Liquid level prediction method and system based on wavelet transform and machine learning

Abstract

The invention provides a liquid level prediction method and system based on wavelet transformation and machine learning, and the method comprises the steps: collecting historical liquid level data, carrying out the wavelet transformation of the historical liquid level data, and obtaining the feature information of the historical liquid level data; the feature information of the historical liquid level data serves as input of a machine learning model, the machine learning model is trained according to the feature information, and a liquid level prediction model is obtained; acquiring current liquid level data in real time, transmitting the current liquid level data to the liquid level prediction model to identify the liquid level data, and outputting a predicted value of the current liquid level; and performing liquid level conversion detection based on the predicted value of the liquid level of the current liquid level, and generating a liquid level detection report. And the accuracy and the real-time performance of detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of liquid level data processing, and in particular to a liquid level prediction method and system based on wavelet transform and machine learning. Background Art

[0002] At present, in industrial production and daily life, liquid level measurement is an important detection task. The accuracy and reliability of liquid level measurement directly affect the stability and safety of the production process. Therefore, by predicting the liquid level measurement data, the production threshold can be effectively set in advance, ensuring the safety and stability of production. At the same time, liquid level prediction helps to optimize the production process, thereby improving production efficiency. However, traditional liquid level prediction methods usually rely on liquid level data detection and artificially output the predicted value of the liquid level through work experience, which wastes a lot of manpower and material resources. There are also physical models or statistical methods to predict the liquid level, which have certain limitations, resulting in high costs and low analysis accuracy. Therefore, the present invention provides a liquid level prediction method and system based on wavelet transform and machine learning. Summary of the invention

[0003] The present invention provides a liquid level prediction method based on wavelet transform and machine learning, which is used for collecting historical liquid level data, thereby performing wavelet transform on the historical liquid level data, effectively obtaining characteristic information of the historical liquid level data, and using the characteristic information of the historical liquid level data as the input of a machine learning model, thereby realizing the training of the machine learning model through the characteristic information, effectively ensuring the effectiveness and accuracy of the liquid level prediction model, and then accurately realizing the effective output of the predicted value of the current liquid level, thereby realizing liquid level change detection through the predicted value of the current liquid level, and then effectively generating a detection report, which not only improves the accuracy and effectiveness of liquid level prediction, but also effectively guarantees the readability of data through the generation of the detection report and is conducive to grasping the current liquid level status as soon as possible, thereby improving the accuracy and real-time performance of the detection.

[0004] The present invention provides a liquid level prediction method based on wavelet transform and machine learning, comprising: Step 1: Collect historical liquid level data, and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level data; Step 2: Use the feature information of the historical liquid level data as the input of the machine learning model, and train the machine learning model based on the feature information to obtain a liquid level prediction model; Step 3: Collect the current liquid level data in real time, and transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; Step 4: Perform liquid level change detection based on the predicted value of the current liquid level and generate a liquid level detection report.

[0005] Preferably, a liquid level prediction method based on wavelet transform and machine learning, in step 1, historical liquid level data is collected, and the historical liquid level data is subjected to wavelet transform to obtain characteristic information of the historical liquid level data, including: Acquire the collection requirements for collecting historical liquid level data, wherein the collection requirements include: data types of the historical liquid level data and the data collection volume corresponding to each data type; Generate a first traversal factor according to the data type of the historical liquid level data in the collection requirement; Generate a second traversal factor according to the data collection amount corresponding to the data type in the collection requirement; Inputting the first traversal factor into the preset database for a first traversal, and locating a data sequence consistent with the data type in the preset database according to the first traversal result; Perform a second traversal on the data sequence with the same data type in the preset database based on the second traversal factor, and collect historical liquid level data in the preset database according to the second traversal result; The historical liquid level data is subjected to wavelet transformation to obtain the characteristic information of the historical liquid level data.

[0006] Preferably, a liquid level prediction method based on wavelet transform and machine learning performs wavelet transform on historical liquid level data to obtain characteristic information of the historical liquid level data, including: Perform wavelet transform on the collected historical liquid level data to obtain the frequency domain characteristic values ​​of the historical liquid level data; Obtain a frequency domain characteristic threshold, and compare the frequency domain characteristic value of the historical liquid level data with the frequency domain characteristic threshold; Based on the comparison results, the data frequency band corresponding to the frequency domain characteristic value of the historical liquid level data that is greater than the frequency domain characteristic threshold is collected and used as the key analysis section; Obtain analysis requirements, perform signal analysis on key analysis sections based on the analysis requirements, and output characteristic information of historical liquid level data corresponding to each analysis requirement.

[0007] Preferably, a liquid level prediction method based on wavelet transform and machine learning, in step 2, the feature information of the historical liquid level data is used as the input of the machine learning model, and the machine learning model is trained according to the feature information to obtain the liquid level prediction model, including: Acquire characteristic information of the historical liquid level data, and analyze the characteristic information based on a preset regression model to determine the subordination of different characteristic information to the liquid level; The feature information is screened based on the relative size relationship between the reference threshold and the degree of subordination to obtain key feature information, and the time domain features and frequency domain features of the key feature information are extracted; Generate index tags based on time domain features, and traverse the spatial information at the same time in the database according to the index tags based on the time and space features to obtain the variable factors corresponding to the key feature information; Based on the time domain features, the frequency domain features and the variable factors are bit-mapped, and an analysis data group is obtained based on the bit-mapped results; Repeatedly obtain the analysis data groups formed by the frequency domain characteristics and variable factors of the key feature information at different times, and bind the analysis data groups at different times to obtain the control data group; Based on the neural network, each analysis data group in the control data group is analyzed to obtain the nonlinear relationship between the frequency domain characteristics and variable factors at different times, and the nonlinear relationship of each analysis data group in the control group is summarized; Performing difference correction on the nonlinear relationship based on the summary result, and obtaining the target nonlinear relationship based on the difference correction result; Iteratively train the machine learning model based on the nonlinear relationship, and obtain an initial liquid level prediction model based on the iterative training results; A reference data layer is allocated in the initial liquid level prediction model, and the frequency domain characteristics of key feature information and the target nonlinear relationship between variable factors are packaged and stored in the reference data layer; The access rights of the reference data layer are adapted based on the storage results, and the liquid level prediction model is obtained based on the adaptation results.

[0008] Preferably, a liquid level prediction method based on wavelet transform and machine learning obtains an initial liquid level prediction model based on iterative training results, comprising: Obtain a test liquid level data set, and obtain a machine learning model after each iterative training based on the number of iterative training; Testing the machine learning model after each iteration of training based on the test liquid level data set, and determining the model loss value of the machine learning model after each iteration of training based on the data difference between the test result and the standard result corresponding to the test liquid level data set; A target iterative training result is selected from the number of iterative training times based on the model loss value, and the target iterative training result is used as the initial liquid level prediction model.

[0009] Preferably, a liquid level prediction method based on wavelet transform and machine learning, in step 3, real-time acquisition of current liquid level data, and transmission of the current liquid level data to a liquid level prediction model to identify the liquid level data, and output a predicted value of the current liquid level, including: Obtain a non-contact sensor and connect the non-contact sensor to a data acquisition terminal; Based on the data acquisition terminal, the current liquid level data collected by the non-contact sensor is acquired in real time. At the same time, the current liquid level data is processed to obtain the standard liquid level data; The standard liquid level data is transmitted to the liquid level prediction model for analysis to determine the data features corresponding to the standard liquid level data; Matching the data features of the standard liquid level data with the reference features stored in the liquid level prediction model, and obtaining the target reference features corresponding to the standard liquid level data according to the matching results, and analyzing the target reference features based on the liquid level prediction model to output the first prediction value of the standard liquid level data corresponding to the next time segment; The reference features of the previous time segment of the standard liquid level data are retrieved in the liquid level prediction model, and the reference features of the previous time segment of the standard liquid level data and the target reference features corresponding to the standard liquid level data are analyzed simultaneously according to the liquid level prediction model, and a second prediction value of the standard liquid level data corresponding to the next time segment is output; The first predicted value is compared with the second predicted value, and the predicted value of the current liquid level is output according to the comparison result.

[0010] Preferably, a liquid level prediction method based on wavelet transform and machine learning compares the first prediction value with the second prediction value and outputs the prediction value of the current liquid level according to the comparison result, including: If the first predicted value is equal to the second predicted value, the first predicted value or the second predicted value is used as the predicted value of the current liquid level; If the first predicted value is not equal to the second predicted value, a target mean of the first predicted value and the second predicted value is calculated, and the target mean is used as the predicted value of the current liquid level.

[0011] Preferably, a liquid level prediction method based on wavelet transform and machine learning, in step 4, performing liquid level transformation detection based on the liquid level prediction value of the current liquid level, and generating a liquid level detection report, includes: Read the predicted value of the current liquid level and the current liquid level data, and calculate the target liquid level difference between the predicted value of the current liquid level and the current liquid level data; Obtain the preset reference liquid level difference interval; Compare the preset reference liquid level difference interval with the target liquid level difference; When the target liquid level difference falls within the preset reference liquid level difference interval, a first liquid level detection report is generated; When the target liquid level difference does not belong to the preset reference liquid level difference interval, a second liquid level detection report is generated.

[0012] Preferably, a liquid level prediction method based on wavelet transform and machine learning comprises: The first liquid level detection report includes: current liquid level data, liquid level prediction value of the current liquid level and target liquid level difference; The second liquid level detection report includes: current liquid level data, liquid level prediction value of the current liquid level, target liquid level difference and alarm signal data.

[0013] The present invention provides a liquid level prediction system based on wavelet transform and machine learning, comprising: The characteristic information determination module is used to collect historical liquid level data and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level signal; A model building module is used to use the characteristic information of the historical liquid level data as the input of the machine learning model, and train the machine learning model according to the characteristic information to obtain a liquid level prediction model; The prediction module is used to collect the current liquid level data in real time, transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; The report generation module is used to record the predicted value of each liquid level, perform liquid level change detection, and generate a liquid level detection report.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. By collecting historical liquid level data, the historical liquid level data is subjected to wavelet transform to effectively obtain the characteristic information of the historical liquid level data. By using the characteristic information of the historical liquid level data as the input of the machine learning model, the machine learning model can be trained through the characteristic information, which effectively guarantees the effectiveness and accuracy of the liquid level prediction model, and then accurately realizes the effective output of the predicted value of the current liquid level, thereby realizing liquid level change detection through the predicted value of the current liquid level, and then effectively generating a test report, which not only improves the accuracy and effectiveness of liquid level prediction, but also effectively guarantees the readability of the data by generating a test report, and is conducive to grasping the current liquid level status as soon as possible, thereby improving the accuracy and real-time performance of the detection.

[0015] 2. By screening the obtained feature information, it is possible to screen out the associated feature information that is strongly correlated with the liquid level. Secondly, the key feature information is analyzed to accurately and effectively determine the non-threaded relationship between the frequency domain characteristics and variable factors of the key feature information at different times, thereby providing a training basis for the machine learning model. Finally, the obtained nonlinear relationship is used to accurately and effectively train the machine learning model, and the liquid level prediction model is effectively acquired, which greatly facilitates liquid level prediction.

[0016] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0017] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Flow chart of a liquid level prediction method based on wavelet transform and machine learning in an embodiment of the present invention; Figure 2 This is a flowchart of step 1 in a liquid level prediction method based on wavelet transform and machine learning in an embodiment of the present invention; Figure 3 It is a structural diagram of a liquid level prediction system based on wavelet transform and machine learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0020] Example 1 This embodiment provides a liquid level prediction method based on wavelet transform and machine learning. Figure 1 As shown, including: Step 1: Collect historical liquid level data, and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level data; Step 2: Use the feature information of the historical liquid level data as the input of the machine learning model, and train the machine learning model based on the feature information to obtain a liquid level prediction model; Step 3: Collect the current liquid level data in real time, and transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; Step 4: Perform liquid level change detection based on the predicted value of the current liquid level and generate a liquid level detection report.

[0021] In this embodiment, the characteristic information of the historical liquid level data includes: frequency characteristic information, which is used to characterize the energy distribution of the liquid level data at different frequencies; time characteristic information, which is used to characterize the rise time, fall time and duration of the liquid level data; amplitude characteristic information, which is used to characterize the maximum and minimum values ​​of the liquid level data; and trend characteristic information, which is used to characterize the rise, fall and stable trend of the liquid level.

[0022] In this embodiment, the current liquid level data may be collected by a non-contact sensor such as an ultrasonic sensor, a radar sensor, etc.

[0023] In this embodiment, the collected liquid level data is subjected to wavelet transformation to extract characteristic information of the liquid level signal. Wavelet transformation can effectively separate the time domain and frequency domain characteristics of the signal, which is convenient for subsequent machine learning algorithm processing.

[0024] The working principle and beneficial effects of the above technical solution are: by collecting historical liquid level data, the historical liquid level data is subjected to wavelet transform to effectively obtain the characteristic information of the historical liquid level data, and by using the characteristic information of the historical liquid level data as the input of the machine learning model, the machine learning model can be trained through the characteristic information, thereby effectively ensuring the effectiveness and accuracy of the liquid level prediction model, and then accurately realizing the effective output of the predicted value of the current liquid level, thereby realizing liquid level change detection through the predicted value of the current liquid level, and then effectively generating a test report, which not only improves the accuracy and effectiveness of the liquid level prediction, but also effectively ensures the readability of the data by generating a test report and is conducive to grasping the current liquid level status as soon as possible, thereby improving the accuracy and real-time performance of the detection.

[0025] Example 2 Based on Example 1, this example provides a liquid level prediction method based on wavelet transform and machine learning, such as Figure 2 As shown, in step 1, historical liquid level data is collected and wavelet transformed to obtain characteristic information of the historical liquid level data, including: S101: Acquire a collection requirement for collecting historical liquid level data, wherein the collection requirement includes: a data type of the historical liquid level data and a data collection amount corresponding to each data type; S102: Generate a first traversal factor according to the data type of the historical liquid level data in the collection requirement; S103: Generate a second traversal factor according to the data collection amount corresponding to the data type in the collection requirement; S104: inputting the first traversal factor into a preset database for a first traversal, and locating a data sequence consistent with the data type in the preset database according to the first traversal result; S105: performing a second traversal on a data sequence with consistent data types in a preset database based on a second traversal factor, and collecting historical liquid level data in the preset database according to the second traversal result; S106: Perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level data.

[0026] In this embodiment, the collection requirements may be set in advance, representing the goal of collecting historical liquid level data, wherein the collection requirements include: the data type of the historical liquid level data, the data collection volume corresponding to each data type, wherein the data type may be int, float, etc.

[0027] In this embodiment, the first traversal factor may be determined according to the data type of the historical liquid level data, and is a tool for locating the data type of the historical liquid level data in a preset database.

[0028] In this embodiment, the second traversal factor may be determined according to the data collection amount corresponding to the data type, and is used to implement a tool for delineating data in a data sequence according to the data collection amount.

[0029] In this embodiment, the preset database may be set in advance.

[0030] The working principle and beneficial effects of the above technical solution are: by obtaining the collection requirements for historical liquid level data, the first traversal factor and the second traversal factor are effectively and accurately generated, and then the first traversal factor and the second traversal factor are effectively used to realize the first traversal in the preset database, and the second traversal after determining the data sequence in the first traversal result, and then the historical liquid level data is collected in the preset database through the second traversal result, and wavelet transformation is performed according to the historical liquid level data to obtain the characteristic information of the historical liquid level data, thereby improving the accuracy and reliability of determining the characteristic information of the historical liquid level data.

[0031] Example 3 Based on Example 2, this example provides a liquid level prediction method based on wavelet transform and machine learning, which performs wavelet transform on historical liquid level data to obtain characteristic information of the historical liquid level data, including: Perform wavelet transform on the collected historical liquid level data to obtain the frequency domain characteristic values ​​of the historical liquid level data; Obtain a frequency domain characteristic threshold, and compare the frequency domain characteristic value of the historical liquid level data with the frequency domain characteristic threshold; Based on the comparison results, the data frequency band corresponding to the frequency domain characteristic value of the historical liquid level data that is greater than the frequency domain characteristic threshold is collected and used as the key analysis section; Obtain analysis requirements, perform signal analysis on key analysis sections based on the analysis requirements, and output characteristic information of historical liquid level data corresponding to each analysis requirement.

[0032] In this embodiment, the frequency domain feature threshold may be set in advance and used as a criterion for dividing the key analysis sections.

[0033] In this embodiment, the key analysis segments have large fluctuations, specific frequency characteristics, etc., and play a key role in understanding the change trend of the liquid level, abnormal conditions, or information related to specific phenomena. They are selected from the overall data for more in-depth and focused analysis, which can more efficiently obtain valuable information and insights.

[0034] In this embodiment, the analysis requirement may be set according to actual needs, so as to obtain characteristic information of the corresponding historical liquid level data.

[0035] The working principle and beneficial effects of the above technical solution are: by performing wavelet transform on the collected historical liquid level data, the frequency domain eigenvalues ​​of the historical liquid level data are effectively obtained, and by comparing the frequency domain eigenvalues ​​of the historical liquid level data with the frequency domain characteristic threshold, the key analysis section is determined, thereby effectively improving the accuracy and efficiency of signal analysis, and then accurately performing signal analysis on the key analysis section through analysis requirements, outputting the characteristic information of the historical liquid level data corresponding to each analysis requirement, ensuring the validity of the obtained characteristic information, and providing a data basis for liquid level prediction.

[0036] Example 4 Based on Example 1, this example provides a liquid level prediction method based on wavelet transform and machine learning. In step 2, the feature information of the historical liquid level data is used as the input of the machine learning model, and the machine learning model is trained according to the feature information to obtain the liquid level prediction model, including: Acquire characteristic information of the historical liquid level data, and analyze the characteristic information based on a preset regression model to determine the subordination of different characteristic information to the liquid level; The feature information is screened based on the relative size relationship between the reference threshold and the degree of subordination to obtain key feature information, and the time domain features and frequency domain features of the key feature information are extracted; Generate index tags based on time domain features, and traverse the spatial information at the same time in the database according to the index tags based on the time and space features to obtain the variable factors corresponding to the key feature information; Based on the time domain features, the frequency domain features and the variable factors are bit-mapped, and an analysis data group is obtained based on the bit-mapped results; Repeatedly obtain the analysis data groups formed by the frequency domain characteristics and variable factors of the key feature information at different times, and bind the analysis data groups at different times to obtain the control data group; Based on the neural network, each analysis data group in the control data group is analyzed to obtain the nonlinear relationship between the frequency domain characteristics and variable factors at different times, and the nonlinear relationship of each analysis data group in the control group is summarized; Performing difference correction on the nonlinear relationship based on the summary result, and obtaining the target nonlinear relationship based on the difference correction result; Iteratively train the machine learning model based on the nonlinear relationship, and obtain an initial liquid level prediction model based on the iterative training results; A reference data layer is allocated in the initial liquid level prediction model, and the frequency domain characteristics of key feature information and the target nonlinear relationship between variable factors are packaged and stored in the reference data layer; The access rights of the reference data layer are adapted based on the storage results, and the liquid level prediction model is obtained based on the adaptation results.

[0037] In this embodiment, the preset regression model is set in advance and is used to determine the subordination of different characteristic information to the liquid level, wherein the subordination degree may be the correlation relationship between different characteristic information and the liquid level.

[0038] In this embodiment, the reference threshold is known in advance and is used as a reference for measuring whether the degree of subordination meets the requirements.

[0039] In this embodiment, the key feature information may be feature information that meets the benchmark requirements after screening the feature information.

[0040] In this embodiment, the time domain features and the frequency domain features may respectively characterize the specific states of the key feature information in the time domain and the frequency domain, including the distribution of values ​​and the like.

[0041] In this embodiment, the index tag may be generated according to the time domain feature and is a reference for accessing the information in the database.

[0042] In this embodiment, the spatiotemporal feature may be a corresponding relationship between the same time and space.

[0043] In this embodiment, the variable factors may be all factors that affect the liquid level at the same time.

[0044] In this embodiment, bit-by-bit mapping may be a one-to-one matching of frequency domain features and variable factors.

[0045] In this embodiment, the analysis data group may be a result obtained by associating the frequency domain features with the variable factors.

[0046] In this embodiment, the control data group may be a result of binding analysis data groups at different times, that is, a plurality of different analysis data groups.

[0047] In this embodiment, the difference correction may be to correct the nonlinear relationship through multiple sets of data analysis results, so as to obtain a correct nonlinear relationship, that is, a target nonlinear relationship.

[0048] In this embodiment, the reference data layer may be a data layer configured in the initial liquid level prediction model, and is used to store the target nonlinear relationship, thereby facilitating the initial liquid level prediction model to perform liquid level prediction operations.

[0049] The working principle and beneficial effects of the above technical solution are: by screening the obtained feature information, it is possible to screen out the associated feature information that is strongly correlated with the liquid level from the feature information; secondly, the key feature information is analyzed to accurately and effectively determine the non-threaded relationship between the frequency domain characteristics and variable factors of the key feature information at different times, thereby providing a training basis for the machine learning model; finally, the machine learning model is accurately and effectively trained through the obtained nonlinear relationship, and the liquid level prediction model is effectively acquired, which provides great convenience for liquid level prediction.

[0050] Example 5 On the basis of Example 4, this embodiment provides a liquid level prediction method based on wavelet transform and machine learning, and obtains an initial liquid level prediction model based on iterative training results, including: Obtain a test liquid level data set, and obtain a machine learning model after each iterative training based on the number of iterative training; Testing the machine learning model after each iteration of training based on the test liquid level data set, and determining the model loss value of the machine learning model after each iteration of training based on the data difference between the test result and the standard result corresponding to the test liquid level data set; A target iterative training result is selected from the number of iterative training times based on the model loss value, and the target iterative training result is used as the initial liquid level prediction model.

[0051] In this embodiment, the test liquid level data set is known in advance and is used to test whether the performance of the initial liquid level prediction model meets the requirements.

[0052] In this embodiment, the standard result may be a result corresponding to the test liquid level data set.

[0053] In this embodiment, the model loss value may be a representation of the performance loss of the machine learning model obtained after each iterative training. A larger value indicates a worse performance of the model.

[0054] In this embodiment, selecting the target iterative training result from the number of iterative training times based on the model loss value may be selecting the result with the smallest model loss value from the iterative training as the target iterative training result.

[0055] The working principle and beneficial effect of the above technical solution are: by testing each iterative training result, the most effective training result is selected from multiple training results as the final initial liquid level prediction model, thereby ensuring the reliability of the final model.

[0056] Example 6 Based on Example 1, this example provides a liquid level prediction method based on wavelet transform and machine learning. In step 3, the current liquid level data is collected in real time, and the current liquid level data is transmitted to the liquid level prediction model to identify the liquid level data, and the predicted value of the current liquid level is output, including: Obtain a non-contact sensor and connect the non-contact sensor to a data acquisition terminal; Based on the data acquisition terminal, the current liquid level data collected by the non-contact sensor is acquired in real time. At the same time, the current liquid level data is processed to obtain the standard liquid level data; The standard liquid level data is transmitted to the liquid level prediction model for analysis to determine the data features corresponding to the standard liquid level data; Matching the data features of the standard liquid level data with the reference features stored in the liquid level prediction model, and obtaining the target reference features corresponding to the standard liquid level data according to the matching results, and analyzing the target reference features based on the liquid level prediction model to output the first prediction value of the standard liquid level data corresponding to the next time segment; The reference features of the previous time segment of the standard liquid level data are retrieved in the liquid level prediction model, and the reference features of the previous time segment of the standard liquid level data and the target reference features corresponding to the standard liquid level data are analyzed simultaneously according to the liquid level prediction model, and a second prediction value of the standard liquid level data corresponding to the next time segment is output; The first predicted value is compared with the second predicted value, and the predicted value of the current liquid level is output according to the comparison result.

[0057] In this embodiment, if the first predicted value is equal to the second predicted value, the first predicted value or the second predicted value is used as the predicted value of the current liquid level; If the first predicted value is not equal to the second predicted value, a target mean of the first predicted value and the second predicted value is calculated, and the target mean is used as the predicted value of the current liquid level.

[0058] In this embodiment, the non-contact sensor is known in advance and is used to collect current liquid level data.

[0059] In this embodiment, the data processing may be preprocessing of the current liquid level data, including operations such as data cleaning, wherein the standard liquid level data is the result obtained after processing the current liquid level data.

[0060] In this embodiment, the data feature may be a specific value corresponding to the standard liquid level data.

[0061] In this embodiment, the reference features are known in advance, stored in the liquid level prediction model, and serve as a reference for liquid level prediction.

[0062] In this embodiment, the target reference feature may be a reference feature corresponding to the standard liquid level data and the reference feature stored in the liquid level prediction model.

[0063] In this embodiment, the first prediction value may be a change prediction value of the current standard liquid level data corresponding to the next time segment output after analysis based on the target reference feature.

[0064] In this embodiment, the second prediction value may be a result obtained by analyzing the reference features of the previous time segment of the standard liquid level data and the target reference features corresponding to the standard liquid level data based on the liquid level prediction model.

[0065] In this embodiment, if the first predicted value is equal to the second predicted value, the first predicted value or the second predicted value is used as the predicted value of the current liquid level; If the first predicted value is not equal to the second predicted value, a target mean of the first predicted value and the second predicted value is calculated, and the target mean is used as the predicted value of the current liquid level.

[0066] The working principle and beneficial effects of the above technical solution are: by obtaining a non-contact sensor and connecting the non-contact sensor to a data acquisition terminal, the current liquid level data can be effectively acquired in real time based on the data acquisition terminal; by processing the current liquid level data, the accuracy and effectiveness of the standard liquid level data can be effectively guaranteed; by transmitting the standard liquid level data to a liquid level prediction model for analysis, the data features corresponding to the standard liquid level data can be determined; by matching the data features of the standard liquid level data with the reference features stored in the liquid level prediction model, the target reference features corresponding to the standard liquid level data can be obtained; and based on the liquid level prediction model, the target reference features are analyzed and output as standard liquid level data. The first prediction value of the next time segment of the standard liquid level data corresponding to the liquid level data is determined by retrieving the reference features of the previous time segment of the standard liquid level data in the liquid level prediction model, and analyzing the reference features of the previous time segment of the standard liquid level data and the target reference features corresponding to the standard liquid level data according to the liquid level prediction model, so as to effectively output the second prediction value of the standard liquid level data corresponding to the next time segment; the first prediction value is compared with the second prediction value, and the prediction value of the current liquid level is output according to the comparison result, which is conducive to ensuring the objectivity of the prediction value of the current liquid level and avoiding data deviation, thereby realizing accurate and effective processing of the current liquid level data and ensuring the accuracy and reliability of the final prediction value.

[0067] Example 7 Based on Example 1, this example provides a liquid level prediction method based on wavelet transform and machine learning. In step 4, liquid level transformation detection is performed based on the liquid level prediction value of the current liquid level to generate a liquid level detection report, including: Read the predicted value of the current liquid level and the current liquid level data, and calculate the target liquid level difference between the predicted value of the current liquid level and the current liquid level data; Obtain the preset reference liquid level difference interval; Compare the preset reference liquid level difference interval with the target liquid level difference; When the target liquid level difference falls within the preset reference liquid level difference interval, a first liquid level detection report is generated; When the target liquid level difference does not belong to the preset reference liquid level difference interval, a second liquid level detection report is generated.

[0068] In this embodiment, the preset reference liquid level difference interval may be set in advance and used as a criterion for generating the first liquid level detection report or the second liquid level detection report.

[0069] In this embodiment, the first liquid level detection report includes: current liquid level data, a liquid level prediction value of the current liquid level, and a target liquid level difference value; The second liquid level detection report includes: current liquid level data, liquid level prediction value of the current liquid level, target liquid level difference and alarm signal data.

[0070] The working principle and beneficial effect of the above technical solution are: by reading the liquid level prediction value of the current liquid level and the current liquid level data, and calculating the target liquid level difference between the current liquid level prediction value and the current liquid level data, and comparing the preset reference liquid level difference interval with the target liquid level difference, the corresponding first liquid level detection report or the second liquid level detection report is effectively generated, thereby facilitating the user to grasp the current liquid level status in time and improving the timeliness of liquid level monitoring.

[0071] Example 8 This embodiment provides a liquid level prediction system based on wavelet transform and machine learning. Figure 3 As shown, including: The characteristic information determination module is used to collect historical liquid level data and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level signal; A model building module is used to use the characteristic information of the historical liquid level data as the input of the machine learning model, and train the machine learning model according to the characteristic information to obtain a liquid level prediction model; The prediction module is used to collect the current liquid level data in real time, transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; The report generation module is used to record the predicted value of each liquid level, perform liquid level change detection, and generate a liquid level detection report.

[0072] The working principle and beneficial effects of the above technical solution are: by collecting historical liquid level data, the historical liquid level data is subjected to wavelet transform to effectively obtain the characteristic information of the historical liquid level data, and by using the characteristic information of the historical liquid level data as the input of the machine learning model, the machine learning model can be trained through the characteristic information, thereby effectively ensuring the effectiveness and accuracy of the liquid level prediction model, and then accurately realizing the effective output of the predicted value of the current liquid level, thereby realizing liquid level change detection through the predicted value of the current liquid level, and then effectively generating a test report, which not only improves the accuracy and effectiveness of the liquid level prediction, but also effectively ensures the readability of the data by generating a test report and is conducive to grasping the current liquid level status as soon as possible, thereby improving the accuracy and real-time performance of the detection.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A liquid level prediction method based on wavelet transform and machine learning, characterized in that: include: Step 1: Collect historical liquid level data, and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level data; Step 2: Use the feature information of the historical liquid level data as the input of the machine learning model, and train the machine learning model based on the feature information to obtain a liquid level prediction model; Step 3: Collect the current liquid level data in real time, and transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; Step 4: Perform liquid level change detection based on the predicted value of the current liquid level and generate a liquid level detection report.

2. The liquid level prediction method based on wavelet transform and machine learning according to claim 1 is characterized in that: In step 1, historical liquid level data is collected and wavelet transformed to obtain characteristic information of the historical liquid level data, including: Acquire the collection requirements for collecting historical liquid level data, wherein the collection requirements include: data types of the historical liquid level data and the data collection volume corresponding to each data type; Generate a first traversal factor according to the data type of the historical liquid level data in the collection requirement; Generate a second traversal factor according to the data collection amount corresponding to the data type in the collection requirement; Inputting the first traversal factor into the preset database for a first traversal, and locating a data sequence consistent with the data type in the preset database according to the first traversal result; Perform a second traversal on the data sequence with the same data type in the preset database based on the second traversal factor, and collect historical liquid level data in the preset database according to the second traversal result; The historical liquid level data is subjected to wavelet transformation to obtain the characteristic information of the historical liquid level data.

3. The liquid level prediction method based on wavelet transform and machine learning according to claim 2 is characterized in that: Perform wavelet transformation on the historical liquid level data to obtain the characteristic information of the historical liquid level data, including: Perform wavelet transform on the collected historical liquid level data to obtain the frequency domain characteristic values ​​of the historical liquid level data; Obtain a frequency domain characteristic threshold, and compare the frequency domain characteristic value of the historical liquid level data with the frequency domain characteristic threshold; Based on the comparison results, the data frequency band corresponding to the frequency domain characteristic value of the historical liquid level data that is greater than the frequency domain characteristic threshold is collected and used as the key analysis section; Obtain analysis requirements, perform signal analysis on key analysis sections based on the analysis requirements, and output characteristic information of historical liquid level data corresponding to each analysis requirement.

4. The liquid level prediction method based on wavelet transform and machine learning according to claim 1 is characterized in that: In step 2, the feature information of the historical liquid level data is used as the input of the machine learning model, and the machine learning model is trained according to the feature information to obtain a liquid level prediction model, including: Acquire characteristic information of the historical liquid level data, and analyze the characteristic information based on a preset regression model to determine the subordination of different characteristic information to the liquid level; The feature information is screened based on the relative size relationship between the reference threshold and the degree of subordination to obtain key feature information, and the time domain features and frequency domain features of the key feature information are extracted; Generate index tags based on time domain features, and traverse the spatial information at the same time in the database according to the index tags based on the time and space features to obtain the variable factors corresponding to the key feature information; Based on the time domain features, the frequency domain features and the variable factors are bit-mapped, and an analysis data group is obtained based on the bit-mapped results; Repeatedly obtain the analysis data groups formed by the frequency domain characteristics and variable factors of the key feature information at different times, and bind the analysis data groups at different times to obtain the control data group; Based on the neural network, each analysis data group in the control data group is analyzed to obtain the nonlinear relationship between the frequency domain characteristics and variable factors at different times, and the nonlinear relationship of each analysis data group in the control group is summarized; Performing difference correction on the nonlinear relationship based on the summary result, and obtaining the target nonlinear relationship based on the difference correction result; Iteratively train the machine learning model based on the nonlinear relationship, and obtain an initial liquid level prediction model based on the iterative training results; A reference data layer is allocated in the initial liquid level prediction model, and the frequency domain characteristics of key feature information and the target nonlinear relationship between variable factors are packaged and stored in the reference data layer; The access rights of the reference data layer are adapted based on the storage results, and the liquid level prediction model is obtained based on the adaptation results.

5. The liquid level prediction method based on wavelet transform and machine learning according to claim 4 is characterized in that: The initial liquid level prediction model is obtained based on the iterative training results, including: Obtain a test liquid level data set, and obtain a machine learning model after each iterative training based on the number of iterative training; Testing the machine learning model after each iteration of training based on the test liquid level data set, and determining the model loss value of the machine learning model after each iteration of training based on the data difference between the test result and the standard result corresponding to the test liquid level data set; A target iterative training result is selected from the number of iterative training times based on the model loss value, and the target iterative training result is used as the initial liquid level prediction model.

6. The liquid level prediction method based on wavelet transform and machine learning according to claim 1 is characterized in that: In step 3, the current liquid level data is collected in real time, and the current liquid level data is transmitted to the liquid level prediction model to identify the liquid level data and output the predicted value of the current liquid level, including: Obtain a non-contact sensor and connect the non-contact sensor to a data acquisition terminal; Based on the data acquisition terminal, the current liquid level data collected by the non-contact sensor is acquired in real time. At the same time, the current liquid level data is processed to obtain the standard liquid level data; The standard liquid level data is transmitted to the liquid level prediction model for analysis to determine the data features corresponding to the standard liquid level data; Matching the data features of the standard liquid level data with the reference features stored in the liquid level prediction model, and obtaining the target reference features corresponding to the standard liquid level data according to the matching results, and analyzing the target reference features based on the liquid level prediction model to output the first prediction value of the standard liquid level data corresponding to the next time segment; The reference features of the previous time segment of the standard liquid level data are retrieved in the liquid level prediction model, and the reference features of the previous time segment of the standard liquid level data and the target reference features corresponding to the standard liquid level data are analyzed simultaneously according to the liquid level prediction model, and a second prediction value of the standard liquid level data corresponding to the next time segment is output; The first predicted value is compared with the second predicted value, and the predicted value of the current liquid level is output according to the comparison result.

7. The liquid level prediction method based on wavelet transform and machine learning according to claim 6 is characterized in that: The first predicted value is compared with the second predicted value, and the predicted value of the current liquid level is output according to the comparison result, including: If the first predicted value is equal to the second predicted value, the first predicted value or the second predicted value is used as the predicted value of the current liquid level; If the first predicted value is not equal to the second predicted value, a target mean of the first predicted value and the second predicted value is calculated, and the target mean is used as the predicted value of the current liquid level.

8. The liquid level prediction method based on wavelet transform and machine learning according to claim 1 is characterized in that: In step 4, a liquid level change detection is performed based on the liquid level prediction value of the current liquid level, and a liquid level detection report is generated, including: Read the predicted value of the current liquid level and the current liquid level data, and calculate the target liquid level difference between the predicted value of the current liquid level and the current liquid level data; Obtain the preset reference liquid level difference interval; Compare the preset reference liquid level difference interval with the target liquid level difference; When the target liquid level difference falls within the preset reference liquid level difference interval, a first liquid level detection report is generated; When the target liquid level difference does not belong to the preset reference liquid level difference interval, a second liquid level detection report is generated.

9. The liquid level prediction method based on wavelet transform and machine learning according to claim 8, characterized in that: The first liquid level detection report includes: current liquid level data, liquid level prediction value of the current liquid level and target liquid level difference; The second liquid level detection report includes: current liquid level data, liquid level prediction value of the current liquid level, target liquid level difference and alarm signal data.

10. A liquid level prediction system based on wavelet transform and machine learning, characterized in that: include: The characteristic information determination module is used to collect historical liquid level data and perform wavelet transformation on the historical liquid level data to obtain characteristic information of the historical liquid level signal; A model building module is used to use the characteristic information of the historical liquid level data as the input of the machine learning model, and train the machine learning model according to the characteristic information to obtain a liquid level prediction model; The prediction module is used to collect the current liquid level data in real time, transmit the current liquid level data to the liquid level prediction model to identify the liquid level data, and output the predicted value of the current liquid level; The report generation module is used to record the predicted value of each liquid level, perform liquid level change detection, and generate a liquid level detection report.

Images