An intelligent supervision system and method for environmental monitoring quality control

Abstract

The present invention discloses an intelligent supervision system and method for environmental monitoring quality control, which specifically relates to the technical field of environmental monitoring. After real-time collection of environmental monitoring data by multiple types of sensors and formatting, the system analyzes the fluctuation of each category of data and identifies the hidden dangers of abnormal fluctuations. If hidden dangers are found, drift verification is performed on the corresponding sensor, noise and drift characteristics are extracted, and data noise index and cumulative drift index are generated. Based on these indexes, fuzzy logic is used to divide the sensors into three categories of severe, moderate and low-level drift. Correction analysis is performed according to the division results, and local or full-category calibration operations are selected to optimize the quality of monitoring data. By introducing an intelligent evaluation mechanism, the present invention can monitor and analyze the drift and data noise level of each sensor in real time, dynamically determine which sensors need to be calibrated, and select local or full-category calibration according to actual conditions to ensure the long-term accuracy and reliability of environmental monitoring data.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and more specifically, to an intelligent supervision system and method for environmental monitoring quality control. Background Art

[0002] With the improvement of environmental protection requirements and the development of intelligent technology, environmental monitoring systems have been widely used in urban management, ecological protection, industrial parks and other fields. These systems rely on a variety of sensors to collect real-time data on environmental parameters such as air quality, water quality, noise, temperature and humidity. However, sensors are prone to drift and increased data noise during long-term use, resulting in a decrease in data accuracy, which in turn affects the effectiveness of environmental monitoring.

[0003] Most of the current environmental monitoring systems use regular calibration or manual calibration. Although this method can solve the drift problem to a certain extent, it has disadvantages such as being time-consuming and wasting resources. Especially when faced with a large number of sensors, it cannot flexibly respond to abnormal drift of individual sensors. In addition, when the environment changes drastically, the existing system cannot respond to the drift problem of the sensor in a timely manner, which may cause distortion of key data. Therefore, the existing technology lacks a calibration mechanism that can be intelligently and dynamically adjusted according to the degree of sensor drift in order to ensure the accuracy of environmental monitoring data. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] An intelligent supervision method for environmental monitoring quality control comprises the following steps:

[0006] Collect environmental monitoring data in real time through multiple types of sensors, format the collected data to ensure data availability, and then analyze the fluctuations of each category of data to identify whether there is a hidden danger of abnormal fluctuations in the target category data;

[0007] If the target category data has the hidden danger of abnormal fluctuation, the sensor corresponding to the target category data is drift verified to obtain the drift verification data set and perform feature extraction. Then, drift analysis is performed based on the feature extraction results to generate the data noise index and cumulative drift index respectively.

[0008] Based on the data noise index and cumulative drift index, fuzzy logic is used to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low-level drift.

[0009] Obtain all data categories with hidden dangers of abnormal fluctuations, obtain the corresponding sensor classification results, perform correction analysis, and select local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition.

[0010] In a preferred embodiment, the analysis steps of the fluctuation situation are:

[0011] Arrange the sensor data of each category in chronological order to form a time series of sensor data of this category, and then obtain the corresponding data of two adjacent time points in the time series and mark them as , , t represents the sequence number in the time series, and then the following calculation is performed:

[0012] ; Indicates the time point corresponding to sequence number t, It represents the value of the data change amplitude, and then uses the Z-score anomaly detection method to measure the degree of deviation of the current data relative to the historical data, and obtains the Z-score value at time point t, and finally obtains a detection sequence containing all Z-score values.

[0013] In a preferred embodiment, identifying whether the target category data has the hidden danger of abnormal fluctuations refers to:

[0014] Each Z-score value in the detection sequence corresponding to the target category data is compared with the preset anomaly threshold, and the sequence number corresponding to each Z-score value greater than the preset anomaly threshold is recorded and summarized as an anomaly point marking set. When the number of data in the anomaly point marking set is non-zero, there is a risk of abnormal fluctuation in the target category data.

[0015] In a preferred embodiment, drift verification refers to:

[0016] Select a calibrated standard sensor of the same type as the target category data as the reference for drift verification. The two sensors perform data collection operations and meet the following conditions:

[0017] Place the target category sensor and the standard sensor in the same environment to ensure that both collect the same type of environmental data at the same time;

[0018] The acquisition time of the two is synchronized to ensure that the two sets of data have the same time stamp.

[0019] In a preferred embodiment, feature extraction refers to extracting noise feature information and drift feature information from the drift verification data set respectively.

[0020] In a preferred embodiment, the logic for obtaining the data noise index is:

[0021] Obtain the environmental data collected in real time by the sensor corresponding to the target category data from the noise feature information, as well as the timestamp of each collection;

[0022] The first-order difference formula is used to calculate the data change amplitude of adjacent time points to form the overall noise characteristics. The formula is: ; is the total amount of environmental data collected in real time, is the overall noise value, Indicates the timestamp corresponding to the data sequence number i, Indicates the environmental data corresponding to the data sequence number i, is the instantaneous variation coefficient, ;

[0023] Use the second-order difference formula to calculate local data fluctuations and capture the fluctuation characteristics in a short period of time. The formula is: ; is the total amount of environmental data collected in real time, is the local noise value;

[0024] The first-order difference and second-order difference results are comprehensively considered using a linear regression model to generate a data noise index, the formula is: ; , , , are the preset regression coefficients. is the data noise index.

[0025] In a preferred embodiment, the logic for obtaining the cumulative drift index is:

[0026] Obtaining the baseline data and monitoring data respectively collected by the standard sensor and the target sensor from the drift characteristic information;

[0027] Time alignment of baseline data and monitoring data to ensure data from the same time points can be compared;

[0028] Calculate the difference between the baseline data and the monitoring data at each time point to obtain a difference sequence;

[0029] Based on the difference sequence, the time decay factor is introduced to calculate the cumulative drift index. The formula is:

[0030] ; , are the start and end time of the collection, is the time decay factor, represents the baseline data at time point t, represents the monitoring data at time point t, Represents the cumulative drift index.

[0031] In a preferred embodiment, the fuzzy logic is used as follows:

[0032] The data noise index and cumulative drift index of the sensor corresponding to the target category data are taken as input variables, the detection drift degree of the sensor corresponding to the target category data is taken as the output variable, the input variables are fuzzified, and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified, and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the detection drift degree under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the detection drift degree of the sensor corresponding to the target category data.

[0033] In a preferred embodiment, obtaining all data categories with hidden dangers of abnormal fluctuations, and correspondingly obtaining sensor classification results, performing correction analysis, and selecting local calibration or full category calibration operation based on the correction analysis results to optimize the monitoring data acquisition quality refers to:

[0034] Get the sensor classification results corresponding to the target category data. Each type of classification result corresponds to a constant value. Mark the constant value corresponding to the classification result of the target sensor numbered i as Ci, and then calculate the optimization value: ; The preset self-influence coefficient of the target sensor number i, is the constant value corresponding to the division result of the target sensor numbered j, N represents the total number of all sensors, The preset inter-sensor influence coefficient for the target sensor number i, is the interference coefficient between the target sensor numbered i and the sensor numbered j, which is obtained by calculating the correlation of the detection data streams of the target sensor numbered i and the sensor numbered j, namely the Pearson correlation coefficient, and then performing absolute value processing. For the optimization value, if there is a sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the full category calibration operation is selected; if there is no sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the local calibration operation is selected.

[0035] In a preferred embodiment, an environmental monitoring quality control intelligent supervision system includes:

[0036] Environmental data collection module, which collects environmental monitoring data in real time through multiple types of sensors and formats the collected data to ensure data availability;

[0037] The fluctuation analysis module analyzes the fluctuation of each category of data and identifies whether the target category data has the hidden danger of abnormal fluctuation;

[0038] Drift verification analysis module: if the target category data has the risk of abnormal fluctuation, the sensor corresponding to the target category data is drift verified to obtain the drift verification data set and perform feature extraction. Then, drift analysis is performed based on the feature extraction results to generate data noise index and cumulative drift index respectively.

[0039] The type classification module uses fuzzy logic to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low drift based on the data noise index and cumulative drift index;

[0040] The correction analysis module obtains all data categories with hidden dangers of abnormal fluctuations, obtains the corresponding sensor classification results, performs correction analysis, and selects local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition.

[0041] Technical effects and advantages of the present invention:

[0042] The present invention can monitor and analyze the drift and data noise level of each sensor in real time by introducing an intelligent evaluation mechanism. By accurately quantifying the drift degree and noise characteristics, the system can dynamically determine which sensors need to be calibrated and select partial or full category calibration according to actual conditions. This dynamic calibration method based on actual data ensures the long-term accuracy and reliability of environmental monitoring data and greatly reduces the data distortion problem caused by long-term drift of sensors.

[0043] Traditional environmental monitoring systems usually rely on regular manual calibration or unified calibration of all sensors, which has the disadvantages of long time, high cost, and waste of resources. In contrast, the present invention can make calibration decisions based on the real-time status of the sensor through intelligent algorithms and select the most suitable calibration operation. This not only reduces unnecessary full system calibration operations, but also greatly saves the time and labor costs required for calibration, significantly improving the overall operating efficiency of the system.

[0044] In actual application scenarios, environmental conditions change in a complex and unpredictable manner, and sensors may drift rapidly in certain extreme cases. The present invention can quickly detect abnormal fluctuations in sensor data and respond in a timely manner. When the data quality of one or more types of sensors is affected, the system can quickly select appropriate calibration operations to avoid distortion and lag of key monitoring data, thereby improving the response speed of the entire monitoring system.

[0045] The present invention significantly reduces the reliance on manual calibration through an intelligent calibration mechanism. Sensor drift detection, data analysis, and calibration decisions are all automatically completed by the system, reducing the frequency and complexity of human intervention. This increase in automation not only reduces the risk of errors caused by manual operation, but also significantly reduces maintenance costs, allowing the system to operate autonomously for a long time and ensure the continuity and accuracy of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0047] Figure 1 The schematic diagram is a schematic diagram of an intelligent supervision method for environmental monitoring quality control in the present invention.

[0048] Figure 2 The schematic diagram of an intelligent supervision system for environmental monitoring quality control in the present invention is shown in FIG. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Reference Figure 1 - Figure 2 The following embodiments are obtained:

[0051] Embodiment 1: A method for intelligent supervision of environmental monitoring quality control, comprising the following steps:

[0052] Environmental monitoring data is collected in real time through multiple types of sensors, and the collected data is formatted to ensure the availability of the data. Then, the fluctuation of each category of data is analyzed to identify whether there is a hidden danger of abnormal fluctuation in the target category data; environmental data is collected in real time through multiple types of sensors (such as air quality sensors, temperature and humidity sensors, noise sensors, etc.). The collected data comes from different sensors, and the format and unit may be different. To ensure the effectiveness of subsequent data processing, the data needs to be formatted, such as unifying units and removing null values. Ensuring that the system can obtain continuous and accurate data input is the basis of intelligent supervision. Data formatting is the premise to ensure that all sensor data can be analyzed uniformly, and to ensure that data from different sources can be integrated and used for subsequent analysis. Through this step, the system can effectively handle different types of data sources to ensure the availability and consistency of monitoring data. By analyzing data fluctuations, the system can detect sudden abnormalities or instability of data, which helps to discover potential problems in advance. If the data fluctuation is too large or abnormal, it may mean that the environmental conditions have changed dramatically, the sensor performance has declined, or external interference has occurred. Identifying hidden dangers of abnormal fluctuations can help the system to carry out risk warnings and troubleshoot problems in advance.

[0053] If the target category data has the hidden danger of abnormal fluctuation, the sensor corresponding to the target category data is drift verified, the drift verification data set is obtained and feature extraction is performed, and then drift analysis is performed based on the feature extraction results to generate data noise index and cumulative drift index respectively; once a data category is identified as having the hidden danger of abnormal fluctuation, the system performs drift verification on the sensor corresponding to the category and collects the drift verification data set. Then, the noise feature information and drift feature information are extracted from the drift verification data set, the state of the sensor is analyzed, and it is determined whether there is data drift. During the long-term use of the sensor, drift may occur due to aging, that is, the measured value gradually deviates from the true value. Drift verification is to detect whether the performance of the sensor is degraded. Through feature extraction, the system can effectively distinguish between noise and long-term drift in the data to ensure the stability and reliability of data quality.

[0054] Based on the data noise index and cumulative drift index, fuzzy logic is used to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low-level drift; after drift verification, the data noise index and cumulative drift index are calculated respectively: the noise index reflects the random fluctuations in the sensor data and describes the stability of the data. The cumulative drift index reflects the long-term drift trend of the sensor data and describes the degree to which the sensor data deviates from the true value over time. Through the noise index and cumulative drift index, the system can quantify two types of problems in sensor data, short-term noise and long-term drift, which provides an accurate basis for subsequent sensor classification and correction analysis, ensuring that the handling of problem sensors is more reasonable and targeted. Based on the noise index and cumulative drift index, fuzzy logic is used to classify the degree of sensor drift and divide it into three types: severe drift, moderate drift, and low-level drift. Fuzzy logic can perform flexible classification according to different degrees of drift through rule set processing.

[0055] Obtain all data categories with hidden dangers of abnormal fluctuations, and obtain the corresponding sensor classification results, perform correction analysis, and select local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition. Obtain all sensor categories with hidden dangers of abnormal fluctuations, and determine the current status of each category of sensors in combination with the drift analysis results (such as severe drift, moderate drift, and low-level drift). The system determines the correction strategy to be taken based on the sensor classification results. Through this step, the system can fully understand which sensors have problems and how serious the problems are, thereby providing complete input information for correction analysis. This helps the system make reasonable correction decisions, avoid blind or excessive corrections, and improve the efficiency and accuracy of the system. Based on the drift classification results, the system performs correction analysis to determine whether to select local calibration (for a few sensors with severe drift) or full category calibration (overall calibration for multiple problem sensors). The correction analysis calculates the optimization value through the optimization formula, and selects the appropriate calibration strategy based on the drift degree and mutual influence between sensors. Correction analysis is an important part of the operation of the entire system, ensuring that the system can self-regulate and maintain the accuracy of monitoring data. By optimizing the strategy and selecting a reasonable calibration method, the system can effectively reduce the problem of inaccurate data caused by sensor drift. Partial calibration avoids unnecessary full calibration and saves resources, while full-category calibration can promptly respond to the performance degradation of the overall sensor. According to the correction analysis results, the system optimizes the monitoring data acquisition process. For sensors that need to be calibrated, the system automatically performs calibration operations, adjusts the sensor's baseline or reinitializes it. Further monitoring of the optimized data ensures the accuracy and stability of the environmental monitoring data. This final step ensures that the system can automatically adjust the performance of the sensor during long-term use to avoid misjudgments or errors caused by data drift or noise. Through calibration operations, the system can maintain continuous high-precision data monitoring, thereby improving the reliability and data quality of the monitoring system.

[0056] The analysis steps of fluctuation are as follows: arrange the sensor data of each category in chronological order to form a time series of sensor data of this category, and then obtain the corresponding data of two adjacent time points in the time series and mark them as , , t represents the sequence number in the time series, and then the following calculation is performed: ; Indicates the time point corresponding to sequence number t, Indicates the value of the data change amplitude, and then uses the Z-score anomaly detection method to measure the deviation of the current data from the historical data, obtains the Z-score value at time point t, and finally obtains a detection sequence containing all Z-score values. The Z-score value calculation formula is: ; Z(t) is the Z-score value, which represents the difference between the data change amplitude Δd(t) at time point t and the historical average change amplitude The Z-score is used to measure the degree to which the current data deviates from the average value. The larger the Z value, the more the current data deviates from the average value, and the greater the possibility of anomalies. Δd(t) represents the amplitude of the data change at time point t, which is usually calculated from the measurement value of the sensor. It reflects the change of the sensor measurement data between time point t and the previous time point. This value is used to evaluate the instantaneous fluctuation of sensor data. It is the average change of historical data, which indicates the average value of data changes in the past period of time. This average value is used as a benchmark to determine whether the current data change range is within the normal range. σ represents the standard deviation of the change range of historical data, which reflects the volatility and dispersion of historical data. The larger the standard deviation, the greater the volatility of historical data; the smaller the standard deviation, the smaller the volatility of historical data. The standard deviation is used to normalize the difference between the current data and the average value, so as to more clearly identify abnormal situations.

[0057] The hidden dangers of identifying whether the target category data has abnormal fluctuations refer to:

[0058] Each Z-score value in the detection sequence corresponding to the target category data is compared with the preset abnormal threshold, and the sequence number corresponding to each Z-score value greater than the preset abnormal threshold is recorded and summarized as an abnormal point mark set. When the number of data in the abnormal point mark set is non-zero, the target category data has the hidden danger of abnormal fluctuation. Compare each calculated Z-score value with the preset abnormal threshold. If the Z-score value of a data point exceeds the threshold, it is considered that the data point may be an abnormal point. Record the data point numbers corresponding to all Z-score values ​​greater than the preset threshold to form an abnormal point mark set. This set contains all data points marked as potentially abnormal. If there is data in the abnormal point mark set (that is, the set is not empty), the system determines that the target category data has the hidden danger of abnormal fluctuation. If the set is empty, it means that no abnormal fluctuation is detected. The Z-score method is a very common preliminary anomaly detection method. Through Z-score comparison, the system can quickly and intuitively screen out points that may have abnormal fluctuations and discover the hidden dangers of abnormal fluctuations in the data in advance. The abnormal point mark set is a standard for preliminarily judging whether the data has potential fluctuation risks. If the set is not empty, it means that there may be more abnormal fluctuations in the data, and the system will perform further processing, such as drift verification, data noise index calculation, etc. The significance of this step is to provide a basis for subsequent in-depth analysis. This set identifies potential anomalies in the data sequence and is an important reference for the system to make preliminary warnings. If the amount of data in the anomaly point set is large, and if there is a priority for analysis in actual use, then computing resources can be concentrated for subsequent processing operations. In actual use, if the amount of data in the anomaly point set is large, for example, more than 5% of the total data volume, a warning signal is generated to remind the user that this sensor has serious hidden dangers.

[0059] Drift verification means: select a calibrated standard sensor of the same type as the sensor corresponding to the target category data as the benchmark for drift verification, and the two sensors perform collection operations to meet the following conditions: place the target category sensor and the standard sensor in the same environment to ensure that both collect the same type of environmental data at the same time; synchronize the collection time of the two to ensure that the two sets of data have the same timestamp. Select a standard sensor of the same type as the sensor corresponding to the target category data, which has been accurately calibrated and can provide accurate benchmark data. The target sensor will compare data with the standard sensor. Using a standard sensor as a benchmark is the basis for ensuring the accuracy and credibility of drift verification. The calibrated standard sensor has a known measurement accuracy. It provides a reference value in the real environment. By comparing it with the target sensor data, it can be detected whether the target sensor has drift. This can effectively distinguish the difference between sensor errors and changes in environmental data.

[0060] Ensure that the target sensor and the standard sensor operate in the same physical environment (same location, same conditions) to ensure that they collect the same environmental data. When the sensor collects data, the conditions of the surrounding environment (such as temperature, humidity, air quality, etc.) will directly affect the measurement results. Therefore, the measurement data of the standard sensor and the target sensor can only be effectively compared in the same environment. If the two are not in the same environment, the data difference may come from environmental changes rather than sensor drift, which will lead to wrong conclusions.

[0061] The acquisition time of the target sensor and the standard sensor is synchronized to ensure that the two sets of data have the same timestamp at the same time point. The synchronization of data timestamps is crucial to ensure that each data point is from the same moment. Even in the same environment, environmental data at different time points may be different. If the timestamps are not synchronized, the comparison will no longer be the measurement results at the same moment, and the drift problem of the sensor cannot be effectively detected. Time synchronization can eliminate the impact of time differences and ensure the accuracy of data comparison.

[0062] By comparing the standard sensor with the target sensor, drift verification can accurately detect whether the target sensor has deviations and ensure the accuracy of environmental monitoring data. Sensors may gradually deviate from the true value over time due to aging or other factors. Drift verification can help the system detect this performance degradation early and avoid erroneous data affecting the monitoring results. By performing drift verification regularly, the system can actively calibrate or replace sensors to maintain long-term stable monitoring performance.

[0063] Feature extraction refers to extracting noise feature information and drift feature information from the drift verification data set. Extracting noise feature information of sensor data from drift verification data is mainly done by detecting short-term fluctuations and high-frequency fluctuations in sensor output data. Noise is usually short-term, irregular, random changes that may be caused by internal or external interference in the sensor. Extracting drift feature information of the sensor from drift verification data determines whether the sensor has drifted by analyzing the long-term trend of the data. Drift features are usually reflected as a gradual deviation of data over time, usually due to sensor aging or environmental changes.

[0064] The logic of obtaining the data noise index is as follows: obtain the environmental data collected in real time by the sensor corresponding to the target category data from the noise feature information, as well as the timestamp of each collection;

[0065] The first-order difference formula is used to calculate the data change amplitude of adjacent time points to form the overall noise characteristics. The formula is: ; is the total amount of environmental data collected in real time, is the overall noise value, Indicates the timestamp corresponding to the data sequence number i, Indicates the environmental data corresponding to the data sequence number i, is the instantaneous variation coefficient, , emphasizing the instantaneous rate of change, a shorter time difference means faster fluctuations, and the instantaneous change coefficient wi is higher. The first-order difference is used to capture the overall change trend and noise fluctuations of sensor data, reflecting the general fluctuations of sensor data.

[0066] Use the second-order difference formula to calculate local data fluctuations and capture the fluctuation characteristics in a short period of time. The formula is: ; is the total amount of environmental data collected in real time, is the local noise value; through the second-order difference operation, the high-frequency noise changes in the sensor data can be better captured. This processing can ignore the linear trend of the data, making the noise part more prominent, which is suitable for detecting short-term data instability. The second-order difference can not only amplify the noise, but also reduce the impact of the long-term trend in the sensor data on the noise calculation. Compared with the previous formula, the new deformation formula pays more attention to data fluctuations in a short period of time rather than long-term changes. By using the square of the time interval in the denominator, the rapidity of noise changes is emphasized. At a shorter time interval, the impact of noise is further amplified, and the random fluctuations of sensor data can be captured more accurately. This formula is suitable for capturing high-frequency noise phenomena in sensors caused by environmental fluctuations or interference. It can be used to effectively identify the noise frequently generated during sensor operation. The second-order difference focuses on capturing high-frequency changes in a short period of time, which is used to reflect local fluctuations in sensor data, especially sharp changes (noise) in detection data.

[0067] The first-order difference and second-order difference results are comprehensively considered using a linear regression model to generate a data noise index, the formula is: ; , , , are the preset regression coefficients. is the data noise index. By calculating the first-order and second-order differences simultaneously, the formula can capture different levels of noise in the sensor data. The first-order difference reflects the global noise, while the second-order difference can more sensitively detect sharp fluctuations in the short term. The linear regression formula combines time t, the overall noise feature F1, and the local noise feature F2 to generate the final noise index ZS. The regression coefficient is determined by fitting the historical data with the purpose of minimizing the fitting error, and is used to adjust the contribution of each feature in the noise index. This regression formula can comprehensively consider the influence of time factors, overall fluctuations, and local fluctuations to generate a more comprehensive noise feature index. When the ZS value is larger, it means that the data fluctuations of the sensor are more drastic, the noise level is higher, and it may be greatly affected by external interference or random fluctuations of the sensor itself. If both F1 (overall noise) and F2 (local noise) are large, it means that the data not only has an overall fluctuation trend, but also has sharp fluctuations in the short term, which may be a signal of drastic environmental changes or sensor abnormalities. ZS is calculated by linear regression and combines different features, so it is a comprehensive measure of the data noise level and reflects the current state of data quality.

[0068] The logic of obtaining the cumulative drift index is as follows: obtain the baseline data and monitoring data collected by the standard sensor and the target sensor respectively from the drift feature information; perform time alignment on the baseline data and the monitoring data to ensure that the data at the same time point are compared;

[0069] Calculate the difference between the baseline data and the monitoring data at each time point to obtain a difference sequence;

[0070] Based on the difference sequence, the time decay factor is introduced to calculate the cumulative drift index. The formula is:

[0071] ; , are the start and end time of the collection, is the time decay factor, which is used to reduce the impact of long-term data on the drift index. represents the baseline data at time point t, represents the monitoring data at time point t, Represents the cumulative drift index.

[0072] The standard sensor data serves as the baseline value of the real environment and is used to evaluate whether the data of the target sensor deviates from the real value. The data of the target sensor may drift over time, so it is necessary to compare it with the baseline data of the standard sensor to determine the degree of drift. And only by comparing the data of the standard sensor and the target sensor at the same time point can the drift of the target sensor be accurately detected. If there is no time synchronization, the data may be affected by the time difference, resulting in inaccurate drift assessment results. At each time point t, the difference between the baseline data and the target sensor monitoring data is calculated to obtain a difference sequence: this difference reflects the deviation between the target sensor and the standard sensor data at each moment. The difference sequence can help quantify the degree of deviation of sensor data over time and is a key data source for evaluating drift. Time decay factor term The introduction of ensures that the system pays more attention to the recent drift, while the drift effect at a more distant time point will gradually decrease. This helps to timely detect the current drift state of the sensor and reduce the impact of past data on the current drift assessment.

[0073] The cumulative drift index is a cumulative assessment of the sensor's deviation from the baseline data over the entire time series. Through this index, the system can quantitatively analyze whether the sensor is gradually losing accuracy and evaluate its overall drift. Over time, the influence of old data on the current drift assessment decreases, allowing the system to focus more on recent drift performance. This approach ensures that the drift assessment results are more accurate and timely, and helps to deal with the performance degradation of sensors in long-term operation. The larger the cumulative drift index, the greater the deviation of the sensor data from the standard baseline, that is, the more serious the drift of the sensor. If the cumulative drift index is large, it means that the sensor's measurement value has deviated greatly from the baseline value for a long time, and there may be serious calibration problems or sensor performance degradation. The larger the cumulative drift index, the sooner the sensor needs to be calibrated or replaced, otherwise it may cause distortion of environmental monitoring data.

[0074] The logic of using fuzzy logic is: take the data noise index and cumulative drift index of the sensor corresponding to the target category data as input variables, take the detection drift degree of the sensor corresponding to the target category data as the output variable, fuzzify the input variables, convert the value of the input variables into a fuzzy set, fuzzify the output variables, convert the output variables into a fuzzy set, formulate fuzzy rules, describe the detection drift degree under different data type combinations, and reason the fuzzified input variables through fuzzy rules to obtain the detection drift degree of the sensor corresponding to the target category data. First, take the data noise index and cumulative drift index of the sensor corresponding to the target category data as input variables, and the detection drift degree of the sensor as the output variable. Fuzzify the input variables, which means converting the specific values ​​of the input variables into fuzzy sets. The fuzzified values ​​may represent different degrees such as low, medium, and high. Then, fuzzify the output variables as well, which means converting the detection drift degree into a fuzzy set, such as low-level drift, moderate drift, or severe drift. On this basis, formulate fuzzy rules. Fuzzy rules describe the detection drift degree under different input data combinations. For example, when the noise index is high and the cumulative drift index is medium, the detected drift may be moderate drift. When both the noise index and the cumulative drift index are high, it may indicate severe drift. Finally, the fuzzified input variables are inferred through the formulated fuzzy rules. The inference process combines the relationship between different fuzzy sets to calculate the final result of the detection drift degree. The result reflects the drift degree of the sensor corresponding to the target category data and helps determine the calibration requirements of the sensor.

[0075] Obtain all data categories with hidden dangers of abnormal fluctuations, obtain sensor classification results accordingly, perform correction analysis, and select local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition. This means:

[0076] Get the sensor classification results corresponding to the target category data. Each type of classification result corresponds to a constant value. Mark the constant value corresponding to the classification result of the target sensor numbered i as Ci, and then calculate the optimization value: ; The preset self-influence coefficient of the target sensor number i, is the constant value corresponding to the division result of the target sensor numbered j, N represents the total number of all sensors, The preset inter-sensor influence coefficient for the target sensor number i, is the interference coefficient between the target sensor numbered i and the sensor numbered j, which is obtained by calculating the correlation of the detection data streams of the target sensor numbered i and the sensor numbered j, namely the Pearson correlation coefficient, and then performing absolute value processing. For the optimization value, if there is a sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the full category calibration operation is selected; if there is no sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the local calibration operation is selected.

[0077] The constant value represents the drift degree or calibration adjustment required for this type of sensor. The following is an example: Assuming the classification result is severe drift, the constant value can be set to a larger value, such as 10, indicating that the sensor requires a greater degree of adjustment. For sensors with moderate drift, the constant value can be set to a medium value, such as 5, indicating that a moderate degree of adjustment is required. For low-level drift types, the constant value may be set to a smaller value, such as 2, indicating that only a small calibration adjustment is required. These constant values ​​are set in advance by the system to evaluate the correction needs of sensors with different drift degrees. The sensor's own classification constant value represents the assessment of the calibration needs of the current sensor's detection results. The self-influence coefficient preset for the target sensor numbered i is used to reflect the degree of influence of the sensor on the overall performance of the system. The larger the value, the greater the influence. The inter-sensor influence coefficient preset for the target sensor numbered i indicates the degree of influence between the sensor and other sensors in the system. is the interference coefficient between the target sensor numbered i and the sensor numbered j, which is obtained by calculating the correlation of the detection data streams of the target sensor numbered i and the sensor numbered j, namely the Pearson correlation coefficient, and then performing absolute value processing, which represents their influence on each other.

[0078] If the optimization value is less than or close to the preset calibration judgment threshold, it means that the drift of the sensor is relatively mild and only affects a part of the system. The system will select a local calibration operation, that is, only adjust the part of the sensor with severe drift. If the optimization value is greater than the preset calibration judgment threshold, it means that the drift of the sensor is not only large, but may also affect other sensors. The system will select a full-category calibration operation to fully calibrate all related sensors. The optimization value not only takes into account the drift of a single sensor, but also the potential impact of the sensor on other sensors in the system. This calculation method ensures the comprehensiveness and rationality of the calibration decision. By introducing correlation parameters and self-influence coefficients between sensors, the system can evaluate whether the drift of a sensor will affect other sensors. If the drift of the sensor affects not only itself but also other sensors, the optimization value will be higher and the system needs to perform a full-category calibration.

[0079] Example of full-category calibration: Assume that there are five sensors in the system, and the optimization value of each sensor is calculated and compared with the system's preset threshold: the optimization value of sensor 1 is 8, the optimization value of sensor 2 is 7, the optimization value of sensor 3 is 10, the optimization value of sensor 4 is 5, and the optimization value of sensor 5 is 6; assume that the preset optimization judgment threshold is 9. Since the optimization value of sensor 3 is 10 greater than or equal to this threshold, the system believes that at least one sensor's optimization value exceeds the threshold. So the system will choose the full-category calibration operation. The system will calibrate all sensors, whether they are sensors with optimization values ​​exceeding the threshold or those that do not exceed the threshold. This is to ensure the overall data consistency and accuracy of the system. The logic of full-category calibration is that as long as there is a sensor with an optimization value greater than or equal to the threshold, the system believes that the drift problem is serious enough and may affect the accuracy of the entire monitoring system, so all sensors are uniformly calibrated.

[0080] Example of partial calibration: Assume that there are five sensors with the following optimized values: Sensor 1 has an optimized value of 7, Sensor 2 has an optimized value of 6, Sensor 3 has an optimized value of 8, Sensor 4 has an optimized value of 5, and Sensor 5 has an optimized value of 7; Assume that the preset optimization judgment threshold is still 9. In this case, no sensor has an optimized value greater than or equal to 9, that is, the optimized values ​​of all sensors are below the threshold of the system. According to the rule, the system will select a partial calibration operation, which means that only those sensors with higher optimized values ​​but still below the threshold will be calibrated. The system believes that the drift of these sensors has not yet reached a level that seriously affects the overall monitoring performance, so only these sensors can be calibrated locally. The logic of local calibration is that if the optimized values ​​of all sensors do not exceed the preset threshold, the system considers that the overall drift problem is relatively small, and only performs local corrections on sensors with obvious drift but not exceeding the threshold to maintain the operating efficiency of the system. Local corrections can be made to the top m sensors ranked by optimization values ​​or p percent of sensors. For example, the value of p can be set to 60. The advantage of local calibration is that only a group of sensors with more prominent problems are corrected, which saves system resources and time, does not interrupt the normal operation of other sensors, and avoids affecting the overall monitoring capability of the system.

[0081] Embodiment 2: An intelligent supervision system for environmental monitoring quality control, comprising:

[0082] Environmental data collection module, which collects environmental monitoring data in real time through multiple types of sensors and formats the collected data to ensure data availability;

[0083] The fluctuation analysis module analyzes the fluctuation of each category of data and identifies whether the target category data has the hidden danger of abnormal fluctuation;

[0084] Drift verification analysis module: if the target category data has the risk of abnormal fluctuation, the sensor corresponding to the target category data is drift verified to obtain the drift verification data set and perform feature extraction. Then, drift analysis is performed based on the feature extraction results to generate data noise index and cumulative drift index respectively.

[0085] The type classification module uses fuzzy logic to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low drift based on the data noise index and cumulative drift index;

[0086] The correction analysis module obtains all data categories with hidden dangers of abnormal fluctuations, obtains the corresponding sensor classification results, performs correction analysis, and selects local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition.

[0087] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0088] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0089] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An intelligent supervision method for environmental monitoring quality control, characterized in that: The following steps are involved: Collect environmental monitoring data in real time through multiple types of sensors, format the collected data to ensure data availability, and then analyze the fluctuations of each category of data to identify whether there is a hidden danger of abnormal fluctuations in the target category data; If the target category data has the hidden danger of abnormal fluctuation, the sensor corresponding to the target category data is drift verified to obtain the drift verification data set and perform feature extraction. Then, drift analysis is performed based on the feature extraction results to generate the data noise index and cumulative drift index respectively. Based on the data noise index and cumulative drift index, fuzzy logic is used to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low-level drift. Obtain all data categories with hidden dangers of abnormal fluctuations, obtain sensor classification results accordingly, perform correction analysis, and select local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition; The logic for obtaining the data noise index is: Obtain the environmental data collected in real time by the sensor corresponding to the target category data from the noise feature information, as well as the timestamp of each collection; The first-order difference formula is used to calculate the data change amplitude of adjacent time points to form the overall noise characteristics. The formula is: ; is the total amount of environmental data collected in real time, is the overall noise value, Indicates the timestamp corresponding to the data sequence number i, Indicates the environmental data corresponding to the data sequence number i, is the instantaneous variation coefficient, ; Use the second-order difference formula to calculate local data fluctuations and capture the fluctuation characteristics in a short period of time. The formula is: ; is the total amount of environmental data collected in real time, is the local noise value; The first-order difference and second-order difference results are comprehensively considered using a linear regression model to generate a data noise index, the formula is: ; , , , are the preset regression coefficients. is the data noise index; The logic for obtaining the cumulative drift index is: Obtaining the baseline data and monitoring data respectively collected by the standard sensor and the target sensor from the drift characteristic information; Time alignment of baseline data and monitoring data to ensure data from the same time points can be compared; Calculate the difference between the baseline data and the monitoring data at each time point to obtain a difference sequence; Based on the difference sequence, the time decay factor is introduced to calculate the cumulative drift index. The formula is: ; , are the start and end time of the collection, is the time decay factor, represents the baseline data at time point t, represents the monitoring data at time point t, Represents the cumulative drift index.

2. According to claim 1, an intelligent supervision method for environmental monitoring quality control is characterized in that: The steps for analyzing fluctuations are: Arrange the sensor data of each category in chronological order to form a time series of sensor data of this category, and then obtain the corresponding data of two adjacent time points in the time series and mark them as , , k represents the sequence number in the time series, and then the following calculation is performed: ; represents the time point corresponding to sequence number k, It represents the amplitude of data change, and then uses the Z-score anomaly detection method to measure the degree of deviation of the current data relative to the historical data, and obtains the Z-score value at the time point corresponding to the sequence number k, and finally obtains a detection sequence containing all Z-score values.

3. According to claim 2, an intelligent supervision method for environmental monitoring quality control is characterized in that: The hidden dangers of identifying whether the target category data has abnormal fluctuations refer to: Each Z-score value in the detection sequence corresponding to the target category data is compared with the preset anomaly threshold, and the sequence number corresponding to each Z-score value greater than the preset anomaly threshold is recorded and summarized as an anomaly point marking set. When the number of data in the anomaly point marking set is non-zero, there is a risk of abnormal fluctuation in the target category data.

4. The method for intelligent supervision of environmental monitoring quality control according to claim 3 is characterized in that: Drift verification refers to: Select a calibrated standard sensor of the same type as the target category data as the reference for drift verification. The two sensors perform data collection and meet the following conditions: Place the target category sensor and the standard sensor in the same environment to ensure that both collect the same type of environmental data at the same time; The acquisition time of the two is synchronized to ensure that the two sets of data have the same time stamp.

5. The method for intelligent supervision of environmental monitoring quality control according to claim 4 is characterized in that: Feature extraction refers to extracting noise feature information and drift feature information from the drift verification dataset respectively.

6. The method for intelligent supervision of environmental monitoring quality control according to claim 5, characterized in that: The logic used in fuzzy logic is: The data noise index and cumulative drift index of the sensor corresponding to the target category data are taken as input variables, the detection drift degree of the sensor corresponding to the target category data is taken as the output variable, the input variables are fuzzified, and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified, and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the detection drift degree under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the detection drift degree of the sensor corresponding to the target category data.

7. The method for intelligent supervision of environmental monitoring quality control according to claim 6 is characterized in that: Obtain all data categories with hidden dangers of abnormal fluctuations, obtain sensor classification results accordingly, perform correction analysis, and select local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition. This means: Get the sensor classification results corresponding to the target category data. Each type of classification result corresponds to a constant value. Mark the constant value corresponding to the classification result of the target sensor numbered r as Cr, and then calculate the optimization value: ; The preset self-influence coefficient of the target sensor numbered r, is the constant value corresponding to the division result of the target sensor numbered j, N represents the total number of all sensors, The preset inter-sensor influence coefficient for the target sensor numbered r, is the interference coefficient between the target sensor numbered r and the sensor numbered j, which is obtained by calculating the correlation of the detection data streams of the target sensor numbered r and the sensor numbered j, namely the Pearson correlation coefficient, and then performing absolute value processing. For the optimization value, if there is a sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the full category calibration operation is selected; if there is no sensor whose optimization value is greater than or equal to the preset optimization judgment threshold, the local calibration operation is selected.

8. An intelligent supervision system for environmental monitoring quality control, implemented based on an intelligent supervision method for environmental monitoring quality control according to any one of claims 1 to 7, characterized in that: include: Environmental data collection module, which collects environmental monitoring data in real time through multiple types of sensors and formats the collected data to ensure data availability; The fluctuation analysis module analyzes the fluctuation of each category of data and identifies whether the target category data has the hidden danger of abnormal fluctuation; Drift verification analysis module: if the target category data has the risk of abnormal fluctuation, the sensor corresponding to the target category data is drift verified to obtain the drift verification data set and perform feature extraction. Then, drift analysis is performed based on the feature extraction results to generate data noise index and cumulative drift index respectively. The type classification module uses fuzzy logic to classify the sensors corresponding to the target category data into three types: severe drift, moderate drift, and low drift based on the data noise index and cumulative drift index; The correction analysis module obtains all data categories with hidden dangers of abnormal fluctuations, obtains the corresponding sensor classification results, performs correction analysis, and selects local calibration or full category calibration operations based on the correction analysis results to optimize the quality of monitoring data acquisition.

Images