Data acquisition method and system based on multi-source water affair sensor

Abstract

The invention relates to the field of data processing, in particular to a data acquisition method and system based on a multi-source water affair sensor, and the method comprises the steps: calculating period relevance and trend relevance, and respectively obtaining abnormal probability predicted values based on a period rule and a trend rule, taking the period relevance and the trend relevance as weights, and carrying out weighted summation on the two abnormal probability predicted values to obtain a comprehensive abnormal probability predicted value; calculating the abnormal relevance of each kind of historical influence factor data, and obtaining the abnormal possibility of the water affair data according to the abnormal relevance; correcting the comprehensive abnormal probability predicted value according to the abnormal possibility of the water affair data to obtain a final abnormal probability predicted value; setting an acquisition frequency according to the final abnormal probability predicted value; and performing water affair data acquisition control based on the acquisition frequency. And the water affair data acquisition frequency is adaptively set, so that abnormal information can be grasped in time while data redundancy is reduced.

Description

Technical Field

[0001] The present invention relates to the field of data processing. More specifically, the present invention relates to a multi-source water sensor data acquisition method and system. Background Art

[0002] In the field of modern water management, it is crucial to ensure the efficient utilization of water resources, water quality safety, and the stable operation of the water supply system. As a key means of obtaining water information, the accuracy, timeliness, and integrity of multi-source water sensor data play a decisive role in water management decisions. By deploying various sensors, such as water level sensors, flow sensors, water quality sensors, etc., at key nodes such as water sources, water supply networks, and sewage treatment plants, important information such as water level changes, water flow velocity, and water quality parameters can be monitored in real time, providing data support for water management.

[0003] However, the existing multi-source water sensor data acquisition methods and systems have certain limitations in practical applications. Traditional sensor data acquisition often uses a fixed acquisition frequency, which fails to fully consider the dynamic characteristics of the operation state of the water system. During the period when the water system operates relatively stably and the probability of water anomalies is small, too high an acquisition frequency will result in the generation of a large amount of redundant data, which not only increases the burden of data storage and transmission, but also consumes too much sensor power and network bandwidth resources, reducing the operating efficiency of the system. On the other hand, at the critical moment when the water system may face abnormal situations, such as a rapid rise in water level caused by heavy rain or sudden water quality pollution, a fixed low acquisition frequency cannot capture the changes in key data in time, resulting in a lag in the monitoring of abnormal situations and making it difficult to take effective countermeasures in time, which may lead to serious water accidents and have an adverse impact on residents' lives and the ecological environment.

[0004] Therefore, there is an urgent practical need to develop a multi-source water sensor data acquisition method and system that can dynamically adjust the sensor acquisition frequency according to the probability of water anomalies. By intelligently setting the acquisition frequency of sensors in different situations, reducing the acquisition frequency when the anomaly probability is small to optimize resource utilization, and increasing the acquisition frequency when the anomaly probability is large to ensure the timeliness and accuracy of data, it will significantly improve the refinement level of water management and the ability to respond to emergencies, effectively avoid resource waste and potential water risks, and promote the development of water management towards a more efficient and intelligent direction. Summary of the Invention

[0005] To solve the problem of how to adaptively adjust the acquisition frequency according to the anomalies of water data, the present invention proposes a multi-source water sensor data acquisition method and system.

[0006] In a first aspect, the present invention provides a multi-source water sensor data acquisition method, including: Obtaining each type of historical water service data at each historical moment and each type of historical influencing factor data at the aligned moment; Splitting the time series sequence composed of the historical water service data into a periodic component and a trend component; calculating the correlation between the feature descriptors and the anomaly probability of all periodic segments of the periodic component, which is denoted as the periodic correlation, obtaining the trend correlation, respectively obtaining the anomaly probability prediction values for future time periods based on the periodic law and the trend law, using the periodic correlation and the trend correlation as the weights of the corresponding anomaly probability prediction values respectively, and performing weighted summation on the two anomaly probability prediction values to obtain the comprehensive anomaly probability prediction value for the future time period, where the future time period includes several future sub-time periods; Calculating the information gain between the historical water service data and each type of historical influencing factor data as the anomaly correlation of each type of historical influencing factor data, obtaining the anomaly possibility of the water service data in the first future sub-time period according to the anomaly correlation and the co-occurrence of each influencing factor data and the abnormal data in the historical water service data; correcting the comprehensive anomaly probability prediction value of the first future sub-time period according to the anomaly possibility of the water service data to obtain the final anomaly probability prediction value; setting the acquisition frequency of the first future sub-time period according to the final anomaly probability prediction value; and performing water service data acquisition control based on the acquisition frequency.

[0007] Preferably, the method for obtaining the feature descriptor of the periodic segment includes: Taking the product of the mean and variance of all historical water service data in the periodic segment as the feature descriptor of the periodic segment.

[0008] Preferably, the method for obtaining the anomaly probability includes: Performing anomaly detection on the historical water service data to obtain the abnormal data in the historical water service data, and denoting the abnormal data in the historical water service data as abnormal water service data; Dividing the number of abnormal water service data in the periodic segment by the total number of data in the periodic segment to obtain the anomaly probability of the periodic segment.

[0009] Preferably, the method for obtaining the trend correlation includes: Performing segmentation processing on the trend component to obtain several trend segments; Obtaining the feature descriptor of the trend segment; Obtaining the anomaly probability of the trend segment; Taking the correlation between the feature descriptors and the anomaly probability of all trend segments in the trend component as the trend correlation.

[0010] Preferably, the method for respectively obtaining the anomaly probability prediction values for future time periods based on the periodic law and the trend law includes: Match the cycle segment at the current moment with each previous cycle segment to obtain a matching value, and use the cycle segments with a matching value greater than a preset threshold as alternative reference segments; use the next cycle segment of the alternative reference segment with the shortest time interval from the current moment as the target cycle segment; use the abnormal probability in the target cycle segment as the abnormal probability prediction value for the future time period based on the cycle law. Use the least squares method to perform polynomial fitting on the abnormal probabilities of all trend segments, and use the fitted polynomial to fit the abnormal probability of the future time period, which is recorded as the abnormal probability prediction value for the future time period based on the trend law.

[0011] Preferably, the method for obtaining the abnormal relevance of each historical influencing factor data includes: Calculate the abnormal information entropy of the historical water service data based on the probability of abnormal water service data and the probability of non-abnormal water service data; divide the influencing factor data into several category levels; obtain the probability of abnormal water service data and the probability of non-abnormal water service data that occur at the alignment moments of all moments when the influencing factor data in any one category level is located, and calculate the information entropy of this category level based on the probability of abnormal water service data and the probability of non-abnormal water service data in this category level; obtain the probability of the influencing factor data in each category level, which is recorded as the probability of each category level, and calculate the information gain based on the probability of each category level and the information entropy of each category level, as the abnormal relevance of each influencing factor data.

[0012] Preferably, the method for obtaining the abnormal possibility of the water service data in the first future sub-period includes: Obtain the category level where the predicted value of each influencing factor data at the alignment moment of each moment in the first future sub-period of the future time period is located, which is recorded as the analysis category level; obtain the probability of abnormal water service data under the analysis category level as the individual co-occurrence of each influencing factor data at each moment and the abnormal data in the historical water service data; use the average value of the individual co-occurrences of all moments in the first future sub-period of the future time period as the comprehensive co-occurrence of the predicted value of each influencing factor data and the abnormal data in the historical water service data; use the proportion of the abnormal relevance of each abnormal factor data as the weight, and perform weighted summation on the comprehensive co-occurrences of the predicted values of all influencing factor data to obtain the abnormal possibility of the water service data in the first future sub-period.

[0013] Preferably, the method for correcting the comprehensive abnormal probability prediction value of the first future sub-period according to the abnormal possibility of the water service data to obtain the final abnormal probability prediction value includes: Multiply the abnormal possibility of the water service data by the comprehensive abnormal probability prediction value of the first future sub-period in the future time period, and then perform normalization processing to obtain the final abnormal probability prediction value of the first future sub-period.

[0014] Preferably, setting the acquisition frequency of the first future sub-period according to the final abnormal probability prediction value includes: Multiplying the final abnormal probability prediction value of the first future sub-period by a preset acquisition frequency to obtain the acquisition frequency of the first future sub-period.

[0015] In a second aspect, the present invention provides a multi-source water sensor data acquisition system, including: a processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned multi-source water sensor data acquisition method is implemented.

[0016] By adopting the above technical solution, the above-mentioned multi-source water sensor data acquisition method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so as to manufacture a terminal device according to the memory and the processor, which is convenient to use.

[0017] The present invention has the following beneficial effects: The present invention adjusts the acquisition frequency of water service data according to the occurrence of abnormalities in the water service data, so that when the probability of abnormality occurrence is relatively large, a larger acquisition frequency is used to collect the water service data, so that the abnormal information can be obtained in a timely manner, providing a basis for making corresponding measures in a timely manner; Furthermore, considering that the probability of occurrence of water service data abnormalities has a certain regularity, the probability of abnormality occurrence in the future period is relatively accurately predicted by analyzing the change law information of historical water service data. The probability of abnormality occurrence in the future period obtained through historical laws reflects more long-term law information; Furthermore, considering that whether water service data abnormalities occur will be affected by other factors, the possibility of water service data abnormalities occurring in the future period is relatively accurately predicted by analyzing the correlation between other factors and the occurrence of water service data abnormalities. The possibility of water service data abnormalities occurring predicted by the correlation between other factors and the occurrence of water service data abnormalities reflects more real-time information.

[0018] Furthermore, the probability of abnormality occurrence is corrected by using the possibility of abnormality occurrence in the future period, so as to combine real-time information and long-term law information, and more accurately predict the occurrence of abnormalities in the future period. Description of the Drawings

[0019] Figure 1 It is a flowchart of the steps of the multi-source water sensor data acquisition method according to an embodiment of the present invention. Detailed Embodiment

[0020] Please refer to Figure 1 , which shows a flowchart of the steps of the multi-source water sensor data acquisition method provided by an embodiment of the present invention. The method includes the following steps: S1: Obtain each type of historical water service data at each historical moment and each type of historical influencing factor data at the aligned moment.

[0021] Preferably, as an example, obtaining each type of historical water service data at each historical moment and each type of historical influencing factor data at the aligned moment includes: Obtain each type of water service data at each historical moment and record it as each type of historical water service data.

[0022] Collect each type of influencing factor data at each historical moment and record it as each type of historical influencing factor data. Match the historical water service data and historical influencing factor data at all historical moments, obtain the moment correspondence relationship between the historical water service data and historical influencing factor data when the matching value is the largest, and perform moment alignment processing on the historical water service data and historical influencing factor data according to the moment correspondence relationship.

[0023] The types of historical water service data include but are not limited to the following aspects: river water level, groundwater level, pH value, dissolved oxygen, conductivity.

[0024] The types of influencing factor data include but are not limited to the following aspects: rainfall, temperature, aquatic organism content, industrial emissions.

[0025] S2: Split the time series composed of the historical water service data into a periodic component and a trend component; calculate the correlation between the characteristic descriptors and the abnormal probability of all periodic segments of the periodic component and record it as the periodic correlation. Obtain the trend correlation, respectively obtain the abnormal probability prediction values for future time periods based on the periodic law and the trend law, use the periodic correlation and the trend correlation as the weights of the corresponding abnormal probability prediction values respectively, and perform weighted summation on the two abnormal probability prediction values to obtain the comprehensive abnormal probability prediction value for the future time period. The future time period includes several future sub - time periods.

[0026] It should be noted that the occurrence of anomalies in water service data has a certain regularity. For example, in summer with more rainfall, the river water level has a greater probability of exceeding the set height and showing anomalies. Therefore, the occurrence of anomalies in water service data has a certain regularity. Thus, the probability of anomalies occurring in future stages can be predicted based on the regularity of the occurrence of water service data anomalies.

[0027] S20: Split the time series composed of the historical water service data into a periodic component and a trend component.

[0028] It should be noted that since the historical water service data contains various laws such as periodicity and trend, analyzing multiple laws mixed together will affect the analysis accuracy. Therefore, each law needs to be analyzed separately. First, perform law splitting.

[0029] Preferably, as an example, splitting the time series composed of the historical water service data into a periodic component and a trend component includes: Using the Seasonal Decomposition of Time Series (STL) method to split the time series composed of the historical water service data into a periodic component and a trend component.

[0030] It should be noted that the method of using the seasonal decomposition method to split the time series composed of the historical water service data into a periodic component and a trend component is a prior art and will not be elaborated here.

[0031] S21: Calculate the correlation between the feature descriptors of all periodic segments of the periodic component and the abnormal probability, denoted as the periodic correlation. Obtain the trend correlation. Respectively obtain the predicted values of the abnormal probability for future time periods based on the periodic law and the trend law. Use the periodic correlation and the trend correlation as the weights of the corresponding predicted values of the abnormal probability, and perform a weighted sum of the two predicted values of the abnormal probability to obtain the comprehensive predicted value of the abnormal probability for the future time period.

[0032] It should be noted that the predicted values of the abnormal probability can be obtained according to both the periodic and trend laws. Therefore, it is necessary to effectively combine the predicted values of the abnormal probability obtained by these laws to obtain a relatively accurate predicted value of the abnormal probability.

[0033] It should be further noted that due to the different correlations between the two laws of period and trend and the occurrence probability of abnormalities, if the correlation between the period and the occurrence probability of abnormal data is relatively large, more reference should be made to the predicted value under the periodic law when predicting the occurrence probability of abnormal data. If the correlation between the trend and the occurrence probability of abnormal data is relatively large, more reference should be made to the predicted value under the trend law when predicting the occurrence probability of abnormal data. Therefore, the predicted values of the abnormal probability obtained by the two laws can be effectively combined based on this.

[0034] Preferably, as an example, calculating the correlation between the feature descriptors of all periodic segments of the periodic component and the abnormal probability, denoted as the periodic correlation. Obtain the trend correlation. Respectively obtain the predicted values of the abnormal probability for future time periods based on the periodic law and the trend law. Use the periodic correlation and the trend correlation as the weights of the corresponding predicted values of the abnormal probability, and perform a weighted sum of the two predicted values of the abnormal probability to obtain the comprehensive predicted value of the abnormal probability for the future time period, including: Denote the correlation between the sequence composed of the feature descriptors of all periodic segments of the periodic component and the sequence composed of the abnormal probability as the periodic correlation.

[0035] Denote the correlation between the sequence composed of the feature descriptors of all trend segments of the trend component and the sequence composed of the abnormal probability as the trend correlation.

[0036] Obtain the abnormal probability prediction value for the future period based on the trend law and the abnormal probability prediction value for the future period based on the periodic law respectively. Take the periodic correlation and the trend correlation as the weights corresponding to the abnormal probability prediction values, and perform weighted summation on the two abnormal probability prediction values to obtain the comprehensive abnormal probability prediction value for the future period.

[0037] It can be understood that the feature descriptor of the periodic segment reflects the information of the periodic segment. If the abnormal probability has a large correlation with the information of the periodic segment, it means that the abnormal probability prediction depends more on the periodic information. Therefore, more periodic information should be referred to when making the abnormal probability prediction, and the weight of the abnormal probability prediction value predicted based on the periodic information should be set larger; the feature descriptor of the trend segment reflects the information of the trend segment. If the abnormal probability has a large correlation with the information of the trend segment, it means that the abnormal probability prediction depends more on the periodic information. Therefore, more trend information should be referred to when making the abnormal probability prediction, and the weight of the abnormal probability prediction value predicted based on the trend information should be set larger.

[0038] The above embodiments involve the periodic segment, the trend segment, the feature descriptors of the periodic segment and the trend segment, as well as the abnormal probability, the future period, the abnormal probability prediction value for the future period based on the trend law, and the abnormal probability prediction value for the future period based on the periodic law. Next, the determination methods of the periodic segment, the trend segment, the feature descriptors of the periodic segment and the trend segment, as well as the abnormal probability, the future period, the abnormal probability prediction value for the future period based on the trend law, and the abnormal probability prediction value for the future period based on the periodic law will be described.

[0039] First, introduce the methods for obtaining the periodic segment and the trend segment.

[0040] Preferably, as an example, the methods for obtaining the periodic segment and the trend segment include: Perform Fourier transform on the periodic component to obtain several frequency components. Take the amplitude of the frequency component as the weight, perform weighted summation on the frequencies of all frequency components to obtain the weighted frequency value, take the reciprocal of the weighted frequency value to obtain the period length, and evenly divide the periodic component into several periodic segments of L, where L represents the period length.

[0041] Take a window with a preset size centered on each data in the trend component, calculate the change rate of each data in the window of each data, calculate the Euclidean distance between the change rates of all data in the window of each data and the change rates of all data in the window of the previous data, which is denoted as the change difference between each data and the previous data. Take the opposite number of the change difference as the exponent and the natural number as the base, perform power calculation to obtain the change similarity between each data and the previous data, and perform clustering processing on all data in the trend component according to the change similarity. The data segment formed by the data in each category obtained by clustering is used as the trend segment.

[0042] It is understandable that the amplitude reflects the proportion of the frequency component in the periodic component. The weighted frequency value obtained by weighting the frequency with the amplitude as the weight can reflect the overall frequency situation of the periodic component. Therefore, segmenting based on the period length obtained from the weighted frequency can better separate each period.

[0043] In addition, the change rate of the data reflects the change trend of the data. By clustering the data in the trend component through the similarity of the change rate, the data with similar change trends are segmented into one segment, and the data with different change trends are segmented into different segments, so as to make each trend segment reflect relatively single trend information.

[0044] Then, the methods for obtaining the feature descriptors and abnormal probabilities of the period segments and trend segments are introduced.

[0045] Preferably, as an example, the methods for obtaining the feature descriptors and abnormal probabilities of the period segments and trend segments include: Taking the product of the mean and variance of all historical water service data in the period segment as the feature descriptor of the period segment; using the LOF algorithm to perform anomaly detection on the historical water service data to obtain the abnormal data in the historical water service data, denoting the abnormal data in the historical water service data as abnormal water service data, and dividing the number of abnormal water service data in the period segment by the total number of data in the period segment to obtain the abnormal probability of the period segment.

[0046] Taking the product of the mean and variance of all historical water service data in the period segment as the feature descriptor of the period segment; using the LOF algorithm to perform anomaly detection on the historical water service data to obtain the abnormal data in the historical water service data, denoting the abnormal data in the historical water service data as abnormal water service data, and dividing the number of abnormal water service data in the period segment by the total number of data in the period segment to obtain the abnormal probability of the period segment.

[0047] Taking the product of the mean and variance of all historical water service data in the trend segment as the feature descriptor of the trend segment; dividing the number of abnormal water service data in the trend segment by the total number of data in the trend segment to obtain the abnormal probability of the trend segment.

[0048] It is understandable that the feature descriptor not only contains value information but also contains data change information. Therefore, the feature descriptor can comprehensively reflect the information situation of each segment, thus providing a basis for accurate correlation analysis.

[0049] Finally, the methods for obtaining the future time period, the predicted value of the abnormal probability of the future time period based on the trend law, and the predicted value of the abnormal probability of the future time period based on the period law are introduced.

[0050] Preferably, as an example, a method for obtaining the future time period and the abnormal probability prediction values of the future time period based on trend rules and the abnormal probability prediction values of the future time period based on periodic rules includes: The upward rounding value of the average of the lengths of all trend segments is denoted as the trend length; the upward rounding value of the average of the trend length and the cycle length is used as the length of the future time period; a time period composed of M consecutive moments starting from the next moment of the current moment is used as the future time period. M represents the length of the future time period.

[0051] The cycle segment where the current moment is located is matched with each previous cycle segment to obtain a matching value, and the cycle segments with a matching value greater than the preset threshold are used as alternative reference segments; the next cycle segment of the alternative reference segment with the shortest time interval from the current moment is used as the target cycle segment; the abnormal probability in the target cycle segment is used as the abnormal probability prediction value of the future time period based on periodic rules; The least squares method is used to perform polynomial fitting on the abnormal probabilities of all trend segments, and the abnormal probability of the future time period fitted by the fitted polynomial is denoted as the abnormal probability prediction value of the future time period based on trend rules.

[0052] It can be understood that the more similar the cycle segment where the current moment is located is to the current moment and the shorter the time interval from the current time, the more similar the cycle change rule is to the current moment. Therefore, the cycle segment with the most similar cycle change rule to the current moment can be obtained through the similarity of the cycle change rule and the interval duration with the current moment. Therefore, the next cycle segment of the cycle segment with the most similar cycle change rule to the current moment can better reflect the cycle change information of the future time period.

[0053] S3: Calculate the information gain between the historical water service data and each historical influencing factor data as the abnormal correlation of each historical influencing factor data, and obtain the abnormal possibility of the water service data in the first future sub-time period according to the abnormal correlation and the co-occurrence of each influencing factor data and the abnormal data in the historical water service data; correct the comprehensive abnormal probability prediction value of the first future sub-time period according to the abnormal possibility of the water service data to obtain the final abnormal probability prediction value; set the acquisition frequency of the first future sub-time period according to the final abnormal probability prediction value; perform water service data acquisition control based on the acquisition frequency.

[0054] S30: Calculate the information gain between the historical water service data and each historical influencing factor data as the abnormal correlation of each historical influencing factor data, and obtain the abnormal possibility of the water service data in the first future sub-time period according to the abnormal correlation and the co-occurrence of each influencing factor data and the abnormal data in the historical water service data.

[0055] It should be noted that the abnormal probability obtained through historical rules cannot reflect the real-time abnormal situation. Therefore, the abnormal probability obtained only through historical rules cannot accurately reflect the abnormal situation at each moment. Since water service anomalies are affected by other factors, for example, precipitation can cause the river water level to rise, the real-time situation of water service anomalies can be analyzed by analyzing the influence relationship of other factors on water service anomalies.

[0056] Preferably, as an example, calculate the information gain between the historical water service data and each historical influencing factor data as the anomaly relevance of each historical influencing factor data, and obtain the probability of water service data anomalies in the first future sub-period according to the anomaly relevance and the co-occurrence of each influencing factor data and abnormal data in the historical water service data, including: First, obtain the anomaly relevance: , where represents the occurrence probability of abnormal water service data, represents the occurrence probability of non-abnormal water service data, represents the logarithmic function with base 2, and S0 represents the abnormal information entropy.

[0057] Among them, in all the alignment times of the moments when the influencing factor data is located in any one category layer, obtain the probability of abnormal water service data and the probability of non-abnormal water service data, which are recorded as the occurrence probability of abnormal water service data under this category layer and the occurrence probability of non-abnormal water service data under this category layer. represents the occurrence probability of abnormal water service data under the i-th category layer, represents the occurrence probability of non-abnormal water service data under the i-th category layer, represents the information entropy of the i-th category layer.

[0058] , represents the number of category layers, represents the information gain.

[0059] Take the information gain as the anomaly relevance of each historical influencing factor data.

[0060] It can be understood that the information gain reflects the determination situation of abnormal water service data under the condition that the influencing factor data is determined. In other words, it reflects the influence situation of the influencing factor data on the anomalies of water service data. The larger this value is, the greater the influence of the influencing factor data on the anomalies of water service data, and thus the greater the anomaly relevance of the influencing factor data.

[0061] Then, based on the anomaly correlation and the co-occurrence of each influencing factor data with the abnormal data in the historical water service data, obtain the probability of water service data anomaly in the first future sub-period.

[0062] Obtain the category layer where the predicted values of each influencing factor data at the alignment moments of each moment in the first future sub-period of the future period are located, and record it as the analysis category layer; obtain the probability of abnormal water service data under the analysis category layer as the individual co-occurrence of each influencing factor data and the abnormal data in the historical water service data at each moment; take the average value of the individual co-occurrences of all moments in the first future sub-period of the future period as the comprehensive co-occurrence of the predicted values of each influencing factor data and the abnormal data in the historical water service data; use the proportion of the anomaly correlation of each abnormal factor data as the weight, and perform a weighted sum of the comprehensive co-occurrences of the predicted values of all influencing factor data to obtain the probability of water service data anomaly in the first future sub-period.

[0063] It can be understood that the individual co-occurrence reflects the probability of abnormal water service data occurring under the condition that the predicted value of the influencing factor data occurs. The anomaly correlation reflects the definite influence of the influencing factor data on the abnormal water service data. By taking the abnormal influence situation as the weight, the individual co-occurrences of all influencing factor data are combined to comprehensively determine the probability of occurrence of abnormal water service data in the future sub-period.

[0064] The above embodiments involve the category layer and the predicted values of each influencing factor data at the alignment moments. Next, the determination methods of the category layer and the predicted values of each influencing factor data at the alignment moments need to be described.

[0065] First, determine the category layer.

[0066] Preferably, as an example, the classification method of the category layer includes: Classify the category layer according to the classification method of the field where each influencing factor data is located. Taking precipitation as an example, when the rainfall is in the interval (0, 10], it is determined as light rain; when the rainfall is in the interval (10, 24.9), it is medium rain; when the rainfall is in the interval (25, 49.9), it is determined as heavy rain; when the rainfall is in the interval (50, 99.9), it is determined as rainstorm; when the rainfall is greater than 100, it is determined as heavy rainstorm. The category layer is light rain, medium rain, heavy rain, rainstorm, and heavy rainstorm.

[0067] Then, determine the predicted values of each influencing factor data at the alignment moments.

[0068] Preferably, as an example, the determination method of the predicted values of each influencing factor data at the alignment moments includes: If the alignment moment is before the current moment, take the actual value of each influencing factor data at the alignment moment as the predicted value of each influencing factor data at the alignment moment.

[0069] If the alignment time is after the current time, the predicted values of each influencing factor data are obtained by using the prediction method in the field where each influencing factor data is located. Taking rainfall as an example, the rainfall predicted value is the rainfall predicted in the existing weather forecast for the alignment time.

[0070] It should be added that the method for obtaining the future sub-periods includes: The future period is evenly divided into K sub-periods, denoted as future sub-periods, where K represents the preset number of divisions.

[0071] It should be noted that the future period has a long span, and the occurrence of anomalies changes in real time. Therefore, it is not accurate to adopt a fixed acquisition frequency throughout the long-span period.

[0072] S31: Modify the predicted value of the comprehensive anomaly probability for the first future sub-period according to the anomaly possibility of the water service data to obtain the final predicted value of the anomaly probability.

[0073] Preferably, as an example, modifying the predicted value of the comprehensive anomaly probability for the first future sub-period according to the anomaly possibility of the water service data to obtain the final predicted value of the anomaly probability includes: Multiply the anomaly possibility of the water service data by the predicted value of the comprehensive anomaly probability for the first future sub-period in the future period and then perform normalization processing to obtain the final predicted value of the anomaly probability for the first future sub-period.

[0074] S32: Set the acquisition frequency for the first future sub-period according to the final predicted value of the anomaly probability.

[0075] Preferably, as an example, setting the acquisition frequency for the first future sub-period according to the final predicted value of the anomaly probability includes: Multiply the final predicted value of the anomaly probability for the first future sub-period by the preset acquisition frequency to obtain the acquisition frequency for the first future sub-period.

[0076] S33: Control the acquisition of water service data based on the acquisition frequency.

[0077] The embodiment of the present invention also discloses a multi-source water service sensor data acquisition system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the multi-source water service sensor data acquisition method according to the present invention is implemented.

[0078] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be elaborated here.

[0079] In the present invention, the foregoing memory may be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as, for example, a resistive random access memory, a dynamic random access memory, a static random access memory, an enhanced dynamic random access memory, a high bandwidth memory, a hybrid memory cube, etc., or any other medium that can be used to store the required information and can be accessed by an application program, a module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

Claims

1. A multi-source water sensor data acquisition method, characterized in that, Including: Obtaining each type of historical water service data at each historical moment and each type of historical influencing factor data at the alignment moment; Splitting the time series sequence composed of the historical water service data into a periodic component and a trend component; calculating the correlation between the characteristic descriptors and the anomaly probability of all periodic segments of the periodic component, which is denoted as the periodic correlation, obtaining the trend correlation, respectively obtaining the anomaly probability prediction values for future time periods based on the periodic law and based on the trend law, using the periodic correlation and the trend correlation as the weights for the corresponding anomaly probability prediction values respectively, and performing weighted summation on the two anomaly probability prediction values to obtain the comprehensive anomaly probability prediction value for the future time period, where the future time period includes several future sub-time periods; Calculating the information gain between the historical water service data and each type of historical influencing factor data as the anomaly correlation of each type of historical influencing factor data, and obtaining the anomaly possibility of the water service data in the first future sub-time period according to the anomaly correlation and the co-occurrence of each influencing factor data and the abnormal data in the historical water service data; Correcting the comprehensive anomaly probability prediction value for the first future sub-time period according to the anomaly possibility of the water service data to obtain the final anomaly probability prediction value; Setting the acquisition frequency for the first future sub-time period according to the final anomaly probability prediction value; Performing water service data acquisition control based on the acquisition frequency.

2. The method for collecting multi-source water sensor data according to claim 1, wherein, The method for obtaining the characteristic descriptor of the periodic segment includes: Taking the product of the mean and variance of all historical water service data in the periodic segment as the characteristic descriptor of the periodic segment.

3. The multi-source water sensor data acquisition method according to claim 1, wherein The method for obtaining the anomaly probability includes: Performing anomaly detection on the historical water service data to obtain the abnormal data in the historical water service data, and denoting the abnormal data in the historical water service data as abnormal water service data; Dividing the number of abnormal water service data in the periodic segment by the total number of data in the periodic segment to obtain the anomaly probability of the periodic segment.

4. The multi-source water sensor data acquisition method according to claim 1, characterized in that, The method for obtaining the trend correlation includes: Performing segmentation processing on the trend component to obtain several trend segments; Obtaining the characteristic descriptor of the trend segment; Obtaining the anomaly probability of the trend segment; Taking the correlation between the characteristic descriptors and the anomaly probability of all trend segments in the trend component as the trend correlation.

5. The method for collecting multi-source water sensor data according to claim 1, characterized in that The method for respectively obtaining the anomaly probability prediction values for future time periods based on the periodic law and based on the trend law includes: Performing matching processing on the periodic segment where the current moment is located with the previous periodic segments to obtain a matching value, and taking the periodic segments with a matching value greater than a preset threshold as alternative reference segments; taking the next periodic segment of the alternative reference segment with the shortest time interval from the current moment as the target periodic segment; taking the anomaly probability in the target periodic segment as the anomaly probability prediction value for the future time period based on the periodic law; Using the least squares method to perform polynomial fitting on the anomaly probabilities of all trend segments, and using the fitted polynomial to fit the anomaly probability for the future time period, which is denoted as the anomaly probability prediction value for the future time period based on the trend law.

6. The method for collecting multi-source water sensor data according to claim 3, characterized in that The method for obtaining the anomaly correlation of each type of historical influencing factor data includes: Calculate the abnormal information entropy of historical water service data based on the probability of abnormal water service data and the probability of non-abnormal water service data; divide the influencing factor data into several category levels; obtain the probability of abnormal water service data and the probability of non-abnormal water service data that do not appear at all alignment times of the influencing factor data in any category level, and record them as the probability of abnormal water service data and the probability of non-abnormal water service data at this category level. Calculate the information entropy of this category level based on the probability of abnormal water service data and the probability of non-abnormal water service data at this category level; obtain the probability of the influencing factor data in each category level and record it as the probability of each category level. Calculate the information gain based on the probability of each category level and the information entropy of each category level, and use it as the abnormal relevance of each influencing factor data.

7. The multi-source water sensor data acquisition method according to claim 6, wherein The obtaining of the abnormal possibility of water service data in the first future sub-period includes: Obtain the category level where the predicted value of each influencing factor data at the alignment time of each moment in the first future sub-period of the future period is located, and record it as the analysis category level; obtain the probability of abnormal water service data under the analysis category level as the individual co-occurrence of each influencing factor data at each moment and the abnormal data in the historical water service data; take the mean value of the individual co-occurrences of all moments in the first future sub-period of the future period as the comprehensive co-occurrence of the predicted value of each influencing factor data and the abnormal data in the historical water service data; use the proportion of the abnormal relevance of each abnormal factor data as the weight, and perform weighted summation on the comprehensive co-occurrence of the predicted values of all influencing factor data to obtain the abnormal possibility of water service data in the first future sub-period.

8. The multi-source water sensor data acquisition method according to claim 1, wherein, The correction of the comprehensive abnormal probability prediction value of the first future sub-period according to the abnormal possibility of water service data to obtain the final abnormal probability prediction value includes: Multiply the abnormal possibility of water service data by the comprehensive abnormal probability prediction value of the first future sub-period in the future period, and then perform normalization processing to obtain the final abnormal probability prediction value of the first future sub-period.

9. The multi-source water sensor data acquisition method according to claim 1, wherein The setting of the acquisition frequency of the first future sub-period according to the final abnormal probability prediction value includes: Multiply the final abnormal probability prediction value of the first future sub-period by the preset acquisition frequency to obtain the acquisition frequency of the first future sub-period.

10. A multi-source water sensor data acquisition system, characterized in that, Includes: A processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the multi-source water service sensor data acquisition method according to any one of claims 1-9 is implemented.

Images