error and random error respectively, and output their respective prediction results. In this way, the system terminal obtains a complete error prediction network, which can simultaneously predict the system error and random error under given environmental elements, providing a basis for subsequent dynamic prediction and dynamic calibration.

[0084] Further, the present application provides the error prediction network training module 32 execution steps: according to the second system error record data set and the second random error record data set, combined with the first system error record data set and the first random error record data set, data fusion is performed to generate a system error identification data set and a random error identification data set, including:

[0085] According to the second system error record data set and the first system error record data set, extract multiple system error record data of the first associated environmental element characteristic value record data; perform feature fusion on the multiple system error record data to generate first system error identification data, and add it to the system error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis; according to the second random error record data set and the first random error record data set, extract multiple random error record data of the first associated environmental element characteristic value record data; perform feature fusion on the multiple random error record data to generate first random error identification data, and add it to the random error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis.

[0086] In an optional embodiment, the system terminal extracts the first associated environmental element characteristic value record data from the associated environmental element characteristic value record data as the first associated environmental element characteristic value record data, and screens out multiple system error records related to the first associated environmental element characteristic value record data from the second system error record data set and the first system error record data set. Subsequently, the screened multiple system error records are subjected to the same vectorization process as described above, and the first system error identification data is obtained based on the vectorization results using the same feature fusion method as described above, and the generated first system error identification data is added to the system error identification data set for subsequent use in neural network training. The system terminal uses the same method to extract multiple random error record data from the second random error record data set and the first random error record data set from the first associated environmental element characteristic value record data, and directly uses the same feature fusion method as described above to fuse the multiple random error record data to obtain the first random error identification data, which is added to the random error identification data set.

[0087] The dynamic prediction module 33 is used to receive the characteristic value of the environmental element from the environmental status monitor 20, perform error analysis on the sensor model in combination with the error prediction network, and generate a system prediction error and a random prediction error.

[0088] Specifically, in the dynamic prediction module 33, the system terminal receives the environmental element characteristic values ​​collected from the environmental status monitor 20 and activates the constructed error prediction network. Subsequently, the obtained environmental element characteristic values ​​are input into the error prediction network, and the error prediction network performs error analysis on the current sensor model according to the internal system error prediction subnetwork and random error prediction subnetwork to generate system prediction error and random prediction error.

[0089] The dynamic calibration module 34 is used to optimize the air quality monitoring data according to the system prediction error and the random prediction error to obtain updated air quality monitoring data.

[0090] Specifically, in the dynamic calibration module 34, after obtaining the system prediction error and the random prediction error, the system terminal judges the random prediction error according to the random error threshold. If the random prediction error meets the random error threshold, the system terminal calibrates the air quality monitoring data according to the system prediction error, that is, the system prediction error is used to perform corresponding addition and subtraction calculations on the air quality monitoring data to obtain the air quality update monitoring data. These data air quality update monitoring data have removed the errors caused by sensor performance deviations, and are therefore closer to the actual air quality conditions.

[0091] Further, the present application provides that the steps performed by the dynamic calibration module 34 include:

[0092] Determine whether the random prediction error is greater than or equal to a random error threshold, wherein the random error threshold is custom configured by the user terminal 10; when the random prediction error is greater than or equal to the random error threshold, mark the air quality monitoring data as abnormal and generate abnormal marking data.

[0093] In a preferred embodiment, the system terminal determines whether the random prediction error has reached a user-defined random error threshold. This random error threshold is a standard for measuring whether the random error is too large, so that it may affect the accuracy of the air quality monitoring data. If the random prediction error is greater than or equal to this random error threshold, the system terminal determines that there is a large uncertainty or error in the air quality monitoring data, and it may no longer be suitable for direct analysis of air quality. Therefore, the system terminal will mark this set of air quality monitoring data as abnormal, generate corresponding abnormal identification data, and send it to the user. The purpose of the abnormal identification is to make the user aware that there may be problems with the air quality monitoring data and that it needs to be used with caution. The user can decide whether to continue using this set of data based on this identification. In this way, the system terminal not only ensures the transparency of the data, but also gives the user the right to choose, making the use of data more flexible and safe.

[0094] Further, the present application provides that the steps performed by the dynamic calibration module 34 include:

[0095] When the random prediction error is less than the random error threshold, the air quality monitoring data is calibrated according to the system prediction error to obtain the air quality updated monitoring data.

[0096] In an optional embodiment, when the random prediction error is less than a user-defined random error threshold, it represents that the random error in the air quality monitoring data is within an acceptable range, and therefore, the air quality monitoring data is considered to be reliable. In this case, the system terminal will use the system prediction error to calibrate the air quality monitoring data to eliminate or reduce the inaccuracy caused by the system error. After calibration, the air quality monitoring data is updated to be closer to the actual environmental conditions, and these updated data are used as air quality update monitoring data, which are more accurate and reliable and can provide users with valuable reference information. Ultimately, these calibrated data will be sent to the user for further analysis, decision-making or application.

[0097] The data interaction component 40 is used to receive the air quality update monitoring data and transmit it back to the user terminal 10.

[0098] Specifically, in the data interaction component 40, after the calibration and update of the air quality monitoring data are completed, the system terminal sends the air quality update monitoring data to the user terminal 10. This return process ensures that the user can obtain optimized and calibrated, more accurate and reliable air quality data in a timely manner.

[0099] In summary, the intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system provided by this application has the following technical effects:

[0100] The multi-dimensional air quality monitoring data optimization system provided by the present application includes a user terminal 10, an environmental state monitor 20, a sensor calibration component 30 and a data interaction component 40. The user terminal 10 receives an air quality monitoring data optimization request including air quality monitoring data and a sensor model. The environmental state monitor 20 monitors the characteristic values ​​of environmental elements in the target area. The sensor calibration component 30 includes an associated element matching module 31, an error prediction network training module 32, a dynamic prediction module 33 and a dynamic calibration module 34. The associated element matching module 31 performs a correlation analysis based on the sensor model to obtain associated environmental elements; the error prediction network training module 32 trains the error prediction network; the dynamic prediction module 33 combines the environmental element characteristic values ​​and the error prediction network to generate a system prediction error and a random prediction error; the dynamic calibration module 34 optimizes the air quality monitoring data according to the prediction error. The data interaction component 40 is used to transmit the updated air quality monitoring data back to the user terminal 10. In addition, the associated element matching module 31 performs sample mining and correlation analysis through the environmental element attribute set, the sensor model and the quality monitoring data attributes to obtain the associated environmental elements. The error prediction network training module 32 trains the system error prediction subnetwork and the random error prediction subnetwork by associating the environmental factor characteristic value record data set, the system error and the random error record data set, and finally generates the error prediction network. The dynamic calibration module 34 optimizes the air quality monitoring data according to the system prediction error and the random prediction error. When the random prediction error exceeds the threshold, the data is marked as abnormal; otherwise, the data is calibrated according to the system prediction error to obtain updated monitoring data. These technical effects jointly solve the technical problem that the traditional calibration method is based on multiple equipment collection or multiple equipment collection, resulting in high calibration cost and low efficiency, and achieves the technical effect of reducing the collection cost and improving the efficiency and accuracy of data calibration by dynamically predicting and calibrating sensor errors.

[0101] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0102] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. An intelligent dynamically calibrated multi-dimensional air quality monitoring data optimization system, characterized in that: include: A user end, the user end is used to receive an air quality monitoring data optimization request, wherein the air quality monitoring data optimization request includes air quality monitoring data and a sensor model; Environmental status monitor, used to monitor the characteristic values ​​of environmental elements in the target area; Sensor calibration kit, including: A correlation element matching module, used to perform correlation analysis on the environmental element attribute set based on the sensor model to obtain the correlation environmental element; An error prediction network training module, used for receiving the associated environmental elements, the sensor model and the quality monitoring data attributes from the associated element matching module, and training the error prediction network; A dynamic prediction module, used for receiving the characteristic value of the environmental element from the environmental status monitor, performing error analysis on the sensor model in combination with the error prediction network, and generating a system prediction error and a random prediction error; A dynamic calibration module, used for optimizing the air quality monitoring data according to the system prediction error and the random prediction error to obtain updated air quality monitoring data; A data interaction component is used to receive the air quality update monitoring data and transmit it back to the user end; The steps of executing the association element matching module include: Using the environmental factor attribute set as a variable, the sensor model as a quantitative value, and the quality monitoring data attribute as a follow-up variable to perform sample mining to obtain a retrieval sample data set, wherein any retrieval sample data includes environmental factor feature record data, sensor control parameter record data, and quality monitoring data; Based on the sensor control parameter record data, the retrieval sample data set is grouped to obtain multiple groups of retrieval sample data; Traversing the multiple groups of search sample data to perform correlation analysis and obtain multiple groups of correlation analysis results; The environmental element attribute sets are sorted according to the multiple groups of association analysis results to obtain the associated environmental elements.

2. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 1, characterized in that: The correlation element matching module performs the following steps: traversing the multiple groups of search sample data to perform correlation analysis and obtain multiple groups of correlation analysis results, including: Extracting a first set of retrieval sample data according to the multiple sets of retrieval sample data, wherein the first set of retrieval sample data includes an environmental element feature record data set and a quality monitoring data set; De-dimensionalizing the quality monitoring data set to construct a parent data sequence; Wherein, the environmental factor feature record data set includes a first attribute environmental factor feature record data set, a second attribute environmental factor feature record data set, and up to an Nth attribute environmental factor feature record data set; Traversing the first attribute environment factor feature record data set, the second attribute environment factor feature record data set, and the Nth attribute environment factor feature record data set to perform dedimensionalization processing, and constructing a first attribute subsequence, a second attribute subsequence, and the Nth attribute subsequence; Grey relational analysis is performed on the parent data sequence, the first attribute subsequence, the second attribute subsequence, and up to the Nth attribute subsequence to generate a first group of relational analysis results, which are added to the multiple groups of relational analysis results.

3. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 1, characterized in that: The associated element matching module performs the steps of: sorting the environmental element attribute sets according to the multiple groups of association degree analysis results to obtain the associated environmental elements, including: Extracting a first attribute association degree set, a second attribute association degree set, and finally an Nth attribute association degree set according to the plurality of sets of association degree analysis results; Traversing the first attribute association degree set, the second attribute association degree set, and the Nth attribute association degree set to perform feature fusion, and obtaining the first attribute feature association degree, the second attribute feature association degree, and the Nth attribute feature association degree, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis, and central tendency analysis; According to the first attribute feature association degree, the second attribute feature association degree, and up to the Nth attribute feature association degree, the associated environmental elements greater than or equal to the association degree threshold are sorted from the environmental element attribute set.

4. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 1, characterized in that: The error prediction network training module execution steps include: Taking the sensor model and the quality monitoring data attributes as static restrictions, taking the associated environmental factors as input variables, taking the systematic error and the random error as output variables, collecting an associated environmental factor characteristic value record data set, a first systematic error record data set, and a first random error record data set, wherein the associated environmental factor characteristic value record data set, the first systematic error record data set, and the first random error record data set correspond one to one; Traversing the associated environmental factor characteristic value record data set, taking the sensor model and the quality monitoring data attribute as static restrictions, assigning a preset number of values ​​to the system error and the random error, and obtaining a second system error record data set and a second random error record data set, wherein the associated environmental factor characteristic value record data set is in a one-to-many relationship with the second system error record data set and the second random error record data set; According to the second system error record data set and the second random error record data set, combined with the first system error record data set and the first random error record data set, data fusion is performed to generate a system error identification data set and a random error identification data set, wherein the system error identification data is a vector and the random error identification data is a modulus; According to the system error identification data set, the associated environmental element feature value record data set is supervised to train a system error prediction subnetwork; According to the random error identification data set, the associated environmental element feature value record data set is supervised to train a random error prediction subnetwork; The input layers of the system error prediction subnetwork and the random error prediction subnetwork are fused to generate the error prediction network.

5. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 4, characterized in that: The error prediction network training module performs the following steps: performing data fusion according to the second system error record data set and the second random error record data set, in combination with the first system error record data set and the first random error record data set, to generate a system error identification data set and a random error identification data set, including: Extracting a plurality of system error record data of first associated environmental element characteristic value record data according to the second system error record data set and the first system error record data set; Performing feature fusion on the plurality of system error record data to generate first system error identification data, and adding the first system error identification data to the system error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis; Extracting a plurality of random error record data of the first associated environmental element characteristic value record data according to the second random error record data set and the first random error record data set; Perform feature fusion on the multiple random error record data to generate first random error identification data, and add the first random error identification data set, wherein the feature fusion is one of mean analysis, mode evaluation, median analysis and central tendency analysis.

6. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 1, characterized in that: The dynamic calibration module execution steps include: Determine whether the random prediction error is greater than or equal to a random error threshold, wherein the random error threshold is custom configured by the user end; When the random prediction error is greater than or equal to the random error threshold, the air quality monitoring data is marked as abnormal and abnormal identification data is generated.

7. The intelligent dynamic calibration multi-dimensional air quality monitoring data optimization system according to claim 6, characterized in that: The dynamic calibration module performs the following steps: When the random prediction error is less than the random error threshold, the air quality monitoring data is calibrated according to the system prediction error to obtain the air quality updated monitoring data.