Intelligent control method of sluice gate based on water conservancy monitoring data

Abstract

The present invention belongs to the field of intelligent control technology, and specifically relates to an intelligent control method for a sluice based on water conservancy monitoring data, the method comprising: collecting monitoring data and water surface video of a water storage body where the sluice is located; estimating flow data of multiple locations to be evaluated based on flow data of multiple monitoring locations in the monitoring data, and inputting the flow data into a water level prediction model together with the monitoring data, and adjusting the sluice opening by the obtained water level prediction value; wherein, estimating flow data of multiple locations to be evaluated comprises: selecting multiple principal components from all components according to stability and discrimination, dividing all principal components into multiple high-frequency components and multiple low-frequency components according to frequency, and determining the estimated value of flow data of the location to be evaluated by the convergence of the location to be evaluated and the adjacent monitoring locations on the low-frequency components, and the convergence of the location to be evaluated and the remaining monitoring locations on the high-frequency components. The present invention improves the accuracy of intelligent control.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and more specifically, to a sluice intelligent control method based on water conservancy monitoring data. Background Art

[0002] Sluice gates are low-head hydraulic structures built on the banks of rivers, canals, reservoirs and lakes with the functions of retaining and discharging water. Sluice gates are an indispensable and important part of water conservancy projects, playing a key role in flood control and drainage, water volume scheduling, irrigation and water supply.

[0003] As an important part of water conservancy projects, the safe operation of sluice gates is directly related to the comprehensive benefits of water conservancy projects and social and economic development. Traditional sluice gate monitoring and control methods have problems such as long manual inspection cycles and low detection efficiency, which can hardly meet the needs of modern water conservancy project management. Therefore, in modern water conservancy project management, the intelligent transformation of sluice gate control systems has become a key trend in the development of the industry.

[0004] In recent years, with the rapid development of sensor technology, communication technology, data processing technology and artificial intelligence technology, the research and development and application of sluice intelligent control systems based on sensor technology, communication technology, data processing technology and artificial intelligence technology have entered a rapid development stage.

[0005] The setting and number of installed sensors are limited by factors such as cost, technical complexity and site environment, resulting in insufficient monitoring data to fully understand the condition of the water area; for example, in large reservoirs or complex river systems, sensors are only set at key locations, and the number of installed sensors is limited. Although local water level data can be obtained, it is difficult to reflect the detailed conditions of the entire water area, which directly affects the accuracy of intelligent control of sluices. Summary of the invention

[0006] In order to solve the technical problem that the setting and installation quantity of the above-mentioned sensors are limited by factors such as cost, technical complexity and on-site environment, resulting in insufficient monitoring data to fully understand the status of the water area, the present invention provides an intelligent control method for a sluice based on water conservancy monitoring data, including: collecting monitoring data and water surface video of the water storage body where the sluice is located; estimating the flow data of multiple locations to be evaluated based on the flow data of multiple monitoring locations in the monitoring data, and inputting them into the water level prediction model together with the monitoring data, and adjusting the sluice opening according to the obtained water level prediction value; estimating the flow data of multiple locations to be evaluated, including: forming a main sequence of the monitoring location / location to be evaluated with the grayscale values ​​of the pixel points corresponding to the monitoring location / location to be evaluated in all video frames; obtaining a pixel point corresponding to the monitoring location in each video frame as the center The grayscale values ​​of the pixels at the same position in the same monitoring area of ​​all video frames are used to form a subordinate sequence of the monitoring position; multiple components of the main sequence and the subordinate sequence are obtained by decomposition; the stability of the component is calculated by the distribution of the main sequence of the monitoring position and all the subordinate sequences on the same component; the distinguishability of the component is calculated by the distribution of the main sequences of different monitoring positions on the same component; multiple main components are selected from all components according to the stability and distinguishability; all the main components are divided into multiple high-frequency components and multiple low-frequency components according to the frequency; for any position to be evaluated, the estimated value of the flow data of the position to be evaluated is determined by the convergence of the position to be evaluated and the adjacent monitoring positions on the low-frequency component, and the convergence of the position to be evaluated and the remaining monitoring positions on the high-frequency component.

[0007] The present invention acquires multi-dimensional data and supplements the visual information of the monitoring data through water surface videos, which helps to understand the changes in the water surface more intuitively. By setting multiple locations to be evaluated between adjacent monitoring locations and obtaining an estimated value of the flow data of the location to be evaluated based on the flow data of multiple monitoring locations in the monitoring data, the problem of a limited number of sensors can be solved. At the same time, the estimated value of the flow data of the location to be evaluated enriches the flow information, realizes flow monitoring of more locations, and improves the comprehensiveness of the system and the accuracy of intelligent control.

[0008] Preferably, the method for setting the position to be evaluated includes: for two adjacent monitoring positions and , , is the number of all monitoring locations; and The distance between ; At two adjacent monitoring locations and There are 4 positions to be evaluated between them. Therefore, the 4 positions to be evaluated are , , , .

[0009] Preferably, the stability of the calculation component includes: ; In the formula, For the The stability of the components, For the The first of all subordinate sequences of the monitoring position The first The mean of the values, For the The first of the main sequence of monitoring positions The first numerical values, To take the absolute value function, is the number of all monitoring locations, is the length of a component, is a natural exponential function.

[0010] In the present invention, the flow data at each location in the local monitoring area are similar, and the components of the subordinate sequences should maintain similar distribution characteristics to the components of the main sequence at the corresponding monitoring location. Based on this characteristic, the distribution of the main sequence of the monitoring location and all the subordinate sequences on the same component is analyzed, and the stability of the distribution of the main sequence of the monitoring location and all the subordinate sequences on the same component is calculated, which provides data support for the subsequent selection of the main component that meets the requirements from all components.

[0011] Preferably, calculating the distinguishability of the components includes: ; In the formula, For the The distinguishability of the components, is the first of the main sequences for all monitoring locations The first The variance of the values, is the first of the main sequences for all monitoring locations The first The mean of the values, is the length of a component.

[0012] In the present invention, there are differences between the flow data at different monitoring locations, and the distribution characteristics of the same component of the main sequences at different monitoring locations are relatively discrete. Based on this characteristic, the distribution of the main sequences of all monitoring locations on the same component is analyzed, and the distinctiveness of the distribution of the main sequences at different monitoring locations on the same component is calculated, which provides data support for the subsequent selection of the main component that meets the requirements from all components.

[0013] Preferably, the method of selecting multiple principal components from all components based on stability and discriminability includes: calculating the mean of the stability of all components and using it as the first threshold; calculating the mean of the discriminability of all components and using it as the second threshold; and taking the components whose stability is greater than the first threshold and whose discriminability is greater than the second threshold as principal components.

[0014] Preferably, the method of determining the estimated value of the flow data at the position to be evaluated includes: calculating the convergence of the position to be evaluated and each monitoring position on each component according to a convergence calculation formula; calculating the trend item of the flow data at the position to be evaluated by summing the convergence of the position to be evaluated and its corresponding adjacent monitoring positions on all low-frequency components; calculating the fluctuation item of the flow data at the position to be evaluated by summing the convergence of the position to be evaluated and its corresponding remaining monitoring positions on all high-frequency components; and taking the sum of the trend item and the fluctuation item of the flow data as the estimated value of the flow data at the position to be evaluated.

[0015] By dividing the low-frequency component and the high-frequency component, the present invention can more comprehensively capture the trend characteristics and change laws of the sluice flow data; the trend item of the low-frequency component can provide a long-term and stable change direction, while the fluctuation item of the high-frequency component can reflect short-term and rapid changes. Combining the two, the flow data of the location to be evaluated can be more accurately predicted, thereby improving the accuracy of sluice control.

[0016] Preferably, the calculation formula of the convergence is: ; In the formula, The position to be evaluated and The monitoring location is The convergence on the principal components, is the length of a component, For the The first of the main sequence of monitoring positions The principal component numerical values, is the first position in the main sequence of the position to be evaluated The principal component numerical values, is the judgment function, if and The difference is less than is equal to 1 if and If the difference is greater than or equal to is equal to 0.

[0017] Preferably, the calculation of the trend item of the flow data of the position to be evaluated includes: taking the sum of the convergence of the position to be evaluated and its corresponding adjacent monitoring position on all low-frequency components as the convergence of the position to be evaluated and its corresponding adjacent monitoring position on the low-frequency component, and recording it as , ; ; In the formula, is the trend item of the traffic data at the location to be evaluated, , It is the flow data of the adjacent monitoring location corresponding to the location to be evaluated.

[0018] Preferably, the calculation of the fluctuation term of the flow data of the position to be evaluated includes: The sum of the convergence of the remaining monitoring positions on all high-frequency components is taken as the position to be evaluated and its corresponding The convergence of the remaining monitoring positions on the high frequency component is recorded as , , the remaining monitoring locations corresponding to the location to be evaluated are indivual, is the number of all monitoring locations; ; In the formula, is the fluctuation term of the flow data at the location to be evaluated, is the position to be evaluated and its corresponding The remaining monitoring locations have similar high-frequency components. is the first position corresponding to the position to be evaluated The flow data of the remaining monitoring locations, It is the trend item of the traffic data at the location to be evaluated.

[0019] The present invention makes full use of the data of all monitoring locations, including adjacent and non-adjacent ones. The flow data of adjacent monitoring locations are used to capture long-term trends, while the flow data of non-adjacent monitoring locations are helpful to capture short-term fluctuations; the existing monitoring data are used to the maximum extent, and the utilization efficiency of the data is improved.

[0020] Preferably, the adjusting of the sluice gate opening by using the obtained water level prediction value comprises: calculating the error between the obtained water level prediction value and the target water level, inputting the error between the water level prediction value and the target water level into a sluice gate opening adjustment model, obtaining the gate opening adjustment amount, and adjusting the sluice gate opening by the gate opening adjustment amount.

[0021] The beneficial effects of the present invention are:

[0022] The present invention acquires multi-dimensional data and supplements the visual information of the monitoring data through water surface videos, which helps to understand the changes in the water surface more intuitively. By setting a plurality of positions to be evaluated between adjacent monitoring positions, and obtaining an estimated value of the flow data of the position to be evaluated based on the flow data of the plurality of monitoring positions in the monitoring data, the problem of a limited number of sensors can be solved. Among them, by dividing the low-frequency component and the high-frequency component, the trend characteristics and change laws of the sluice flow data can be captured more comprehensively. At the same time, the present invention makes maximum use of the existing monitoring data, improves the utilization efficiency of the data, and more accurately predicts the flow data of the position to be evaluated. The estimated value of the flow data of the position to be evaluated enriches the flow information, realizes flow monitoring of more positions, and improves the comprehensiveness of the system and the accuracy of intelligent control. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart schematically illustrating a sluice intelligent control method based on water conservancy monitoring data in the present invention;

[0024] Figure 2 is a flow chart schematically illustrating step S2;

[0025] Figure 3 is a schematic diagram schematically showing a monitoring location and a location to be evaluated. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0028] The embodiment of the present invention discloses a sluice intelligent control method based on water conservancy monitoring data, referring to Figure 1 , including steps S1 to S3:

[0029] S1. Collect monitoring data and water surface video of the water storage body where the sluice is located.

[0030] It should be noted that the role of sluices in water conservancy projects is indeed very important, especially in flood control, drainage and water resource allocation. In order to realize intelligent control of sluices, it is necessary to collect water level data, flow data and rainfall data of the water storage body where the sluice is located through water level sensors, flow meters and rain gauges. At the same time, water surface videos of the water storage body where the sluice is located are collected to provide comprehensive data support for subsequent intelligent control.

[0031] Specifically, monitoring data of the water storage body where the sluice is located is collected, and the monitoring data includes water level data, flow data and rainfall data, which are collected by water level sensors, flow meters and rain gauges respectively; the data sampling interval is 5 minutes; wherein, the water level sensor is installed in the river or water storage body upstream and downstream of the sluice, usually near the gate, so as to monitor the water level change in real time, and the water level data is collected by the water level sensor, and the water level data can reflect the height change of the water body; the flow meter is installed in the water supply pipeline or flow structure of the sluice, and is used to measure the flow data passing through the sluice, and the flow data refers to the volume of water flowing through the sluice per unit time; the rain gauge is installed in an open area near the sluice, and is avoided as much as possible from being blocked or affected by wind, and the rainfall data is collected by the rain gauge, and the rainfall data refers to the total amount of precipitation within a certain period of time.

[0032] Through the data collection of these three sensors, the intelligent control system of the sluice gate can evaluate various information such as water conditions and weather in real time, thereby achieving fast and accurate adjustment and ensuring the safe and efficient operation of water conservancy facilities.

[0033] Furthermore, the location where the flow meter is installed is recorded as the monitoring location, and the flow data of multiple monitoring locations can be obtained through the flow meter.

[0034] Furthermore, multiple cameras are installed to monitor and collect water surface videos of the water reservoir where the sluice is located. The water surface videos collected by the multiple cameras are required to cover the entire water surface of the water reservoir where the sluice is located. Therefore, by reasonably arranging the positions and angles of the cameras, comprehensive monitoring of the water surface can be achieved to avoid monitoring blind spots.

[0035] Among them, since the camera is installed near the sluice, it may be affected by water vapor, rain, etc., so it must have good waterproof performance to ensure normal operation in a humid environment.

[0036] It should be noted that, through the acquisition of multi-dimensional data, water surface videos supplement the visual information of monitoring data, which helps to understand water surface changes more intuitively; combining monitoring data and water surface videos enriches water situation information and provides a basis for intelligent control.

[0037] S2. Estimate the flow data of multiple locations to be evaluated based on the flow data of multiple monitoring locations in the monitoring data to obtain estimated values ​​of the flow data of the locations to be evaluated.

[0038] It should be noted that the setting and number of installed sensors are limited by factors such as cost, technical complexity and site environment, resulting in insufficient monitoring data to fully understand the condition of the water area; therefore, setting multiple locations to be evaluated between adjacent monitoring locations and obtaining an estimated value of the flow data at the location to be evaluated based on the flow data of multiple monitoring locations in the monitoring data can solve the problem of limited number of sensors.

[0039] Specifically, for two adjacent monitoring locations and , , is the number of all monitoring locations; and The distance between , the distance is in meters; at two adjacent monitoring locations and There are 4 positions to be evaluated between them. Therefore, the 4 positions to be evaluated are , , , .

[0040] For example, in Figure 3 In the schematic diagram of the monitoring positions and the positions to be evaluated shown, the black dots represent the monitoring positions. There are 6 monitoring positions in total, namely monitoring position W1, monitoring position W2, monitoring position W3, monitoring position W4, monitoring position W5 and monitoring position W6. The white dots represent the positions to be evaluated, and these 4 positions to be evaluated are the positions to be evaluated set between monitoring position W3 and monitoring position W4.

[0041] Furthermore, the pixel points corresponding to each monitoring position / position to be evaluated are marked in the water surface video. Since the monitoring position / position to be evaluated is fixed and the position and angle of the camera are also fixed, the position of the pixel points corresponding to each monitoring position / position to be evaluated is the same in all video frames of the water surface video.

[0042] According to the flow data of multiple monitoring locations in the monitoring data, the flow data of multiple locations to be evaluated are estimated to obtain the estimated value of the flow data of the locations to be evaluated. The flowchart of step S2 refers to Figure 2 , including steps S201 to S205, specifically:

[0043] S201, obtaining a master sequence and a subordinate sequence of each monitoring position, and obtaining a master sequence of each position to be evaluated.

[0044] Specifically, the grayscale values ​​of the pixels corresponding to the monitoring positions in all video frames form the main sequence of the monitoring positions, and each monitoring position has only one main sequence; the grayscale values ​​of the pixels corresponding to the position to be evaluated in all video frames form the main sequence of the position to be evaluated, and each position to be evaluated has only one main sequence; the main sequence reflects the grayscale value changes of the monitoring position / position to be evaluated.

[0045] Furthermore, a monitoring area centered on a pixel point corresponding to the monitoring position in each video frame is obtained, the size of the monitoring area is equal to 7×7, and the position of the monitoring area corresponding to the same monitoring position in all video frames is the same.

[0046] Furthermore, the grayscale values ​​of the pixels at the same position in the same monitoring area of ​​all video frames are combined to form a subordinate sequence of the monitoring position. Since the monitoring area contains pixels at multiple different positions, there are multiple subordinate sequences for each monitoring position; the subordinate sequence reflects the changes in the grayscale values ​​of the pixels in the monitoring area.

[0047] S202: Obtain multiple components of a main sequence and a subordinate sequence by decomposition.

[0048] It should be noted that Empirical Mode Decomposition (EMD) is a data-driven decomposition method that can decompose complex signals into multiple intrinsic mode functions (IMFs), also called components; each component represents a vibration mode of a different time scale in the signal, and the number of decomposed components depends on the complexity of the signal itself.

[0049] Specifically, the main sequence of each monitoring position and each position to be evaluated is decomposed by empirical mode decomposition to obtain multiple components corresponding to each main sequence; the multiple subordinate sequences of each monitoring position are decomposed by empirical mode decomposition to obtain multiple components corresponding to the multiple subordinate sequences.

[0050] The components obtained by decomposition are from low order to high order, that is, the frequency is from high to low, wherein the high-frequency components represent high-frequency noise, and the low-frequency components represent low-frequency trends.

[0051] It should be noted that by decomposing multiple components of the main sequence and the subordinate sequence, a complex sequence is decomposed into multiple simple components, which is convenient for separate analysis and processing.

[0052] S203, calculating the stability of the component by the distribution of the main sequence of the monitoring position and all the subordinate sequences on the same component; calculating the distinguishability of the component by the distribution of the main sequences of different monitoring positions on the same component.

[0053] It should be noted that the traffic data at each location in the local monitoring area are similar. Therefore, the subordinate sequences of the monitoring location are the sequences of other locations in the adjacent monitoring area, and the components of these subordinate sequences should maintain similar distribution characteristics to the components of the main sequence of the corresponding monitoring location. Based on this feature, the distribution of the main sequence of the monitoring location and all the subordinate sequences on the same component is analyzed, and the stability of the distribution of the main sequence of the monitoring location and all the subordinate sequences on the same component is calculated.

[0054] Specifically, the stability of the component is calculated by monitoring the distribution of the main sequence of the position and all the subordinate sequences on the same component; The calculation formula for the stability of each component is as follows:

[0055] ;

[0056] In the formula, For the The stability of the components, For the The first of all subordinate sequences of the monitoring position The first The mean of the values, For the The first of the main sequence of monitoring positions The first numerical values, To take the absolute value function, is the number of all monitoring locations, is the length of a component, is a natural exponential function.

[0057] in, Indicates The first of all subordinate sequences of the monitoring position The first The mean of the values ​​and The first of the main sequence of monitoring positions The first The greater the difference, the less stable the distribution of the main sequence at the monitoring position and all the subordinate sequences on the same component.

[0058] It should be noted that there are differences between the flow data at different monitoring locations. Therefore, for the main sequences at different monitoring locations, the distribution characteristics of these main sequences on the same component are relatively discrete. Based on this feature, the distribution of the main sequences at all monitoring locations on the same component is analyzed, and the distinctiveness of the distribution of the main sequences at different monitoring locations on the same component is calculated.

[0059] Specifically, the component discrimination is calculated by the distribution of the main sequence at different monitoring locations on the same component; The calculation formula of the discriminability of each component is as follows:

[0060] ;

[0061] In the formula, For the The distinguishability of the components, is the first of the main sequences for all monitoring locations The first The variance of the values, is the first of the main sequences for all monitoring locations The first The mean of the values, is the length of a component.

[0062] in, The first The distribution dispersion of the components, variance The larger the mean The smaller it is, the more discrete the distribution of the components is, and accordingly, the greater the distinguishability of the components is.

[0063] S204, selecting a plurality of principal components from all components according to stability and distinguishability; and dividing all principal components into a plurality of high-frequency components and a plurality of low-frequency components according to frequency.

[0064] It should be noted that the flow data at each location in the local monitoring area are similar, therefore, the components between different sequences in the local monitoring area are stable, so when selecting the principal component, the component with greater stability, that is, less stability, should be selected as the principal component; the flow data at different monitoring locations are different, therefore, the components between the main sequences at different monitoring locations are distinguishable, so when selecting the principal component, the component with greater distinguishability, that is, greater discrimination, should be selected as the principal component.

[0065] Specifically, the mean of the stability of all components is calculated and used as the first threshold; the mean of the discriminability of all components is calculated and used as the second threshold; and the components whose stability is greater than the first threshold and whose discriminability is greater than the second threshold are taken as the principal components.

[0066] It should be noted that the stability of the component is calculated by the distribution of the main sequence of the monitoring position and all the subordinate sequences on the same component. Therefore, the stability reflects the stability and consistency between the main sequence and the subordinate sequences of the component at the monitoring position. The smaller the stability, the more consistent the distribution of the component on the main sequence and the subordinate sequences at the monitoring position, and the more suitable it is as a reference component for prediction; the discriminability of the component is calculated by the distribution of the main sequences of different monitoring positions on the same component. Therefore, the discriminability reflects the difference and inconsistency between the components at different monitoring positions. The greater the discriminability, the smaller the distribution difference of the component between different monitoring positions, and the more suitable it is as a reference component for prediction; the selection of the main component can improve the accuracy and reliability of subsequent evaluation.

[0067] Furthermore, all principal components are divided into a plurality of high-frequency components and a plurality of low-frequency components according to the frequencies, including: sorting all principal components in descending order of frequency, taking the first 50% of the principal components with the largest frequencies after sorting as high-frequency components, and taking the last 50% of the principal components with the smallest frequencies after sorting as low-frequency components.

[0068] It should be noted that the principal components are divided into high-frequency components and low-frequency components. The high-frequency components can reflect the short-term fluctuations of the data, while the low-frequency components can reflect the long-term trends of the data.

[0069] S205. For any position to be evaluated, determine the estimated value of the flow data of the position to be evaluated by the convergence of the low-frequency components between the position to be evaluated and the adjacent monitoring positions, and the convergence of the high-frequency components between the position to be evaluated and the remaining monitoring positions.

[0070] Specifically, for any position to be evaluated, the monitoring position on the left side of the position to be evaluated and adjacent to the position to be evaluated and the monitoring position on the right side of the position to be evaluated and adjacent to the position to be evaluated are used as the adjacent monitoring positions corresponding to the position to be evaluated; therefore, there are two adjacent monitoring positions corresponding to each position to be evaluated, and accordingly, there are two other monitoring positions corresponding to each position to be evaluated. indivual, is the number of all monitoring locations.

[0071] For example, in Figure 3 In the schematic diagram of the monitoring positions and the positions to be evaluated, the adjacent monitoring positions of the four positions to be evaluated are monitoring position W3 and monitoring position W4, and the remaining monitoring positions include monitoring position W1, monitoring position W2, monitoring position W5 and monitoring position W6.

[0072] Furthermore, for any location to be evaluated, the estimated value of the flow data of the location to be evaluated is determined by the convergence of the location to be evaluated and the adjacent monitoring locations on the low-frequency component, and the convergence of the location to be evaluated and the remaining monitoring locations on the high-frequency component: including:

[0073] 1. According to the convergence calculation formula, calculate the convergence of the location to be evaluated and each monitoring location on each component.

[0074] The calculation formula of convergence is: ; In the formula, The position to be evaluated and The monitoring location is The convergence on the principal components, is the length of a component, For the The first of the main sequence of monitoring positions The principal component numerical values, is the first position in the main sequence of the position to be evaluated The principal component numerical values, is the judgment function, if and The difference is less than is equal to 1 if and If the difference is greater than or equal to is equal to 0.

[0075] 2. The sum of the convergences of the position to be evaluated and its corresponding adjacent monitoring position on all low-frequency components is taken as the convergence of the position to be evaluated and its corresponding adjacent monitoring position on the low-frequency component, and is recorded as , .

[0076] 3. Associate the position to be evaluated with its corresponding The sum of the convergence of the remaining monitoring positions on all high-frequency components is taken as the position to be evaluated and its corresponding The convergence of the remaining monitoring positions on the high frequency component is recorded as , , the remaining monitoring locations corresponding to the location to be evaluated are indivual, is the number of all monitoring locations.

[0077] 4. Calculate the trend item of the flow data of the location to be evaluated through the convergence of the low-frequency components between the location to be evaluated and the adjacent monitoring location, as well as the flow data of the adjacent monitoring location.

[0078] It should be noted that the low-frequency component reflects the long-term trend of the data and can reflect the overall direction of the flow data. By analyzing the convergence of the low-frequency components between the location to be evaluated and the adjacent monitoring locations, we can understand the similarities in the long-term trends of these locations. Since adjacent locations are geographically close, their long-term trends often have certain similarities. Therefore, the flow data of the adjacent monitoring locations can be used to infer the trend items of the flow data of the location to be evaluated.

[0079] The calculation formula for the trend item of the flow data at the location to be evaluated is:

[0080] ;

[0081] In the formula, is the trend item of the traffic data at the location to be evaluated, , is the flow data of the adjacent monitoring location corresponding to the location to be evaluated, , It is the convergence of the low-frequency components between the position to be evaluated and its corresponding adjacent monitoring position.

[0082] 5. Calculate the fluctuation term of the flow data of the location to be evaluated through the convergence of the high-frequency components between the location to be evaluated and the other monitoring locations, as well as the flow data of the other monitoring locations.

[0083] It should be noted that the high-frequency component represents the short-term fluctuation of the data and can capture the rapid changes in flow data; the convergence of the high-frequency components between the location to be evaluated and the other monitoring locations reflects the similarity of these locations in short-term fluctuations; therefore, the flow data of the other monitoring locations can be used to calculate the fluctuation term of the flow data of the location to be evaluated.

[0084] Among them, the calculation formula for the fluctuation term of the flow data at the location to be evaluated is:

[0085] ;

[0086] In the formula, is the fluctuation term of the flow data at the location to be evaluated, , is the position to be evaluated and its corresponding , The remaining monitoring locations have similar high-frequency components. is the first position corresponding to the position to be evaluated The flow data of the remaining monitoring locations, It is the trend item of the traffic data at the location to be evaluated.

[0087] It should be noted that by dividing all components into low-frequency components and high-frequency components, and dividing all monitoring locations into adjacent monitoring locations and remaining monitoring locations, the trend characteristics and change laws of the sluice flow data can be fully captured, and the existing monitoring data can be used to the maximum extent to improve the efficiency of data utilization.

[0088] 6. The sum of the trend item and the fluctuation item of the flow data of the location to be evaluated is used as the estimated value of the flow data of the location to be evaluated.

[0089] It should be noted that the combination of trend items and fluctuation items improves the accuracy of the estimated value of the flow data at the location to be evaluated.

[0090] S3. The monitoring data and the estimated value of the flow data at the location to be evaluated are input into the water level prediction model to obtain the water level prediction value, and the sluice gate opening is adjusted according to the water level prediction value.

[0091] In this embodiment, the water level prediction model adopts a long short-term memory network (LSTM) structure, and the model input includes the monitoring data of the past 24 hours and the estimated values ​​of the flow data of all the locations to be evaluated in the past 24 hours, wherein the long short-term memory network consists of 4 layers, namely: 1 input layer, 2 hidden layers and 1 output layer, and the loss function adopted is the root mean square error; in other embodiments, other neural network models can be selected to train the water level prediction model, such as RNN network (Rerrent Neural Network) and GRU network (Gated Recurrent Unit).

[0092] Specifically, the error between the predicted water level and the target water level is calculated, the error between the predicted water level and the target water level is input into a sluice gate opening adjustment model, the gate opening adjustment amount is obtained, and the sluice gate opening is adjusted according to the gate opening adjustment amount.

[0093] In this embodiment, the sluice gate opening adjustment model is based on the fuzzy PID control strategy. The input of the model is the error between the water level prediction value and the target water level, and the output is the gate opening adjustment amount. The membership function of the fuzzy set adopts the Gaussian function, and there are 7 fuzzy subsets in total.

[0094] It should be noted that the estimated value of the flow data at the location to be evaluated enriches the flow information, realizes flow monitoring of more locations, and improves the comprehensiveness of the system and the accuracy of intelligent control.

Claims

1. A sluice intelligent control method based on water conservancy monitoring data, characterized in that: include: Collect monitoring data and water surface video of the water reservoir where the sluice is located; According to the flow data of multiple monitoring locations in the monitoring data, the flow data of multiple locations to be evaluated are estimated, and the data are input into the water level prediction model together with the monitoring data, and the sluice opening is adjusted according to the obtained water level prediction value; Estimating the flow data of multiple locations to be evaluated, including: forming a main sequence of the monitoring location / location to be evaluated by gray values ​​of pixels corresponding to the monitoring location / location to be evaluated in all video frames; obtaining a monitoring area centered on the pixel corresponding to the monitoring location in each video frame, and forming a subordinate sequence of the monitoring location by gray values ​​of pixels at the same position in the same monitoring area of ​​all video frames; obtaining multiple components of the main sequence and the subordinate sequence by decomposition; and calculating the stability of the component by the distribution of the main sequence of the monitoring location and all subordinate sequences on the same component; The distinguishability of the components is calculated through the distribution of the main sequences of different monitoring locations on the same component; multiple main components are selected from all components based on stability and distinguishability; all main components are divided into multiple high-frequency components and multiple low-frequency components based on frequency; for any position to be evaluated, the estimated value of the flow data at the position to be evaluated is determined through the convergence of the position to be evaluated and the adjacent monitoring locations on the low-frequency components, and the convergence of the position to be evaluated and the remaining monitoring locations on the high-frequency components.

2. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The method for setting the position to be evaluated includes: For two adjacent monitoring locations and , , is the number of all monitoring locations; and The distance between ; At two adjacent monitoring locations and There are 4 positions to be evaluated between them. Therefore, the 4 positions to be evaluated are , , , .

3. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The stability of the calculation components includes: ； In the formula, For the The stability of the components, For the The first of all subordinate sequences of the monitoring position The first The mean of the values, For the The first of the main sequence of monitoring positions The first numerical values, To take the absolute value function, is the number of all monitoring locations, is the length of a component, is a natural exponential function.

4. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: Calculate the distinctiveness of the components, including: ； In the formula, For the The distinguishability of the components, is the first of the main sequences for all monitoring locations The first The variance of the values, is the first of the main sequences for all monitoring locations The first The mean of the values, is the length of a component.

5. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The method of selecting a plurality of principal components from all components according to stability and distinguishability comprises: The mean of the stability of all components is calculated and used as the first threshold; the mean of the discriminability of all components is calculated and used as the second threshold; the components whose stability is greater than the first threshold and whose discriminability is greater than the second threshold are taken as the principal components.

6. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: Determining an estimated value of the flow data of the location to be evaluated includes: According to the calculation formula of convergence, the convergence of the position to be evaluated and each monitoring position on each component is calculated; The trend item of the flow data of the location to be evaluated is calculated by summing the convergence of the location to be evaluated and its corresponding adjacent monitoring locations on all low-frequency components; the fluctuation item of the flow data of the location to be evaluated is calculated by summing the convergence of the location to be evaluated and its corresponding remaining monitoring locations on all high-frequency components; the sum of the trend item and the fluctuation item of the flow data is used as the estimated value of the flow data of the location to be evaluated.

7. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The calculation formula of the convergence is: ; In the formula, The position to be evaluated and The monitoring location is The convergence on the principal components, is the length of a component, For the The first of the main sequence of monitoring positions The principal component numerical values, is the first position in the main sequence of the position to be evaluated The principal component numerical values, is the judgment function, if and The difference is less than is equal to 1 if and If the difference is greater than or equal to is equal to 0.

8. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The step of calculating the trend item of the flow data of the location to be evaluated includes: The sum of the convergences of the position to be evaluated and its corresponding adjacent monitoring position on all low-frequency components is taken as the convergence of the position to be evaluated and its corresponding adjacent monitoring position on the low-frequency component, and is recorded as , ; ； In the formula, is the trend item of the traffic data at the location to be evaluated, , It is the flow data of the adjacent monitoring location corresponding to the location to be evaluated.

9. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The calculation of the fluctuation item of the flow data of the location to be evaluated includes: The position to be evaluated and its corresponding The sum of the convergence of the remaining monitoring positions on all high-frequency components is taken as the position to be evaluated and its corresponding The convergence of the remaining monitoring positions on the high frequency component is recorded as , , the remaining monitoring locations corresponding to the location to be evaluated are indivual, is the number of all monitoring locations; ； In the formula, is the fluctuation term of the flow data at the location to be evaluated, is the position to be evaluated and its corresponding The remaining monitoring locations have similar high-frequency components. is the first position corresponding to the position to be evaluated The flow data of the remaining monitoring locations, It is the trend item of the traffic data at the location to be evaluated.

10. The sluice intelligent control method based on water conservancy monitoring data according to claim 1 is characterized in that: The step of adjusting the sluice gate opening by using the obtained water level prediction value comprises: The error between the predicted water level and the target water level is calculated, and the error between the predicted water level and the target water level is input into the sluice opening adjustment model to obtain the gate opening adjustment amount, and the sluice opening is adjusted according to the gate opening adjustment amount.

Images