Environment monitoring method and system based on data fusion technology

Abstract

The invention discloses an environment monitoring method and system based on a data fusion technology, and belongs to the technical field of environment monitoring, and the method comprises the steps: synchronously collecting different types of environment characteristic data at a fixed sampling period, and forming an original data set; preprocessing the original data set; performing data fusion on the same type of environment feature data in the preprocessed original data set to generate a single type of state evaluation value; generating a joint feature matrix according to the single-type state evaluation value; establishing an environment quality evaluation model according to the joint feature matrix, and generating an environment quality index; monitoring the environment according to the environment quality index; according to the method, the problems of low data reliability and insufficient multi-source fusion of a traditional method are solved by synchronously collecting environment feature data, preprocessing and removing abnormal values, fusing and generating a single-type evaluation value and a combined feature matrix and establishing a dynamic weight model, the data reliability is improved, the multi-source data fusion capability is enhanced, and the method is suitable for large-scale popularization and application. Therefore, the accuracy of environment monitoring is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of environmental monitoring, and particularly relates to an environmental monitoring method and system based on data fusion technology. BACKGROUND

[0002] With the continuous improvement of environmental protection requirements and the increasing concern for ecological health, the demand for accurate, comprehensive and real-time monitoring of environmental conditions has become increasingly urgent.

[0003] Currently, environmental monitoring mainly relies on deploying various sensor networks to collect heterogeneous data such as temperature, humidity, and pollutant concentration, and usually uses threshold-based judgment or simple statistical average methods for data processing and quality assessment.

[0004] However, the existing monitoring methods have significant shortcomings: first, the original sensor data is easily affected by noise and interference, and traditional preprocessing methods are difficult to effectively eliminate outliers, resulting in low data reliability and affecting the evaluation basis; second, the processing of different types of environmental parameters is often conducted in isolation or insufficiently integrated, failing to fully exploit the temporal and spatial correlations and complementary information among multi-source data, leading to one-sidedness in comprehensive environmental state assessment and reducing the accuracy of environmental monitoring; in addition, the evaluation model generally uses fixed weight coefficients, which cannot dynamically adjust the contribution of each parameter according to environmental conditions, resulting in poor adaptability and reduced accuracy of the evaluation results in complex and variable scenarios. SUMMARY

[0005] To address the shortcomings of the prior art, the present application provides an environmental monitoring method and system based on data fusion technology, which solves the above problems.

[0006] To achieve the above purpose, the present application is implemented by the following technical solution: an environmental monitoring method based on data fusion technology, comprising the following steps:

[0007] Synchronously collecting different types of environmental characteristic data at a fixed sampling period to form an original data set;

[0008] Preprocessing the original data set;

[0009] Data fusion on the same type of environmental characteristic data in the preprocessed original data set to generate a single-type state evaluation value;

[0010] Generating a joint feature matrix according to the single-type state evaluation value;

[0011] Establishing an environmental quality evaluation model according to the joint feature matrix to generate an environmental quality index;

[0012] Monitoring the environment according to the environmental quality index.

[0013] On the basis of the above technical solutions, the application further provides the following optional technical solutions.

[0014] Further technical solutions: the pre-processing manner of the original data set specifically includes:

[0015] According to the single environmental feature data in the original data set, a data validity condition value is generated;

[0016] Specifically, the data validity condition value K is generated through the formula:

[0017] ;

[0018] The data validity condition value K is generated;

[0019] In the formula, x i represents the single environmental feature data in the original data set, μ represents the mean value of all data of the same type as the environmental feature data x i ;

[0020] According to the data validity condition value K, the environmental feature data in the original data set is removed.

[0021] Further technical solutions: the generation manner of the single-type state evaluation value specifically includes:

[0022] Through the formula:

[0023]

[0024] The single-type state evaluation value P k is generated;

[0025] In the formula, P k-1 represents the single-type state evaluation value at k-1 moment, u k represents the control input vector at k moment, Z k represents the single-type data vector at k moment, K k represents the Kalman gain matrix at k moment, A represents the state transition matrix, B represents the control input matrix, and H represents the observation matrix.

[0026] Further technical solutions: the generation manner of the joint feature matrix specifically includes:

[0027] Through the formula:

[0028] ;

[0029] The joint feature matrix F is generated;

[0030] In the formula, T represents an ambient temperature state evaluation value in the single-type state evaluation value, H represents an ambient humidity state evaluation value in the single-type state evaluation value, IAQI (G CO2 ) represents an air quality sub-index, G CO2 represents an air quality state evaluation value in the single-type state evaluation value, log (PM 2.5 +1) represents a log-transformed value of PM 2.5 concentration, PM 2.5 represents a PM 2.5 state evaluation value in the single-type state evaluation value, M n represents the remaining single-type state evaluation values in the single-type state evaluation value.

[0031] Further technical solutions: the expression of the environment quality evaluation model is specifically as follows:

[0032] ;

[0033] wherein, EQI represents an environment quality index, W j represents a dynamic weight coefficient, F j represents a normalized feature value of data in a joint feature matrix, j represents an environment feature data type, and n is the number of environment feature data types.

[0034] Further technical solutions: the generation manner of the dynamic weight coefficient W j is specifically as follows:

[0035] acquiring historical environment feature data;

[0036] establishing a dynamic adjustment model according to the historical environment feature data to generate the dynamic weight coefficient W j .

[0037] Further technical solutions: the expression of the dynamic adjustment model is specifically as follows:

[0038] ;

[0039] wherein, σ J 2 represents a variance of the jth historical data, σ b 2 represents a variance of the jth historical data, b represents a total number of data features participating in fusion, and j is an index of the total number of data features participating in fusion.

[0040] An environment monitoring system based on a data fusion technology, the system comprising:

[0041] A data acquisition unit is configured to synchronously acquire different types of environmental characteristic data at a fixed sampling period to form a raw data set;

[0042] A data processing unit is configured to preprocess the raw data set;

[0043] A data preliminary fusion unit is configured to perform data fusion on the same type of environmental characteristic data in the preprocessed raw data set to generate a single-type state evaluation value;

[0044] A feature fusion unit is configured to generate a joint feature matrix according to the single-type state evaluation value;

[0045] A quality analysis unit is configured to establish an environmental quality evaluation model according to the joint feature matrix to generate an environmental quality index;

[0046] A monitoring unit is configured to monitor the environment according to the environmental quality index.

[0047] Further technical solutions: the data processing unit specifically comprises:

[0048] An effective analysis module is configured to generate a data effective condition value according to a single environmental characteristic data in the raw data set;

[0049] A rejection module is configured to reject the environmental characteristic data in the raw data set according to the data effective condition value.

[0050] Further technical solutions: the quality analysis unit comprises:

[0051] A historical data acquisition module is configured to acquire historical environmental characteristic data;

[0052] A result output module is configured to establish a dynamic adjustment model according to the historical environmental characteristic data to generate a dynamic weight coefficient.

[0053] The present application provides an environmental monitoring method and system based on data fusion technology, which has the following advantages compared with the prior art:

[0054] The present application synchronously acquires environmental characteristic data, preprocesses and rejects abnormal values, generates a single-type evaluation value and a joint feature matrix, and establishes a dynamic weight model, thereby solving the problems of low data reliability and insufficient multi-source fusion of the traditional method, improving the data reliability and enhancing the multi-source data fusion capability, and thus improving the accuracy of environmental monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A flowchart of an environmental monitoring method based on data fusion technology is provided for the embodiments of the present application.

[0056] Figure 2The flow block diagram of S2 provided for the embodiment of the present application.

[0057] Figure 3 The flow block diagram of the data elimination mode in S2 provided for the embodiment of the present application.

[0058] Figure 4 The flow block diagram of S5 provided for the embodiment of the present application.

[0059] Figure 5 The structural schematic diagram of an environmental monitoring system based on data fusion technology provided for the embodiment of the present application.

[0060] Figure 6 The module block diagram of a data processing unit provided for the embodiment of the present application.

[0061] Figure 7 The module block diagram of a quality analysis unit provided for the embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to make the objects, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0063] In the prior art, environmental monitoring mainly relies on sensor networks to collect heterogeneous data such as temperature, humidity, and pollutant concentration, and adopts threshold judgment or simple statistical average method for quality evaluation. The existing method has defects such as insufficient data preprocessing leading to abnormal value interference on evaluation results, isolated processing of multi-source data causing one-sided evaluation, and fixed weight model being difficult to adapt to complex scene changes. For example, in the industrial park environmental monitoring scene, the PM2.5 data collected by the sensor often fluctuates abnormally due to equipment interference, and the traditional method directly eliminates or retains the abnormal data, which leads to the deviation of the subsequent evaluation results from the true pollution level; at the same time, the correlation between temperature, humidity and pollutant concentration has not been effectively mined, and each parameter is analyzed independently during evaluation, which cannot reflect the overall change of the environmental state; in addition, the fixed weight model cannot dynamically adjust the contribution of the temperature parameter in seasons with large diurnal temperature difference, resulting in the decrease of the accuracy of the evaluation results.

[0064] To solve the above problems, first of all, it is found that the traditional data preprocessing only relies on simple threshold filtering, and cannot distinguish between real environment mutation and device noise. By analyzing the distribution characteristics of sensor data, a preprocessing method based on statistical outlier detection is proposed. Secondly, in view of the problem of insufficient multi-source data fusion, it is realized that different types of environmental parameters have spatio-temporal correlation, and a unified feature expression framework needs to be established. Finally, in view of the limitation of fixed weight model, a dynamic adjustment mechanism based on data volatility is proposed. The solution ideas of the three problems finally form a complete technical chain including data cleaning, feature fusion and dynamic evaluation.

[0065] Therefore, the present application proposes an environment monitoring method comprising the following steps, and the specific implementation of the present application is described in detail below in combination with specific embodiments.

[0066] Please refer to Figure 1 The present application proposes an environment monitoring method based on data fusion technology, comprising the following steps:

[0067] S1: Synchronously collecting different types of environmental feature data with a fixed sampling period to form an original data set;

[0068] S2: Preprocessing the original data set;

[0069] S3: Data fusion of the same type of environmental feature data in the preprocessed original data set to generate a single type state evaluation value;

[0070] S4: Generating a joint feature matrix according to the single type state evaluation value;

[0071] S5: Establishing an environmental quality evaluation model according to the joint feature matrix to generate an environmental quality index;

[0072] S6: Monitoring the environment according to the environmental quality index;

[0073] Among them, the fixed sampling period synchronous collection means that all sensors collect data according to a unified time reference, which can realize clock synchronization by using a GPS time module, so as to ensure that the temperature, humidity and pollutant concentration data have the same time stamp;

[0074] Data preprocessing means identifying abnormal data by statistical method, which can be realized by calculating the absolute deviation of data points from the mean value, and when the deviation exceeds the set multiple standard deviation, it is determined as an outlier;

[0075] Single type state evaluation value means fusion calculation of time series data of the same type of environmental parameters, which can be realized by using Kalman filter algorithm, and the evaluation value at the current time is dynamically corrected by state transition matrix and observation matrix;

[0076] The joint feature matrix refers to the standardized features integrating multiple types of environmental parameters, and can specifically combine the temperature evaluation value, the humidity evaluation value, the pollutant sub-index, etc. into a multi-dimensional matrix in a vector splicing manner.

[0077] The environmental quality evaluation model refers to a linear combination model based on dynamic weights, and can specifically calculate the weight coefficients of each parameter by using the inverse variance weighting method.

[0078] Specifically, the time alignment of multi-source data is ensured by the clock synchronization module, laying a foundation for subsequent fusion; in the preprocessing stage, the absolute deviation of each data point from the mean value of the same type of data is calculated, and when the deviation exceeds three standard deviations, it is determined as an abnormal value and is eliminated, effectively eliminating the influence of device noise; the Kalman filtering algorithm is applied to the cleaned temperature data sequence, the evaluation value at the previous moment and the current observation value are combined, and the current temperature state evaluation value is dynamically updated through the state transition matrix and the Kalman gain matrix, thereby enhancing the data continuity; after the evaluation values of various parameters are normalized, a joint feature matrix containing temperature, humidity, PM2.5 logarithmic concentration and CO2 sub-index is spliced; the variances of various parameters are calculated according to historical data, and the feature matrix is weighted and summed with the inverse variance as the weight, thereby finally generating an environmental quality index that automatically adjusts with data volatility.

[0079] Compared with the prior art, the traditional method uses independent sensor data processing, resulting in single evaluation dimension, while the present application integrates multi-dimensional environmental state information through a joint feature matrix; the prior art uses a fixed threshold to eliminate abnormal data, which may mistakenly delete real mutation values, while the present application more accurately identifies noise data based on a dynamic monitoring method of statistical distribution; the conventional evaluation model uses artificial setting of weights, which is difficult to adapt to scene changes, while the present application automatically calculates the weight coefficients through data variance, so that the model can still maintain evaluation accuracy when the concentration of pollutants suddenly increases or the temperature and humidity fluctuate sharply.

[0080] Through the above technical solutions, the present application effectively cleans the sensor data, eliminating the influence of abnormal values on the evaluation results; establishes a correlation analysis framework for multi-source environmental parameters, improving the comprehensiveness of comprehensive evaluation; and constructs a self-adaptive weight adjustment mechanism, enhancing the robustness of the model under different environmental conditions; specifically, it accurately identifies and eliminates abnormal PM2.5 data in the presence of device interference; dynamically increases the evaluation weight of the temperature parameter in an environment with large diurnal temperature difference; and accurately reflects the deterioration trend of environmental quality in a sudden pollution event through multi-parameter joint analysis.

[0081] Please refer to Figure 2 and Figure 3 The present application further proposes a way of preprocessing the original data set, which specifically includes:

[0082] S2.1: generating a data validity condition value according to single environmental characteristic data in the original data set;

[0083] Specifically, the data validity condition value K is generated by the formula:

[0084]

[0085]

[0086] In the formula, x i represents single environmental characteristic data in the original data set, and μ represents the mean value of all data of the same type as the environmental characteristic data x i

[0087] S2.2: removing environmental characteristic data in the original data set according to the data validity condition value K;

[0088] The data validity condition value K refers to the absolute difference value of single environmental characteristic data and the mean value of data of the same type, which can be realized by calculating the absolute deviation of each data point from the mean value of the whole data to quantify the difference between the data and the group distribution.

[0089] The mean value μ of data of the same type refers to the average value of the same environmental parameter in the same time window, which can be obtained by using sliding average or batch calculation, and reflects the central tendency of the current data set.

[0090] The data removal mechanism refers to screening abnormal data according to the K value, which can be realized by setting a fixed threshold or dynamically adjusting the removal rule to ensure the reliability of the data set by excluding data points deviating too much.

[0091] Specifically, in the data preprocessing stage, for each collected environmental characteristic data, the absolute difference value between it and the mean value of data of the same type is calculated to obtain the corresponding K value. The larger the value, the higher the degree of deviation of the data point from the group distribution, and the greater the possibility of abnormality. By setting a reasonable K value threshold or using a dynamic adjustment rule, for example, data more than twice the standard deviation of the mean value is considered as an abnormal value, the system automatically removes such data from the original data set. This process avoids the defect that the fixed threshold in the traditional method cannot adapt to the change of data distribution, and overcomes the problem that simple statistical average is sensitive to abnormal values, ensuring that the preprocessed data has higher consistency and reliability, laying a foundation for subsequent fusion and evaluation.

[0092] ​​​Compared with the prior art, the traditional preprocessing method usually adopts a fixed threshold or abnormality detection based on statistical average, and cannot effectively cope with dynamic changes in data distribution, and is prone to cause false rejection or missed rejection; and the application can more accurately identify abnormal values and improve the flexibility and accuracy of data screening by dynamically calculating the absolute difference between the data points and the mean value and combining an adaptive rejection mechanism.

[0093] Through the above technical solutions, the application realizes effective identification and rejection of abnormal values in original environmental feature data, and significantly improves the reliability of the data preprocessing stage. By dynamically quantifying the degree of data deviation and adaptively adjusting the rejection rule, the false judgment problem caused by the fixed threshold or simple average of the traditional method is avoided, and the data set relied on subsequent data fusion and quality evaluation has higher accuracy and consistency.

[0094] The application further proposes a generation method of a single-type state evaluation value, which specifically includes:

[0095] Through the formula:

[0096]

[0097] Generate a single-type state evaluation value P k ;

[0098] In the formula, P k-1 represents the single-type state evaluation value at k-1 time, u k represents the control input vector at k time, Z k represents the single-type data vector at k time, K k represents the Kalman gain matrix at k time, A represents the state transition matrix, B represents the control input matrix, and H represents the observation matrix.

[0099] The state transition matrix A refers to a matrix describing the dynamic relationship of the single-type environmental state changing over time, and can be realized by using the state transition parameters in the linear system model, for transmitting the evaluation value at k-1 time to the current time and retaining the time sequence correlation of historical data.

[0100] The control input vector u k refers to an input variable reflecting external intervention or environmental condition change, and can be realized by a control signal collected by a sensor or a preset adjustment parameter, for introducing the influence of external factors on the current state.

[0101] The Kalman gain matrix K k is a matrix for dynamically adjusting the weight of observation data according to the real-time data noise characteristics, and can be calculated through the covariance matrix in the Kalman filtering algorithm, for balancing the credibility of historical prediction values and current observation values and suppressing noise interference.

[0102] The observation matrix H refers to a conversion matrix for mapping state variables to an observation space, which can be realized by using parameters in a linear observation model, and is used for comparing historical state prediction values with actual observation data to generate a residual correction term.

[0103] Specifically, the technical scheme realizes dynamic optimization and fusion of the same type of environmental characteristic data through a recursive formula. First, the state transition matrix A is used to pass the single-type state evaluation value P k-1 at the k-1 time to the current time, retaining the time sequence continuity of the environmental state; second, the external intervention variable u k is introduced through the control input matrix B, and finally the optimal evaluation value P k that takes into account the historical state and real-time observation is generated through the triple action of state transition, control input and error correction.

[0104] Compared with the prior art, the traditional method usually uses a sliding average or a fixed coefficient to fuse the same type of data, which cannot distinguish the dynamic correlation between the historical data and the current observation, and is prone to evaluation deviation when the sensor noise fluctuates; while the present application can optimize the fusion weight in real time according to the data quality through the dynamic adjustment mechanism of the Kalman gain matrix.

[0105] Through the above technical scheme, the present application effectively solves the problem of insufficient correlation between the historical state and real-time observation in the fusion of the same type of environmental characteristic data, and realizes dynamic optimization and fusion in complex scenes such as sensor noise interference and external condition mutation; specifically, the single-type state evaluation value can reflect both the time sequence evolution law and the real-time change characteristics of the environmental parameters; through the adaptive adjustment of the Kalman gain matrix, the interference of abnormal data on the evaluation result is suppressed; the introduction of the control input vector enables the external intervention factors to be quantitatively fused, enhancing the environmental adaptability of the evaluation model.

[0106] The present application further proposes a generation method of a joint feature matrix, which specifically includes:

[0107] Through the formula:

[0108] ;

[0109] to generate a joint feature matrix F;

[0110] In the formula, T represents the environmental temperature state evaluation value in the single-type state evaluation value, H represents the environmental humidity state evaluation value in the single-type state evaluation value, IAQI(G CO2 ) represents the air quality index, G CO2 represents the air quality state evaluation value in the single-type state evaluation value, and log(PM 2.5+1) represents PM 2.5 logarithmic transformation value of the concentration, PM 2.5 represents PM in the single-type state evaluation value 2.5 state evaluation value, M n represents the remaining single-type state evaluation value in the single-type state evaluation value;

[0111] The environmental temperature state evaluation value T refers to the standardized evaluation result obtained by preprocessing and fusing temperature sensor data, which can be realized by using sliding average filtering and normalization processing after collecting data by the temperature sensor, and is used to represent the influence degree of the environmental temperature on the overall quality.

[0112] The environmental humidity state evaluation value H refers to the quantitative index obtained by removing outliers and data fusion of humidity sensor data, which can be realized by using Kalman filtering algorithm to denoise the original humidity data, and is used to reflect the dynamic change of the humidity parameter in the environmental state. CO2 IAQI (G is a standardized expression of the CO2 concentration data converted into air quality index, which can be realized by using piecewise linear interpolation method to map the CO2 concentration to the preset air quality index interval, and is used to eliminate the influence of different pollution dimension differences on the fusion process.

[0113] log(PM 2.5 +1) is a characteristic value obtained by nonlinear transformation of PM 2.5 concentration data, which can be realized by using logarithmic function to smooth the original concentration value, and is used to suppress the interference of high concentration outliers on model training.

[0114] The remaining single-type state evaluation value M n refers to the fusion result of other environmental parameters except temperature, humidity, CO2 and PM 2.5 , which can be generated by using the same data fusion algorithm as T and H, and is used to expand the dimension of the feature matrix to cover more environmental impact factors.

[0115] Specifically, when generating the joint feature matrix, first, the temperature and humidity evaluation values are directly included in the matrix as the basic physical quantity, retaining their original magnitude characteristics; for CO2 concentration data, it is converted into a standard air quality evaluation system by using exponential conversion, solving the problem of inconsistent evaluation standards of different pollutants; for PM 2.5The concentration is logarithmically transformed, which reduces the negative influence of extremely high values on the model while retaining the trend of concentration change; and finally, other parameter fusion results are integrated to form a multi-dimensional feature vector. Through the combination of exponential conversion, nonlinear transformation and multi-dimensional expansion, originally isolated environmental parameters form a quantifiable and comparable correlation in a unified feature space, providing multi-source data input for subsequent models that contain physical quantities, standardized pollutant indicators and derived features;

[0116] Compared with the prior art, the traditional method usually processes parameters such as temperature, humidity and pollutant concentration separately or simply splices them, without considering the dimensional difference and nonlinear relationship between data, which makes it difficult for the model to accurately capture the synergistic effect of multiple parameters. The present application effectively fuses the evaluation results of each parameter in the same matrix by constructing a joint feature matrix, significantly improving the correlation and comparability between multi-source data.

[0117] Through the above technical solution, the present application can express the features of temperature, humidity, concentrations of multiple pollutants and other environmental parameters in the same multi-dimensional space, solve the one-sided evaluation problem caused by isolated processing of data in traditional methods, and improve the accuracy of state judgment in complex environments.

[0118] The present application further proposes an expression of the environmental quality evaluation model, which is specifically:

[0119] ;

[0120] Wherein, EQI represents the environmental quality index, W j represents the dynamic weight coefficient, F j represents the normalized feature value of data in the joint feature matrix, j represents the type of environmental feature data, and n is the number of environmental feature data types;

[0121] The dynamic weight coefficient refers to a variable that dynamically adjusts the weight of each environmental parameter according to the statistical characteristics of historical data. Specifically, it can be realized by calculating the weight distribution ratio based on the variance of historical data. By giving higher weight to data with low variance, the model can give priority to environmental parameters with strong stability;

[0122] The normalized feature value refers to standardized data after eliminating dimensional differences. Specifically, it can be realized by using maximum and minimum normalization or Z-score standardization. By converting different types of parameters such as temperature, humidity and PM2.5 concentration to a unified numerical interval, the comparability of multi-source data in weighted calculation is ensured;

[0123] The environmental quality index refers to a quantitative indicator reflecting the comprehensive environmental state;

[0124] Specifically, the technical scheme constructs a linear combination model of dynamic weight and normalized features, dynamically calculates the weight coefficient of each parameter based on the historical data variance at each evaluation; for example, when the historical data variance of a certain type of environmental parameter is low, it indicates that the parameter has high stability, and the system automatically allocates a larger weight; when a certain type of parameter has abnormal fluctuations, the increase in variance leads to a decrease in weight, thereby suppressing the influence of abnormal data on the evaluation result.

[0125] At the same time, after all the environmental parameters are normalized, the unit of the temperature parameter is Celsius degree, and the unit of the PM 2.5 The micrograms per cubic meter of different dimensions of the parameters are eliminated, so that the humidity percentage data and the CO2 concentration data can be weighted in the same dimension;

[0126] By continuously updating the historical data window and recalculating the variance, the dynamic weight coefficient can be automatically adjusted with the change of season or pollution characteristics, for example, the weight proportion of the PM 2.5 Parameter is increased in the high-haze period in winter, and the weight of the temperature parameter is increased in the high-temperature period in summer.

[0127] Compared with the prior art, the traditional environmental quality evaluation model uses fixed weight coefficients, which cannot reflect the dynamic change characteristics of each environmental parameter under different regional and seasonal conditions, for example, the same weight is used when there is a significant difference in the influence of humidity parameters in coastal areas and humidity parameters in inland areas on air quality;

[0128] And the present application can automatically optimize the weight distribution according to the fluctuation characteristics of the historical data through the dynamic weight mechanism, for example, when the variance of the CO2 concentration data in a certain area continuously increases, the system automatically reduces its weight proportion to avoid interference with the evaluation result caused by abnormal fluctuations, which is more suitable for complex and variable actual environmental scenarios than the fixed weight model;

[0129] Through the above technical scheme, the present application realizes the dynamic optimization and configuration of the weight of the environmental quality evaluation model, and solves the adaptability defects of the fixed weight system in the time and space change scenarios.

[0130] Please refer to Figure 4 The present application further proposes a generation method of the dynamic weight coefficient W j , which specifically includes:

[0131] S5.1: Obtain historical environmental feature data;

[0132] S5.2: Establish a dynamic adjustment model according to the historical environmental feature data to generate a dynamic weight coefficient W j ;

[0133] The dynamic weight coefficient refers to a contribution degree parameter automatically adjusted according to the environmental parameter change characteristic, and can be specifically realized by adopting a historical data variance inverse proportion distribution mode, and the stability difference of different environmental parameters is reflected through variance calculation.

[0134] The dynamic adjustment model refers to a weight generation mechanism constructed based on historical data statistical characteristics, and can be specifically realized by adopting a variance reciprocal normalization method, and the weight is allocated through variance reciprocal proportion, so that the parameter with high stability obtains a higher weight.

[0135] Specifically, in the environmental quality evaluation process, historical environmental characteristic data is continuously collected and stored to form a data set reflecting the fluctuation characteristics of different environmental parameters. The dynamic adjustment model calculates the variance value of each type of historical data, and the variance value represents the fluctuation amplitude of the parameter in the historical period. The parameter with small variance indicates that its historical performance is stable, and a higher proportion is given in weight allocation; the parameter with large variance indicates that its volatility is strong, and the weight proportion is correspondingly reduced.

[0136] The weight coefficient is normalized to ensure that the sum is 1, and the finally generated dynamic weight coefficient is applied to the comprehensive evaluation model to realize real-time optimization and adjustment of the contribution degree of each environmental parameter.

[0137] Compared with the prior art, the traditional method uses fixed weight coefficients and cannot distinguish the stability difference of environmental parameters, resulting in that the evaluation result is greatly disturbed by the fluctuation parameters; the dynamic weight mechanism driven by variance in the present application can automatically identify and suppress the negative influence of high fluctuation parameters, and at the same time, enhance the decision weight of stable parameters, thereby improving the adaptability of the evaluation model to different environmental scenes.

[0138] The expression of the dynamic adjustment model is specifically:

[0139] ;

[0140] Wherein, σ J 2 Indicates the variance of the jth historical data, σ b 2 Indicates the variance of the jth historical data, b indicates the total number of data characteristics participating in fusion, and j is the index of the total number of data characteristics participating in fusion.

[0141] The dynamic adjustment model refers to a mathematical model for generating weight coefficients according to historical data statistical characteristics, and can be specifically realized by adopting a variance reciprocal proportion calculation mode. Variance is used to measure data volatility, and the data category with small variance obtains a higher proportion in weight allocation, thereby improving the contribution degree of stable data.

[0142] The historical data variance refers to various types of data volatility indexes calculated by historical environmental characteristic data, which can be calculated by using a time series analysis method, and is used to reflect the data reliability difference.

[0143] Specifically, the dynamic adjustment model automatically optimizes the weight distribution by analyzing the historical data variance. The data category with smaller variance indicates that it has lower historical volatility and higher stability, and thus is given a larger weight. The data category with larger variance has insufficient reliability, and the weight is correspondingly reduced. The model calculates the weight reference value of each category by the inverse proportion of the variance, and then generates a dynamic weight coefficient with a sum of 1 through normalization processing. For example, in the seasonal change scenario, the variance of temperature data may increase due to the increase of diurnal temperature difference, at which time the model automatically reduces the weight of the temperature parameter and increases the weight of the humidity parameter with smaller variance, so as to adapt to the change of environmental conditions.

[0144] Compared with the prior art, the traditional method uses fixed weight coefficients, which cannot dynamically adjust the parameter contribution according to the data volatility, resulting in deviation of the evaluation result when the data reliability changes. The present application introduces a dynamic adjustment mechanism based on variance, which can adaptively optimize the weight distribution and avoid evaluation errors of fixed weights in the scenario of sudden pollution events or regional differences.

[0145] Through the above technical solutions, the present application solves the problem that the fixed weight coefficient in the environmental quality evaluation model cannot be dynamically adjusted, and improves the adaptability of the evaluation result to different environmental conditions. Through the variance-driven weight distribution mechanism, the model can automatically suppress the interference of data with larger volatility and enhance the decision-making influence of stable parameters, so as to output more accurate environmental quality index in complex and variable scenarios.

[0146] Please refer to Figure 5 The present application further proposes an environmental monitoring system based on data fusion technology, which comprises:

[0147] A data acquisition unit 10 is used to synchronously acquire different types of environmental characteristic data at a fixed sampling period to form an original data set.

[0148] A data processing unit 20 is used to preprocess the original data set.

[0149] A data preliminary fusion unit 30 is used to fuse the same type of environmental characteristic data in the preprocessed original data set to generate a single-type state evaluation value.

[0150] A feature fusion unit 40 is used to generate a joint feature matrix according to the single-type state evaluation value.

[0151] A quality analysis unit 50 is used to establish an environmental quality evaluation model according to the joint feature matrix to generate an environmental quality index.

[0152] The monitoring unit 60 is used to monitor the environment according to the environmental quality index.

[0153] See also Figure 6 The present invention further proposes the data processing unit 20, which specifically includes:

[0154] The validity analysis module 21 is used to generate a data validity condition value based on the single environmental feature data in the original data set;

[0155] The elimination module 22 is used to eliminate the environmental characteristic data in the original data set according to the data validity condition value.

[0156] See also Figure 7 The present invention further proposes that the mass analysis unit 50 includes:

[0157] A historical data acquisition module 51 is used to acquire historical environmental feature data;

[0158] The result output module 52 is used to establish a dynamic adjustment model based on historical environmental characteristic data and generate a dynamic weight coefficient.

[0159] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An environmental monitoring method based on data fusion technology, characterized in that: The following steps are involved: Different types of environmental characteristic data are collected synchronously at a fixed sampling period to form an original data set; Preprocess the original data set; Perform data fusion on the same type of environmental feature data in the preprocessed original data set to generate a single type of state assessment value; Generate a joint feature matrix based on the single-type state evaluation value; Establish an environmental quality assessment model based on the joint characteristic matrix and generate an environmental quality index; Monitor the environment based on the environmental quality index.

2. The environmental monitoring method based on data fusion technology according to claim 1, characterized in that: The method of preprocessing the original data set specifically includes: Generate data validity condition values ​​based on the single environmental feature data in the original data set; Specifically, through the formula: ； Generate data valid condition value K; In the formula, x i It represents a single environmental feature data in the original data set, and μ represents all the environmental feature data x i The mean of data of the same type; According to the data validity condition value K, the environmental feature data in the original data set is eliminated.

3. The environmental monitoring method based on data fusion technology according to claim 1, characterized in that: The method for generating the single-type status evaluation value specifically includes: By formula: Generate a single type state evaluation value P k ; In the formula, P k-1 It represents the single type state evaluation value at time k-1, u k It represents the control input vector at time k, Z k Represents a single type of data vector at time k, K k represents the Kalman gain matrix at time k, A represents the state transfer matrix, B represents the control input matrix, and H represents the observation matrix.

4. The environmental monitoring method based on data fusion technology according to claim 1, characterized in that: The generation method of the joint feature matrix specifically includes: By formula: ； Generate joint feature matrix F; In the formula, T represents the ambient temperature state evaluation value in the single type state evaluation value, H represents the ambient humidity state evaluation value in the single type state evaluation value, IAQI (G CO2 ) represents the air quality index, G CO2 It represents the air quality status assessment value in the single type status assessment value, log (PM 2.5 +1) means PM 2.5 Logarithmic transformation of concentration, PM 2.5 Represents PM in a single type of state assessment value 2.5 State assessment value, M n It represents the remaining single-type status evaluation values ​​in the single-type status evaluation value.

5. The environmental monitoring method based on data fusion technology according to claim 1, characterized in that: The expression of the environmental quality assessment model is specifically: ； Among them, EQI stands for Environmental Quality Index, W j It represents the dynamic weight coefficient, F j It represents the normalized eigenvalue of the data in the joint feature matrix, j represents the type of environmental feature data, and n is the number of environmental feature data types.

6. The environmental monitoring method based on data fusion technology according to claim 5, characterized in that: The dynamic weight coefficient W j The generation methods include: Obtain historical environmental characteristics data; Establish a dynamic adjustment model based on historical environmental characteristic data and generate a dynamic weight coefficient W j .

7. The environmental monitoring method based on data fusion technology according to claim 6, characterized in that: The expression of the dynamic adjustment model is specifically: ； Among them, σ J 2 It represents the variance of the j-th historical data, σ b 2 It represents the variance of the j-th historical data, b represents the total number of data features involved in the fusion, and j is the index of the total number of data features involved in the fusion.

8. An environmental monitoring system based on data fusion technology, characterized in that: The system is used to execute the environmental monitoring method based on data fusion technology as described in any one of claims 1 to 7, specifically comprising: A data acquisition unit is used to synchronously collect different types of environmental characteristic data at a fixed sampling period to form an original data set; A data processing unit, used for preprocessing the original data set; The data preliminary fusion unit is used to fuse the same type of environmental feature data in the preprocessed original data set to generate a single type state assessment value; A feature fusion unit is used to generate a joint feature matrix based on single-type state evaluation values; A quality analysis unit is used to establish an environmental quality assessment model based on the joint feature matrix and generate an environmental quality index; The monitoring unit is used to monitor the environment according to the environmental quality index.

9. The environmental monitoring system based on data fusion technology according to claim 8, characterized in that: The data processing unit specifically includes: The validity analysis module is used to generate data validity condition values ​​based on the individual environmental feature data in the original data set; The elimination module is used to eliminate the environmental feature data in the original data set according to the data validity condition value.

10. The environmental monitoring system based on data fusion technology according to claim 8, characterized in that: The mass analysis unit comprises: A historical data acquisition module is used to obtain historical environmental feature data; The result output module is used to establish a dynamic adjustment model based on historical environmental characteristic data and generate dynamic weight coefficients.

Images