Model construction method for water conservancy information analysis

Abstract

The invention relates to the technical field of data processing, in particular to a model construction method for water conservancy information analysis, and the method comprises the steps: obtaining region information corresponding to a target monitoring region and a plurality of water conservancy data; according to the difference data comparison value, water conservancy data is selected according to the number of similar monitoring parameters or parameter characteristic values; segmenting monitoring time periods corresponding to the selected water conservancy data based on the first division point and the second division point; according to a characteristic difference value between a water level characteristic value corresponding to the third divided section and a preset water level characteristic value, determining whether segmented optimization is carried out or not; and taking the segmented water conservancy data as training data to construct a target model. According to the invention, the accuracy of model risk prediction can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a model construction method for water conservancy information analysis. Background Technology

[0002] In water conservancy projects, when a sudden rise in water level or a surge in water flow occurs, it is often difficult to respond quickly and effectively, and to allow sufficient time to address potential risks. This undoubtedly increases safety hazards significantly. Although existing technologies have attempted to use models for prediction, the lack of a data segmentation mechanism that matches the dynamic characteristics of hydrological events during the model training phase results in inadequate data partitioning and selection methods, thus affecting prediction performance and limiting accuracy. Therefore, how to select and segment training data to improve the accuracy of risk prediction is a problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN117371337A discloses a method and system for constructing a water conservancy model based on digital twins. The method includes: collecting data from collection points located within a detection area and establishing a water conservancy digital twin model using a convolutional neural network model; performing data analysis on several sets of monitoring data obtained from the monitoring points to obtain a data quality set; generating stability coefficients from the data quality set; selecting target data; using the target data as input; obtaining corresponding prediction data using the trained water conservancy digital twin model; and generating corresponding error coefficients based on the degree of deviation between the prediction data and the actual data. If the error coefficients exceed an error threshold, a corresponding correction scheme is selected based on the parameter characteristics and algorithm characteristics of the abnormal module to correct the water conservancy twin model. However, the above method has the following problems: relying solely on generating stability coefficients from the data quality set to select target data fails to fully consider the diversity and complexity of data under different monitoring scenarios, and the lack of data segmentation leads to poor accuracy in model risk prediction. Summary of the Invention

[0004] To address this, the present invention provides a model construction method for water conservancy information analysis, which overcomes the problems of existing technologies that rely solely on generating stability coefficients from data quality sets to screen target data, failing to fully consider the diversity and complexity of data under different monitoring scenarios, and resulting in poor accuracy in model risk prediction due to the lack of data segmentation.

[0005] To achieve the above objectives, this invention provides a model construction method for water conservancy information analysis, comprising:

[0006] Acquire regional information and some water conservancy data corresponding to the target monitoring area;

[0007] Based on the comparison values ​​of the difference data, water conservancy data are selected according to the number of similar monitoring parameters or the characteristic values ​​of the parameters;

[0008] The first division point is determined based on the characteristic coefficient corresponding to the first time period, and the second division point is determined based on the weather parameter characteristic value of the first reference interval. Based on the first division point and the second division point, the monitoring time period corresponding to each selected water conservancy data is segmented.

[0009] Whether to perform segmentation optimization is determined based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value.

[0010] In the segmented optimization, the correlation combination is determined based on the correlation reference value, the correlation degree of parameters or the proportion of similar parameters, and the division point adjustment or feature segment selection is carried out according to the combination feature value corresponding to each correlation combination to segment the monitoring period corresponding to the water conservancy data.

[0011] The segmented water conservancy data will be used as training data to build the target model.

[0012] Furthermore, the method for confirming the difference data comparison values ​​includes:

[0013] Similar data combinations are determined based on information correlation, and the combination volatility is determined based on the standard deviation of the combination evaluation value corresponding to each similar data combination. The difference between the combination volatility and the preset combination volatility is then used to determine the difference data comparison value.

[0014] Furthermore, if the difference data comparison value is greater than or equal to the preset difference data comparison value, then water conservancy data are selected based on the number of similar monitoring parameters;

[0015] When selecting water conservancy data based on the number of similar monitoring parameters, select water conservancy data whose number of similar monitoring parameters corresponding to the regional information is greater than the preset number of similar monitoring parameters.

[0016] Furthermore, if the difference data comparison value is less than the preset difference data comparison value, then water conservancy data is selected based on the parameter characteristic value;

[0017] When selecting water conservancy data based on parameter characteristic values, data selection and analysis are performed for each parameter. When selecting and analyzing data for a single parameter, the water conservancy data are sorted in ascending order of the coefficient of variation corresponding to that parameter, and water conservancy data are selected according to the preset interval number corresponding to that parameter.

[0018] The number of preset intervals corresponding to a single parameter is determined based on the coefficient of variation comparison value corresponding to that parameter.

[0019] Furthermore, the first dividing point is determined based on the characteristic coefficients corresponding to the first time period, including:

[0020] Divide the monitoring period corresponding to the water conservancy data into three equal parts to obtain two equal division points, and then divide it into the first period, the second period, and the third period in order from morning to evening;

[0021] If the characteristic coefficient corresponding to the first time period is greater than or equal to the preset characteristic coefficient, then the earlier division point is recorded as the first division point.

[0022] If the characteristic coefficient corresponding to the first time period is less than the preset characteristic coefficient, then the first time period is increased based on the absolute value of the difference between the characteristic coefficient and the preset characteristic coefficient, and the last time point of the first time period after the increase adjustment is recorded as the first division point.

[0023] Furthermore, the second dividing point is determined based on the weather parameter characteristic values ​​of the first reference interval, including:

[0024] If the weather parameter characteristic value of the first reference interval is greater than or equal to the preset weather parameter characteristic value, then the later time division point is recorded as the second division point.

[0025] If the weather parameter characteristic value of the first reference interval is less than the preset weather parameter characteristic value, then the third time period is reduced based on the absolute value of the difference between the weather parameter characteristic value and the preset weather parameter characteristic value, and the earliest time point in the third time period after the reduction adjustment is recorded as the second division point.

[0026] The first reference interval is the time segment between the first division point and the later division point.

[0027] Furthermore, based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value, it is determined whether to perform segmentation optimization, including:

[0028] If the feature difference is greater than or equal to the preset feature difference, then no segmentation optimization is required;

[0029] If the feature difference is less than the preset feature difference, then segmented optimization is performed.

[0030] Furthermore, based on the correlation reference value, the correlation combination is determined according to the correlation degree of parameters or the proportion of similar parameters, including:

[0031] If the associated reference value is greater than or equal to the preset associated reference value, the associated combination is determined based on the parameter association degree.

[0032] If the associated reference value is less than the preset associated reference value, the associated combination is determined based on the proportion of similar parameters.

[0033] Furthermore, if the combination feature value corresponding to a single association combination is greater than or equal to the preset combination feature value, then the division point is adjusted for that association combination;

[0034] During the division point adjustment, adjustments are made to the first and third paragraph divisions based on the paragraph comparison values, and the two adjusted dividing points are used as the division points.

[0035] Furthermore, if the feature value corresponding to a single association combination is less than the preset feature value, then the feature segments for that association combination are selected in equal parts.

[0036] In the selection of feature paragraphs, the length of the feature paragraph is determined based on the comparison difference between the paragraph comparison value and the preset paragraph comparison value, and the feature paragraphs are selected based on the paragraph relevance. The trisection points corresponding to each feature paragraph are used as the division points.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows: In the technical solution of the present invention, the difference data comparison value effectively reflects the degree of difference between the evaluation value fluctuation level of each similar data combination and the historical effective fluctuation level. Then, based on the difference data comparison value, water conservancy data is selected adaptively according to the number of similar monitoring parameters or parameter characteristic values. This makes the selection of water conservancy data more in line with the actual application scenario, avoids the problem of insufficient data representativeness or disconnection from regional characteristics caused by a single selection standard, and thus improves the pertinence and effectiveness of the selected data. This provides a higher quality training data foundation for the subsequent construction of water conservancy information analysis models, thereby improving the accuracy and reliability of model analysis results.

[0038] Furthermore, in this invention, the monitoring time periods corresponding to the selected water conservancy data are segmented based on the first and second division points. This helps to improve the pertinence and accuracy of water conservancy data time period division, ensuring that the division results not only conform to the actual characteristics of monitoring parameters and weather parameters, but also flexibly respond to data patterns under different scenarios, providing a more reliable time period division basis for the subsequent construction of target models.

[0039] Furthermore, in this invention, the correlation reference value reflects the overall degree of correlation between selected water conservancy data. Then, based on the correlation reference value, the correlation combination is adaptively selected according to the parameter correlation degree or the proportion of similar parameters. This allows the division of correlation combinations to accurately capture the deep correlation between data when the overall correlation degree is high, and to build an effective correlation from the perspective of parameter commonality when the overall correlation degree is low, by using the proportion of similar parameters. Ultimately, the determination of correlation combinations is more in line with the actual correlation state of the data.

[0040] Furthermore, in this invention, the combined feature values ​​corresponding to each associated combination effectively reflect the overall characteristics of each segment within the associated combination and the degree of deviation from the reference standard. Then, based on the combined feature values, the division point adjustment or the selection of feature segments is determined, which is conducive to improving the accuracy and flexibility of water conservancy data division. This ensures that the division results can be optimized by fine-tuning when the features are stable, and the deviation can be corrected by reconstructing segments when the feature deviation is large, thus providing more reliable basic data support for the subsequent construction of the target model. Attached Figure Description

[0041] Figure 1 This is a module connection diagram of the hydrogen fuel cell vehicle safety early warning system supported by vehicle networking big data of the present invention;

[0042] Figure 2 This is a flowchart illustrating the process of determining the selection of water conservancy data based on the number of similar monitoring parameters or parameter characteristic values ​​according to the comparison value of difference data in this invention;

[0043] Figure 3 This is a flowchart illustrating the process of determining whether to perform segmentation optimization based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value in this invention.

[0044] Figure 4 This is a flowchart illustrating the process of adjusting the division points or selecting feature segments based on the combination feature values ​​corresponding to the associated combinations, as per the present invention. Detailed Implementation

[0045] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0046] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0047] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0048] Please see Figures 1 to 4 As shown, this invention provides a model construction method for water conservancy information analysis, comprising:

[0049] Acquire regional information and some water conservancy data corresponding to the target monitoring area;

[0050] Based on the comparison values ​​of the difference data, water conservancy data are selected according to the number of similar monitoring parameters or the characteristic values ​​of the parameters;

[0051] The first division point is determined based on the characteristic coefficient corresponding to the first time period, and the second division point is determined based on the weather parameter characteristic value of the first reference interval. Based on the first division point and the second division point, the monitoring time period corresponding to each selected water conservancy data is segmented.

[0052] Whether to perform segmentation optimization is determined based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value.

[0053] In the segmented optimization, the correlation combination is determined based on the correlation reference value, the correlation degree of parameters or the proportion of similar parameters, and the division point adjustment or feature segment selection is carried out according to the combination feature value corresponding to each correlation combination to segment the monitoring period corresponding to the water conservancy data.

[0054] The segmented water conservancy data will be used as training data to build the target model.

[0055] The application scenario of this invention is the selection and segmentation of training data when building a model for water conservancy information analysis; the target monitoring area is the catchment area of ​​a river or reservoir that needs to be predicted for water level.

[0056] The regional information consists of the monitoring parameters corresponding to each time point within a preset time period of the target monitoring area. The monitoring parameters include, but are not limited to, gate opening, average flow velocity, and average water level.

[0057] It is understood that each monitoring area in this invention is equipped with several monitoring points. The location and number of monitoring points can be set by the user. The average flow velocity and average water level correspond to the average flow velocity and average water level of each monitoring point at a single moment, respectively. The flow velocity and water level of each monitoring point are measured by a flow meter and a water level gauge, respectively. The gate opening is converted into an electrical signal by installing a rotary encoder or a wire displacement sensor on the motor or transmission mechanism of the gate opening and closing machine, and the gate opening is output in real time. This is a common technical means used by those skilled in the art, and will not be described in detail.

[0058] Individual water conservancy data includes parameters corresponding to each time point monitored in real time for a single monitoring area. The parameters include monitoring parameters and weather parameters. Weather parameters include, but are not limited to, rainfall, temperature, air pressure, and wind speed. Rainfall, temperature, air pressure, and wind speed are obtained by referencing meteorological station or forecast data, which will not be elaborated on in detail.

[0059] For a single water conservancy data or regional information, take the start time of the water conservancy data or regional information as the starting point, set an interval point every 1 hour, and record the starting point and each interval point as a time point;

[0060] The monitoring period corresponding to a single water conservancy data point is [the earliest time point corresponding to the water conservancy data point, and the latest time point corresponding to the water conservancy data point]. It should be noted that the length of the monitoring period corresponding to each water conservancy data point is the same. The length of the monitoring period corresponding to a single water conservancy data point is the time length between the earliest time point and the latest time point of the water conservancy data point.

[0061] This invention includes several historical records. Each historical record records at least one instance of the selection and segmentation of training data during the model construction for water conservancy information analysis, including the combined volatility, difference data comparison value, number of similar monitoring parameters, coefficient of variation comparison value, and number of intervals. Each historical record also has a corresponding qualification mark, which records whether the selection and segmentation of training data during the model construction for water conservancy information analysis meets the user's requirements. The qualification mark can be recorded manually. It is understood that the user can determine whether the selection and segmentation of training data during the model construction for water conservancy information analysis meets the requirements based on self-defined indicators. Self-defined indicators can be, but are not limited to, the number of errors, which will not be elaborated here. The number of errors is the cumulative number of times the target model makes incorrect predictions of water level.

[0062] When segmenting the monitoring periods corresponding to the selected water conservancy data based on the first and second division points, the data is divided into the first segment, the second segment, and the third segment in order from morning to evening.

[0063] Specifically, the methods for confirming the difference data comparison values ​​include:

[0064] Similar data combinations are determined based on information correlation, and the combination volatility is determined based on the standard deviation of the combination evaluation value corresponding to each similar data combination. The difference between the combination volatility and the preset combination volatility is then used to determine the difference data comparison value.

[0065] The process of determining similar data combinations based on information correlation includes: performing cluster analysis on each water conservancy data; when performing cluster analysis on a single water conservancy data, the water conservancy data is recorded as the target water conservancy data; other water conservancy data not recorded in the similar data combination are recorded as reference water conservancy data; reference water conservancy data with information correlation greater than the preset information correlation with the target water conservancy data and the target water conservancy data are recorded in a similar data combination; and cluster analysis continues on other water conservancy data not recorded in the similar data set until all water conservancy data are recorded in a similar data combination, at which point the cluster analysis stops.

[0066] The correlation between any two water conservancy data is the minimum value among the mean convergence of each monitoring parameter. The average value of the monitoring parameter corresponding to each time point of one water conservancy data is denoted as a1, and the average value of the monitoring parameter corresponding to each time point of another water conservancy data is denoted as a2. The mean convergence of a single monitoring parameter = |a1-a2| / the larger value between a1 and a2.

[0067] The user can determine the preset information correlation value according to the actual application scenario. The greater the user's need to improve the accuracy of the difference data comparison value judgment, the greater the preset information correlation value. One preset information correlation value is provided, with a preset information correlation of 70%.

[0068] The combined evaluation value corresponding to a single similar data combination is the average of the maximum water level mean values ​​corresponding to each water conservancy data in that similar data combination.

[0069] The maximum average water level corresponding to a single water conservancy data point is the maximum value among the average water levels at each time point in that water conservancy data point;

[0070] The portfolio volatility is the standard deviation of the portfolio evaluation values ​​corresponding to each similar data combination;

[0071] Difference data comparison value = Combined volatility - Preset combined volatility;

[0072] The detection method selects historical records of water conservancy data based on data feature values, and records the average value of the combined volatility corresponding to each historical record that meets the user's needs as the preset combined volatility.

[0073] Understandably, information correlation effectively reflects the similarity of different water conservancy data in parameter dimensions, and then similar data combinations are determined based on information correlation. The combined evaluation value corresponding to the similar data combination effectively reflects the average level of extreme water levels within the similar data combination. Then, the combined volatility is determined based on the standard deviation of the combined evaluation value corresponding to each similar data combination. The combined volatility effectively reflects the dispersion of extreme water level differences between individual similar data combinations. Then, the difference between the combined volatility and the preset combined volatility is used to determine the difference data comparison value. The difference data comparison value effectively reflects the difference between the extreme water level fluctuation of different similar data combinations and the historical effective fluctuation benchmark.

[0074] Specifically, if the difference data comparison value is greater than or equal to the preset difference data comparison value, then water conservancy data are selected based on the number of similar monitoring parameters;

[0075] When selecting water conservancy data based on the number of similar monitoring parameters, select water conservancy data whose number of similar monitoring parameters corresponding to the regional information is greater than the preset number of similar monitoring parameters.

[0076] Among them, the value of the preset difference data comparison value can be determined by the user according to the actual application scenario. The smaller the value of the preset difference data comparison value, the greater the user's need to select water conservancy data based on the number of similar monitoring parameters. A method for determining the value of the preset difference data comparison value is provided, which detects the historical records of users selecting water conservancy data based on the number of similar monitoring parameters, and records the average value of the difference data comparison value corresponding to the historical records that can meet the user's needs as the preset difference data comparison value.

[0077] The number of similar monitoring parameters corresponding to a single water conservancy data point and regional information is the total number of similar monitoring parameters corresponding to a single water conservancy data point and regional information.

[0078] For a single monitoring parameter, if the mean convergence of a single water conservancy data point and the corresponding regional information for that monitoring parameter is greater than the preset mean convergence, then that monitoring parameter is a similar monitoring parameter corresponding to the single water conservancy data point and the regional information.

[0079] It should be noted that the calculation method for the mean convergence of a single monitoring parameter corresponding to a single water conservancy data point and regional information is the same as the calculation method for the mean convergence of a single monitoring parameter corresponding to any two water conservancy data points, and will not be elaborated further.

[0080] The user can determine the value of the preset mean convergence degree according to the actual application scenario. The greater the user's need to improve the accuracy of similar monitoring parameter judgment, the greater the value of the preset mean convergence degree. A method for determining the preset mean convergence degree is provided, with the preset mean convergence degree being 78%.

[0081] The user can determine the value of the preset number of similar monitoring parameters according to the actual application scenario. The greater the user's need to improve the model training accuracy, the larger the value of the preset number of similar monitoring parameters. A method for setting the value of the preset number of similar monitoring parameters is provided. The method detects the historical records of water conservancy data selected by the user based on the number of similar monitoring parameters, and records the average value of the number of similar monitoring parameters corresponding to the historical records that can meet the user's needs as the preset number of similar monitoring parameters.

[0082] Specifically, if the difference data comparison value is less than the preset difference data comparison value, then water conservancy data is selected based on the parameter characteristic value;

[0083] When selecting water conservancy data based on parameter characteristic values, data selection and analysis are performed for each parameter. When selecting and analyzing data for a single parameter, the water conservancy data are sorted in ascending order of the coefficient of variation corresponding to that parameter, and water conservancy data are selected according to the preset interval number corresponding to that parameter.

[0084] The number of preset intervals corresponding to a single parameter is determined based on the coefficient of variation comparison value corresponding to that parameter.

[0085] For a single parameter, the coefficient of variation of that parameter in a single set of water conservancy data is equal to the standard deviation of the parameter value at each time point in the water conservancy data, divided by the average value of the parameter value at each time point in the water conservancy data.

[0086] The span value of the coefficient of variation for a single parameter is the absolute value of the difference between the maximum and minimum values ​​of the coefficient of variation for that parameter in each set of water conservancy data.

[0087] Let b be the range of the coefficient of variation for a single parameter, and let b0 be the average of the ranges of the coefficient of variation for all parameters; the coefficient of variation comparison value = b / b0;

[0088] The number of preset intervals corresponding to a single parameter is the smallest integer greater than or equal to w, where w = coefficient of variation comparison value / average of the coefficient of variation comparison values ​​corresponding to each parameter in each historical record that can meet the user's needs × (average of the number of preset intervals corresponding to each parameter in each historical record that can meet the user's needs).

[0089] The sequence of water conservancy data sorted according to the coefficient of variation corresponding to the parameter from smallest to largest is recorded as the reference sequence corresponding to the parameter.

[0090] When selecting water conservancy data according to the preset interval number corresponding to this parameter, the first water conservancy data in the reference sequence corresponding to this parameter is used as the first selected water conservancy data. Water conservancy data is selected according to the preset interval number corresponding to this parameter until the number of water conservancy data after the last selected water conservancy data is less than the preset interval number, then the selection of water conservancy data for this parameter is stopped.

[0091] The preset interval number corresponding to a single parameter is the number of water conservancy data points between two adjacent water conservancy data points selected in the reference sequence corresponding to that parameter.

[0092] It is understandable that when the difference data comparison value is greater than or equal to the preset difference data comparison value, it means that the fluctuation of the combined evaluation value of the current similar data combination exceeds the historical fluctuation level that meets the requirements, and the difference in evaluation value between the combinations is more significant. At this time, it is necessary to prioritize ensuring that the selected water conservancy data matches the core characteristics of the target area. Therefore, water conservancy data is selected according to the number of similar monitoring parameters.

[0093] When the difference data comparison value is less than the preset difference data comparison value, it indicates that the fluctuation of the combined evaluation value of each similar data combination is lower than the historical effective fluctuation level, and the difference of the evaluation value between each combination is small. At this time, more attention should be paid to the characteristic distribution of the data itself to ensure the representativeness and diversity of the selected data. Therefore, water conservancy data are selected according to the parameter characteristic value.

[0094] Specifically, the first dividing point is determined based on the characteristic coefficients corresponding to the first time period, including:

[0095] Divide the monitoring period corresponding to the water conservancy data into three equal parts to obtain two equal division points, and then divide it into the first period, the second period, and the third period in order from morning to evening;

[0096] If the characteristic coefficient corresponding to the first time period is greater than or equal to the preset characteristic coefficient, then the earlier division point is recorded as the first division point.

[0097] If the characteristic coefficient corresponding to the first time period is less than the preset characteristic coefficient, then the first time period is increased based on the absolute value of the difference between the characteristic coefficient and the preset characteristic coefficient, and the last time point of the first time period after the increase adjustment is recorded as the first division point.

[0098] Among them, the characteristic coefficient corresponding to the first time period is the average value of the sub-characteristic values ​​corresponding to each monitoring parameter in the first time period;

[0099] For a single monitoring parameter, the monitoring parameter is denoted as the target monitoring parameter, and the sub-feature value corresponding to the target monitoring parameter is the standard deviation of the reference value corresponding to the target monitoring parameter in each selected water conservancy data.

[0100] For a single water conservancy data, the average value of the target monitoring parameter corresponding to each time point in the first time period of the water conservancy data is recorded as the mean value of the target monitoring parameter corresponding to the water conservancy data, and the standard deviation of the target monitoring parameter corresponding to each time point in the first time period of the water conservancy data is recorded as the fluctuation value of the target monitoring parameter corresponding to the water conservancy data. The earlier equal division point is recorded as the first division point and the historical record that can meet the user's needs is recorded as the first reference historical record.

[0101] The reference value of the target monitoring parameter in a single water conservancy data point = the mean value of the target monitoring parameter corresponding to the water conservancy data point / the average value of the mean values ​​of the target monitoring parameters corresponding to each water conservancy data point in each first reference historical record + the fluctuation value of the target monitoring parameter corresponding to the water conservancy data point / the average value of the fluctuation value of the target monitoring parameter corresponding to each water conservancy data point in each first reference historical record;

[0102] The user can determine the value of the preset feature coefficient according to the actual application scenario. The smaller the value of the preset feature coefficient, the greater the user's need to mark the earlier division point as the first division point. A method for determining the value of the preset feature coefficient is provided to detect the historical records in which the user marks the earlier division point as the first division point, and record the historical records that can meet the user's needs as the preset feature coefficient.

[0103] When adjusting the increase for the first time period based on the absolute value of the difference between the characteristic coefficient and the preset characteristic coefficient, the increase value for the first time period is the smallest integer greater than or equal to w0, where w0 = |characteristic coefficient - preset characteristic coefficient| / preset characteristic coefficient × number of time points in the first time period × ratio coefficient, and the ratio coefficient is 1 / 4.

[0104] When increasing the adjustment for the first time period, the calculated number of time points (i.e., the increase value of the first time period) is added backward from the end of the first time period (i.e. the latest time point of the original first time period) to form the first time period after the increase adjustment.

[0105] It is understandable that the characteristic coefficients corresponding to the first time period can effectively reflect the overall characteristic performance of the monitoring parameters within the first time period. When the characteristic coefficients corresponding to the first time period are greater than or equal to the preset characteristic coefficients, it indicates that the characteristic performance of the first time period has reached the expected standard, and the monitoring data contained therein can better reflect the typical characteristics of the time period. There is no need to adjust the time period range. The earlier division point is recorded as the first division point.

[0106] When the characteristic coefficient corresponding to the first time period is less than the preset characteristic coefficient, it indicates that the characteristic performance of the first time period has not met the expected standard. The existing time period range has failed to fully reflect the monitoring characteristics that the time period should have and is not representative enough. It is necessary to expand the time period range to enhance the integrity of its characteristics. The last time point of the first time period after the adjustment is recorded as the first division point.

[0107] Specifically, the second division point is determined based on the weather parameter characteristic values ​​of the first reference interval, including:

[0108] If the weather parameter characteristic value of the first reference interval is greater than or equal to the preset weather parameter characteristic value, then the later time division point is recorded as the second division point.

[0109] If the weather parameter characteristic value of the first reference interval is less than the preset weather parameter characteristic value, then the third time period is reduced based on the absolute value of the difference between the weather parameter characteristic value and the preset weather parameter characteristic value, and the earliest time point in the third time period after the reduction adjustment is recorded as the second division point.

[0110] The first reference interval is the time segment between the first division point and the later division point.

[0111] Among them, the weather parameter characteristic value of the first reference interval is the average value of the weather characteristic value corresponding to each weather parameter in the first reference interval;

[0112] For a single weather parameter, the weather parameter is denoted as the target weather parameter, and the sub-weather feature value corresponding to the target weather parameter is the standard deviation of the weather reference value corresponding to the target weather parameter in each selected water conservancy data.

[0113] For a single water conservancy data, the average value of the target weather parameter corresponding to each time point in the first reference interval of the water conservancy data is recorded as the mean value of the target weather parameter corresponding to the water conservancy data, and the standard deviation of the target weather parameter corresponding to each time point in the first reference interval of the water conservancy data is recorded as the characteristic value of the target weather parameter corresponding to the water conservancy data. The historical records that can meet the user's needs and are recorded as the second reference historical records are detected and recorded as the second reference historical records.

[0114] The weather reference value corresponding to the target weather parameter in a single water conservancy data point = the mean value of the target weather parameter corresponding to the water conservancy data point / the average value of the mean values ​​of the target weather parameters corresponding to each water conservancy data point in each second reference historical record + the characteristic value of the target weather parameter corresponding to the water conservancy data point / the average value of the characteristic values ​​of the target weather parameters corresponding to each water conservancy data point in each second reference historical record;

[0115] The user can determine the value of the preset weather parameter feature value according to the actual application scenario. The smaller the value of the preset weather parameter feature value, the greater the user's demand to mark the later time division point as the second division point. A method for determining the value of the preset weather parameter feature value is provided, which detects the historical records in which the user marks the later time division point as the second division point, and records the average value of the weather parameter feature value corresponding to the historical records that can meet the user's needs as the preset weather parameter feature value.

[0116] The decrease value for the third period is the smallest integer greater than or equal to w1, where w1 = |weather parameter characteristic value - preset weather parameter characteristic value| / preset weather parameter characteristic value × number of time points in the first reference interval × ratio coefficient, and the ratio coefficient is 1 / 4;

[0117] When reducing the third time period, the calculated number of time points (i.e., the reduction value of the third time period) is shortened from the beginning of the third time period (i.e., the earliest time point of the original third time period) to form the shortened third time period.

[0118] It is understandable that the weather parameter characteristic values ​​of the first reference interval can effectively reflect the overall characteristics of the weather parameters within the first reference interval. When the weather parameter characteristic values ​​of the first reference interval are greater than or equal to the preset weather parameter characteristic values, it indicates that the weather parameter characteristics of the interval have reached the expected standard, and the weather data contained therein can better support the original setting of the second division point. There is no need to adjust the range of the third time period, and the later division points are recorded as the second division points.

[0119] When the weather parameter characteristic value of the first reference interval is less than the preset weather parameter characteristic value, it indicates that the weather parameter characteristic of the interval has not met the expected standard. The existing third time period range has not fully reflected the proper weather characteristic correlation. It is necessary to enhance the specificity of the characteristics by shortening the time period. The earliest time point in the adjusted third time period is recorded as the second division point.

[0120] Specifically, the decision to perform segmentation optimization is based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value, including:

[0121] If the feature difference is greater than or equal to the preset feature difference, then no segmentation optimization is required;

[0122] If the feature difference is less than the preset feature difference, then segmented optimization is performed.

[0123] Wherein, the feature difference = the water level feature value corresponding to the third segment - the preset water level feature value;

[0124] The water level characteristic value corresponding to the third division segment is the standard deviation of the interval water level value corresponding to each selected water conservancy data, and the interval water level value corresponding to a single water conservancy data is the maximum value of the average water level value at each time point in the third division segment of that water conservancy data.

[0125] The preset water level characteristic value is the average value of the water level characteristic values ​​corresponding to each historical record that does not require segmentation optimization and can meet the user's needs.

[0126] The preset feature difference value can be determined by the user based on the actual application scenario. The greater the user's need to improve training accuracy, the smaller the preset feature difference value should be. A method for setting the preset feature difference value is provided. Historical records that do not require segmentation optimization are detected, and the average value of the feature difference corresponding to the historical records that can meet the user's needs is recorded as the preset feature difference value.

[0127] It is understandable that the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value can effectively reflect the difference between the water level feature of the third segment and the preset value. When the feature difference is less than the preset feature difference, it means that the gap between the water level feature of the current third segment and the expected standard has not reached an acceptable range, and the water level data contained therein is insufficient to support the training accuracy. The existing segmentation method is difficult to meet the user's accuracy requirements, so segmentation optimization is required.

[0128] Specifically, the association combination is determined based on the correlation reference value, the correlation degree of parameters, or the proportion of similar parameters, including:

[0129] If the associated reference value is greater than or equal to the preset associated reference value, the associated combination is determined based on the parameter association degree.

[0130] If the associated reference value is less than the preset associated reference value, the associated combination is determined based on the proportion of similar parameters.

[0131] Among them, each selected water conservancy data is denoted as the selected water conservancy data, and the associated reference value is the average number of associated water conservancy data corresponding to each selected water conservancy data;

[0132] The associated water conservancy data corresponding to a single selected water conservancy data point is the total number of selected water conservancy data points whose parameter correlation degree with the selected water conservancy data point is greater than the preset parameter correlation degree.

[0133] The correlation degree between any two selected water conservancy data points is the minimum value among the correlation coefficients of each parameter in the two selected water conservancy data points.

[0134] For a single parameter, the formula for calculating the correlation coefficient r corresponding to that parameter in two selected hydraulic data is:

[0135]

[0136] Where n is the number of time points in a single selected water resources dataset; xi and yi are the values ​​of this parameter corresponding to the i-th time point in two selected water resources datasets, respectively. This represents the average value of the parameter at each time point in the selected water conservancy data corresponding to xi. Let yi be the average value of the parameter at each time point in the selected water conservancy data, i = 1, 2, 3, ..., n;

[0137] The value of the preset association reference value can be determined by the user according to the actual application scenario. The larger the value of the preset association reference value, the greater the user's need to determine the association combination based on the proportion of similar parameters. A method for determining the value of the preset association reference value is provided, which detects the historical records of users determining the association combination based on the proportion of similar parameters, and records the average value of the association reference value corresponding to the historical records that can meet the user's needs as the preset association reference value.

[0138] When determining the association combination based on the correlation degree of parameters or the proportion of similar parameters, association analysis is performed on each selected water conservancy data. When performing association analysis on a single selected water conservancy data, the selected water conservancy data is recorded as the target selected water conservancy data, and the other selected water conservancy data besides the target selected water conservancy data is recorded as the reference selected water conservancy data. The reference selected water conservancy data that meet the preset conditions and the target selected water conservancy data are recorded as an association combination. The association analysis continues to be performed on the selected water conservancy data that are not recorded in the association combination until all selected water conservancy data are recorded in the association combination, and then the association analysis stops.

[0139] When determining the association combination based on the parameter correlation degree, the preset condition is that the parameter correlation degree with the target selected water conservancy data is greater than the preset parameter correlation degree;

[0140] When determining the associated combination based on the proportion of similar parameters, the preset condition is that the proportion of similar parameters with the target selected water conservancy data is greater than the preset proportion of similar parameters.

[0141] The percentage of similar parameters corresponding to any two selected water conservancy data points = the number of similar parameters corresponding to the two selected water conservancy data points / the total number of parameters corresponding to the two selected water conservancy data points;

[0142] For any single parameter in two selected water conservancy data sets, if the correlation coefficient corresponding to the parameter is greater than the preset correlation coefficient, then the parameter is a similar parameter corresponding to the two selected water conservancy data sets.

[0143] The value of the preset correlation coefficient can be determined by the user according to the actual application scenario. The greater the user's need to improve the accuracy of similarity parameter judgment, the larger the value of the preset correlation coefficient. One preset correlation coefficient value is provided, which is 0.8.

[0144] Users can determine the values ​​of preset parameter correlation degree and preset similar parameter quantity ratio according to the actual application scenario. The greater the user's need to improve the similarity of data in the associated combination, the greater the values ​​of preset parameter correlation degree and preset similar parameter quantity ratio. One preset parameter correlation degree is 0.6 and the preset similar parameter quantity ratio is 75%.

[0145] It is understandable that the correlation reference value can effectively reflect the overall correlation between the selected water conservancy data. When the correlation reference value is greater than or equal to the preset correlation reference value, it means that each selected water conservancy data has a lot of related water conservancy data on average, and the overall correlation is high. At this time, the correlation combination can be determined by the parameter correlation degree, which can more accurately capture the deep correlation between the data.

[0146] When the correlation reference value is less than the preset correlation reference value, it means that there are fewer correlation data for each selected water conservancy data on average, and the overall correlation degree is low. At this time, we should switch to determining the correlation combination based on the proportion of similar parameters. This can make up for the problem of insufficient overall correlation by using the common characteristics at the parameter level, which is more in line with the actual correlation status of the data.

[0147] Specifically, if the combination feature value corresponding to a single association combination is greater than or equal to the preset combination feature value, then the division point is adjusted for that association combination.

[0148] During the division point adjustment, adjustments are made to the first and third paragraph divisions based on the paragraph comparison values, and the two adjusted dividing points are used as the division points.

[0149] Among them, historical records that can meet user needs and adjust the division points for related combinations will be recorded as reference records;

[0150] The comparison value of the first segment is equal to (the monitoring parameter characterization value - the average value of the monitoring parameter characterization value of each associated combination in each reference record) / the average value of the monitoring parameter characterization value of each associated combination in each reference record; the monitoring parameter characterization value of a single associated combination is the average value of the fluctuation of the mean value of each monitoring parameter in the first segment of that associated combination.

[0151] The comparison value of the second segment is equal to (weather parameter characterization value - average weather parameter characterization value of each associated combination in each reference record) / average weather parameter characterization value of each associated combination in each reference record; the weather parameter characterization value of a single associated combination is the average value of the mean fluctuation of the parameters corresponding to each weather parameter in the second segment of that associated combination.

[0152] The mean fluctuation of a single parameter in the associated combination is the standard deviation of the mean of the parameter corresponding to each selected water conservancy data in the associated combination; the mean of the parameter corresponding to a single selected water conservancy data is the average value of the parameter at each time point in a single segment of the selected water conservancy data.

[0153] The comparison value of the third segment is equal to (water level characterization value - average water level characterization value of each associated combination in each reference record) / average water level characterization value of each associated combination in each reference record; the water level characterization value is the standard deviation of the parameter mean of the third segment of the selected water conservancy data in the associated combination when the parameter is the water level mean.

[0154] The combined feature value is the average of the paragraph comparison values ​​corresponding to each segment.

[0155] The user can determine the value of the preset combination feature value according to the actual application scenario. The smaller the value of the preset combination feature value, the greater the user's need to adjust the division point. A method for setting the value of the preset combination feature value is provided, which detects the user's historical records of division point adjustments and records the average value of the combination feature values ​​corresponding to each related combination in the historical records that can meet the user's needs as the preset combination feature value.

[0156] It should be noted that for the first or third segment division, if the paragraph comparison value corresponding to a single segment division is greater than the average of the paragraph comparison values ​​corresponding to all segment divisions, then the adjustment should be reduced for that segment division.

[0157] If the paragraph comparison value corresponding to a single segment is less than the average of the paragraph comparison values ​​corresponding to all segments, then the adjustment is increased for that segment.

[0158] If the paragraph comparison value corresponding to a single segmentation stage is equal to the average of the paragraph comparison values ​​corresponding to all segmentation stages, then no adjustment is needed for that segmentation stage.

[0159] Adjustments are made for the first and third paragraph divisions. When adjusting a single paragraph division, the adjustment amount for a single division stage is the smallest integer greater than or equal to w2, where w2 = |paragraph comparison value corresponding to this division stage - average value of paragraph comparison values ​​corresponding to each division stage| / average value of paragraph comparison values ​​corresponding to each division stage × average length of each division stage × ratio coefficient, where the ratio coefficient is 1 / 4.

[0160] The latest time point in the segment after adjustment for the first segment and the earliest time point in the segment after adjustment for the third segment are recorded as the dividing points. The monitoring period corresponding to the water conservancy data is segmented based on the two dividing points.

[0161] Specifically, if the feature value of a single associated combination is less than the preset feature value, then the feature segments of that associated combination are selected in equal parts.

[0162] In the selection of feature paragraphs, the length of the feature paragraph is determined based on the comparison difference between the paragraph comparison value and the preset paragraph comparison value, and the feature paragraphs are selected based on the paragraph relevance. The trisection points corresponding to each feature paragraph are used as the division points.

[0163] Wherein, the comparison difference = |combined feature value - preset combined feature value|;

[0164] Detect historical records that are selected by dividing them into characteristic segments and can meet the user's needs, and record them as valid historical records.

[0165] The feature paragraph length is the smallest integer greater than or equal to w3, where w3 = the alignment difference corresponding to a single association combination / the average of the alignment differences corresponding to each association combination in each valid historical record × the average of the feature paragraph lengths corresponding to each association combination in each valid historical record.

[0166] The length of a feature paragraph is the number of time points within a single time paragraph.

[0167] When selecting feature segments for a single water conservancy data set, first randomly select any segment of water conservancy data with the length of the feature segment;

[0168] When selecting other water conservancy data, the segment with the lowest correlation to the selected feature segment is chosen as the feature segment;

[0169] The paragraph correlation degree between a single segment and the selected feature segments is the average of the parameter correlation degrees between the segment and each selected feature segment. It should be noted that the calculation method of the parameter correlation degree between the segment and the selected feature segments is the same as the calculation method of the parameter correlation degree between the two water conservancy data. The only difference is that the number of time points included in the two data is different. This is something that is easy for those skilled in the art to understand, and will not be elaborated on in detail.

[0170] For a single water conservancy data point, the data is split and analyzed in chronological order from earliest to latest. When splitting and analyzing a single time point, a data segment with that time point as the starting point and a length equal to the characteristic segment length is recorded as a split segment. The splitting and analysis stops when the number of time points that have not been split and analyzed is less than the characteristic segment length.

[0171] Understandably, the combined feature values ​​intuitively quantify the degree of deviation between the overall characteristics of each segment in the associated combination and the reference standard;

[0172] When the combined feature value of a single association combination is greater than or equal to the preset combined feature value, it means that the overall feature of each segment of the association combination is within an acceptable range from the reference standard, and the feature performance of the existing segment has a basis for further optimization. At this time, by adjusting the division points in a targeted manner, the division can be made to fit the data details better while retaining the existing segment framework. Therefore, the division points of the association combination are adjusted.

[0173] When the combined feature value is less than the preset combined feature value, it indicates that the overall features of each segment deviate significantly from the reference standard. The existing features of the segments are no longer sufficient to support effective adjustment of the segmentation points. Continuing to use the original segment framework may lead to the accumulation of deviations. At this time, it is necessary to select feature segments equally and re-examine more representative feature segments to correct the deviations. The segmentation basis should be reconstructed at the segment level to improve the feature matching degree of the associated combination.

[0174] The three segments of water conservancy data are divided into three sections in chronological order from earliest to latest: the first section, the second section, and the third section.

[0175] The steps for using segmented water conservancy data as training data to build a target model include constructing a training sample set, feature extraction, dataset partitioning, and setting hyperparameters. When constructing the training sample set, based on the three segments of data, the training sample set is constructed according to an "input-output" logic: the input for a single sample is the monitoring parameter sequence of the first segment (including gate opening, average flow velocity, and average water level, etc.) and the weather parameter sequence of the second segment (including rainfall, temperature, air pressure, and wind speed, etc.), and the output is the maximum average water level of the third segment. During feature extraction, statistical features of the first and second segments are extracted. The statistical features corresponding to the first segment include, but are not limited to, the peak value (such as maximum flow velocity) and standard deviation of each parameter. (Reflecting fluctuation stability) and cumulative values ​​(such as total rainfall); the statistical features corresponding to the second paragraph include, but are not limited to, the rate of change of weather parameters (such as hourly temperature changes) and the duration of extreme values ​​(such as the number of hours of heavy rain ≥50mm / h). Principal component analysis or factor analysis is used to compress high-dimensional features into low-dimensional principal components, reducing redundancy and improving training efficiency; when dividing the dataset, it is divided into training set (70%), validation set (10%), and test set (20%) in a 7:1:2 ratio. Users can set hyperparameters according to their actual needs. Hyperparameters include, but are not limited to, vector dimension, learning rate, and number of iterations. These are all common techniques used by those skilled in the art, and will not be elaborated on in detail.

[0176] It should be noted that the length of the first and second paragraphs entered by the user is not fixed and can be flexibly set according to the actual analysis needs. At the same time, the weather monitoring parameters for the second paragraph can be obtained by the user from the weather forecast.

[0177] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A model construction method for water conservancy information analysis, characterized in that, include: Acquire regional information and some water conservancy data corresponding to the target monitoring area; Based on the comparison values ​​of the difference data, water conservancy data are selected according to the number of similar monitoring parameters or the characteristic values ​​of the parameters; The first division point is determined based on the characteristic coefficient corresponding to the first time period, and the second division point is determined based on the weather parameter characteristic value of the first reference interval. Based on the first division point and the second division point, the monitoring time period corresponding to each selected water conservancy data is segmented. Whether to perform segmentation optimization is determined based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value. In the segmented optimization, the correlation combination is determined based on the correlation reference value, the correlation degree of parameters or the proportion of similar parameters, and the division point adjustment or feature segment selection is carried out according to the combination feature value corresponding to each correlation combination to segment the monitoring period corresponding to the water conservancy data. The segmented water conservancy data will be used as training data to build the target model.

2. The model construction method for water conservancy information analysis according to claim 1, characterized in that, The methods for confirming the difference data comparison values ​​include: Similar data combinations are determined based on information correlation, and the combination volatility is determined based on the standard deviation of the combination evaluation value corresponding to each similar data combination. The difference between the combination volatility and the preset combination volatility is then used to determine the difference data comparison value.

3. The model construction method for water conservancy information analysis according to claim 2, characterized in that, If the difference data comparison value is greater than or equal to the preset difference data comparison value, then water conservancy data are selected according to the number of similar monitoring parameters; When selecting water conservancy data based on the number of similar monitoring parameters, select water conservancy data whose number of similar monitoring parameters corresponding to the regional information is greater than the preset number of similar monitoring parameters.

4. The model construction method for water conservancy information analysis according to claim 3, characterized in that, If the difference data comparison value is less than the preset difference data comparison value, then water conservancy data is selected based on the parameter characteristic value; When selecting water conservancy data based on parameter characteristic values, data selection and analysis are performed for each parameter. When selecting and analyzing data for a single parameter, the water conservancy data are sorted in ascending order of the coefficient of variation corresponding to that parameter, and water conservancy data are selected according to the preset interval number corresponding to that parameter. The number of preset intervals corresponding to a single parameter is determined based on the coefficient of variation comparison value corresponding to that parameter.

5. The model construction method for water conservancy information analysis according to claim 1, characterized in that, The first dividing point is determined based on the characteristic coefficients corresponding to the first time period, including: Divide the monitoring period corresponding to the water conservancy data into three equal parts to obtain two equal division points, and then divide it into the first period, the second period, and the third period in order from morning to evening; If the characteristic coefficient corresponding to the first time period is greater than or equal to the preset characteristic coefficient, then the earlier division point is recorded as the first division point. If the characteristic coefficient corresponding to the first time period is less than the preset characteristic coefficient, then the first time period is increased based on the absolute value of the difference between the characteristic coefficient and the preset characteristic coefficient, and the last time point of the first time period after the increase adjustment is recorded as the first division point.

6. The model construction method for water conservancy information analysis according to claim 5, characterized in that, The second division point is determined based on the weather parameter characteristic values ​​of the first reference interval, including: If the weather parameter characteristic value of the first reference interval is greater than or equal to the preset weather parameter characteristic value, then the later time division point is recorded as the second division point. If the weather parameter characteristic value of the first reference interval is less than the preset weather parameter characteristic value, then the third time period is reduced based on the absolute value of the difference between the weather parameter characteristic value and the preset weather parameter characteristic value, and the earliest time point in the third time period after the reduction adjustment is recorded as the second division point. The first reference interval is the time segment between the first division point and the later division point.

7. The model construction method for water conservancy information analysis according to claim 6, characterized in that, Whether to perform segmentation optimization is determined based on the feature difference between the water level feature value corresponding to the third segment and the preset water level feature value, including: If the feature difference is greater than or equal to the preset feature difference, then no segmentation optimization is required; If the feature difference is less than the preset feature difference, then segmented optimization is performed.

8. The model construction method for water conservancy information analysis according to claim 7, characterized in that, The association combination is determined based on the correlation reference value, the correlation degree of parameters, or the proportion of similar parameters, including: If the associated reference value is greater than or equal to the preset associated reference value, the associated combination is determined based on the parameter association degree. If the associated reference value is less than the preset associated reference value, the associated combination is determined based on the proportion of similar parameters.

9. The model construction method for water conservancy information analysis according to claim 8, characterized in that, If the combined feature value corresponding to a single associated combination is greater than or equal to the preset combined feature value, then the division point is adjusted for that associated combination; During the division point adjustment, adjustments are made to the first and third paragraph divisions based on the paragraph comparison values, and the two adjusted dividing points are used as the division points.

10. The model construction method for water conservancy information analysis according to claim 8, characterized in that, If the feature value of a single associated combination is less than the preset feature value, then the feature segments of that associated combination are selected in equal parts. In the selection of feature paragraphs, the length of the feature paragraph is determined based on the comparison difference between the paragraph comparison value and the preset paragraph comparison value, and the feature paragraphs are selected based on the paragraph relevance. The trisection points corresponding to each feature paragraph are used as the division points.

Images