River flow measurement lead fish control method and system

Abstract

The invention relates to the technical field of river channel flow measurement, in particular to a river channel flow measurement lead fish control method and system, and the method comprises the following steps: S1, horizontally moving a lead fish to a position right above a to-be-detected depth point through a winch system, calculating a distance L, and then lowering the lead fish; s2, detecting and uploading the deviation angle and the flow velocity in real time; s3, processing information, adjusting the position of the fish lead to a to-be-detected depth point, collecting the flow velocity, analyzing a flow velocity result, dynamically selecting and training a prediction model, and calculating compensation power according to the prediction model and the power compensation model; and S4, adjusting the position of the fish lead according to the compensation power until the deviation angle is within the prediction error range, and recording the flow velocity of an ideal point. A prediction model is dynamically selected according to the change condition of real-time information, the prediction precision is improved compared with a fixed model, the accuracy and stability of positioning of the fish lead under complex water flow are improved by establishing a power compensation model and combining an intelligent closed-loop control mechanism fed back in real time, and the method is an innovation and improvement of a traditional control mode.

Description

Technical Field

[0001] The present invention relates to the technical field of river flow measurement, and specifically, to a method and system for controlling a lead fish for river flow measurement. Background Art

[0002] River flow monitoring is an important basic task in water conservancy projects, water resources management, and flood control and disaster reduction. As a commonly used flow measurement device, the core task of the lead fish is to stably place the carried ADCP probe at the target depth point to obtain accurate real-time flow velocity data. However, the existing lead fish control technology has the following core problems: First, the operator needs to visually observe the projection position of the lead fish on the water surface and estimate the position directly above the target point by combining the release length of the winch steel rope. Affected by environmental factors such as light and water flow fluctuations, the positioning error usually exceeds 50 cm. In scenarios such as deep water areas or complex riverbed terrains (such as hidden reefs and deep pools), the reliability of manual judgment further decreases, resulting in a horizontal offset of 1 - 2 meters between the actual point and the ideal point after the lead fish is lowered, directly affecting the flow measurement accuracy of the ADCP probe; Second, the river flow velocity shows significant dynamic change characteristics (such as sudden flood peaks, tidal cycles, ship wakes, etc.), but the existing lead fish control technology only relies on the static balance between the self-weight (gravity) of the lead fish and the steel rope tension, lacking the ability of active compensation. When the river impact force is greater than the component force of the lead fish gravity along the water flow direction (such as in a rapid flow environment with a flow velocity > 1.5 m / s), the lead fish will deviate significantly, and the deviation angle can reach 15° to 30°, resulting in a large amount of lateral movement noise in the flow velocity data collected by the ADCP probe and unable to reflect the true flow velocity of the target point. In traditional methods, the operator needs to wait for the water flow to calm down temporarily or manually adjust the steel rope length, which takes up to several minutes and seriously affects the flow measurement efficiency; Third, most of the existing lead fish control systems are open-loop designs (such as setting the release length of the steel rope and then standing still), lacking a real-time feedback mechanism. For example, when the water flow direction suddenly changes, the deviation angle of the lead fish may increase to the critical value within a few seconds, but the control system cannot automatically trigger power compensation, resulting in the failure of the flow measurement task. Field tests show that in rivers with frequent flow velocity fluctuations, the effective data collection rate is only 65% - 75%, which cannot meet the high-precision monitoring requirements.

[0003] In view of this, a method and system for controlling a lead fish for river flow measurement are proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for controlling a lead fish for river flow measurement to solve the problem that there is a position deviation between the lead fish and the target depth point and the lead fish cannot be stable at the target depth point in different river flow velocity environments.

[0005] To solve the above technical problems, the present invention provides a method for controlling a lead fish for river flow measurement, including the following steps: S1. Through a winch system, first horizontally move the lead fish to directly above the depth point of the river to be detected, then calculate the distance from the lead fish to the depth point of the river to be detected, denoted as L. By releasing the steel cable, lower the lead fish to the depth point of the river to be detected; S2. During the process of lowering the lead fish, use the angle sensor built in the lead fish to detect the offset angle between the steel cable and the lead fish and the lowering dimension of the steel cable, and upload the offset angle and the lowering dimension to the information processing module. Until the lowering dimension of the steel cable is L, record the actual river depth point where the lead fish is fixed as the actual point, and the depth point of the river to be detected as the ideal point; S3. According to the offset angle, use the power output by the power module of the lead fish to move the lead fish within the allowable error range of the offset angle, regarded as the position of the lead fish at the ideal point. At the same time, collect the flow velocity information of the lead fish at the ideal point and input it into the information processing module. The information processing module selects a prediction model according to the analysis result of the flow velocity information, uses the prediction model to predict the flow velocity at the ideal point, and calculates the predicted compensation power at the ideal point according to the power compensation model; S4. Control the power module of the lead fish to output according to the predicted compensation power, so that the lead fish is stable within the error range of the ideal point, and collect the real-time flow velocity of the current position through the ADCP probe of the lead fish.

[0006] As a further improvement of this technical solution, in S1, the winch system includes a motor, a drum, a transmission device, a control circuit and a control terminal. Both ends of the steel cable are respectively connected to the drum and the lead fish. The control terminal issues an instruction to the control circuit to make the transmission device drive the drum to horizontally move the lead fish to directly above the depth point of the river to be detected.

[0007] As a further improvement of this technical solution, in S1, through the GPS positioning system and the coordinate information of the depth point of the river to be detected, determine L by calculating the vertical distance between the current position of the lead fish and the point to be detected. After determining the distance L, the control terminal issues an instruction to the control circuit to make the motor drive the drum to rotate, release the steel cable, and the lead fish is lowered to the depth point of the river to be detected under the action of gravity.

[0008] As a further improvement of this technical solution, in S3, the information processing module includes a data access unit, a data processing unit, a model adaptation training unit and a power compensation model, where The data access unit is used to receive in real time the flow velocity information collected by the lead fish and the offset angle information between the steel cable and the lead fish. When the offset angle exceeds the allowable error range of the offset angle, trigger the cache mechanism, and the subsequent data is stored in the buffer area. The buffer area adopts a circular queue structure, and the maximum capacity is a preset group of data; The data processing unit is connected to the data access unit and is used to extract and clean the information in the buffer area, and perform feature extraction on the information after data cleaning based on the sliding window technology. The fluctuation range of the offset angle and the spectral characteristics of the flow velocity are obtained for each window, and the change amplitudes of the two are statistically calculated to form an analysis result and upload it; The model adaptation training unit is connected to the data processing unit. The model adaptation training unit includes a prediction model library. According to the uploaded analysis result, an appropriate prediction model is selected from the prediction model library; the offset angle and flow velocity information between the lead fish and the steel wire rope of the actual point are input into the prediction model, and the predicted flow velocity of the ideal point is output and uploaded to the dynamic compensation model; The dynamic compensation model is connected to the model adaptation training unit and is used to receive the predicted flow velocity information and calculate the compensation power.

[0009] As a further improvement of this technical solution, the prediction model library includes models of linear regression, random forest, LSTM, and support vector regression. The historical database labels the corresponding historical data according to the model selection conditions, and each type of model is trained with the historical data corresponding to the labels in the historical database; The model selection conditions are as follows: based on the change amplitudes of the offset angle fluctuation range and the flow velocity spectral characteristics, when the change amplitudes of both are < the preset value and stable, a linear regression model is selected; when the change amplitude of one of them ≥ the preset value and the other also shows an abnormal change synchronously, a random forest model is selected; when the analysis result shows that the changes in each window have a time series pattern, an LSTM model is selected; when the linear regression model, random forest model, and LSTM model are not satisfied, a support vector regression model is selected.

[0010] As a further improvement of this technical solution, according to the uploaded analysis result, an appropriate prediction model is selected from the prediction model library as the model used for the current window. At the same time, the data within the current window is divided into a front sub-interval and a rear sub-interval according to the time period, and the change amplitudes of the offset angle fluctuation range and the flow velocity spectral characteristics of the front sub-interval and the rear sub-interval are calculated respectively, and the mutation coefficients of the front sub-interval and the rear sub-interval are calculated according to the change amplitudes , if the change amplitudes of the two in the rear sub-interval meet the corresponding conditions in the model selection conditions, the corresponding model is used for prediction in the next window; if the change amplitudes of the two in the rear sub-interval do not meet any of the model selection conditions, but the mutation coefficient is within the preset tolerance range, the pre-activated state of the corresponding model is triggered, and the corresponding model is used for prediction in the next window.

[0011] As a further improvement of the technical solution, in S4, by inputting the compensation power parameters into the control system of the lead fish power module, the power module drives the propulsion device of the lead fish to generate thrust according to the instructions issued by the control system to adjust the position of the lead fish. At the same time, the angle sensor continuously collects the deviation angle of the lead fish and feeds it back to the control system to continuously compare the difference between the real-time angle and the zero value. When the deviation angle decreases to within the preset allowable error range and the stable holding time ≥ the preset value, it is determined that the lead fish is stable at the ideal point, and the control system triggers the ADCP probe to start measuring the real-time flow velocity at the ideal point. The measurement data is transmitted to the data processing unit through the communication module of the lead fish for storage. At the same time, the control system monitors the deviation angle of the lead fish in real time. If the fluctuation value of the deviation angle ≤ the preset value, the recorded data is confirmed to be valid. If an abnormality occurs, the power output is immediately readjusted until the stable condition is met and then the ADCP probe is triggered to record again.

[0012] A control system for a river flow measurement lead fish, which is used to implement the above-mentioned river flow measurement lead fish control method, includes: A winch system, which is used to move the lead fish directly above the point of the depth of the river to be detected, calculate the distance from the lead fish to the point of the depth of the river to be detected, the lowering dimension of the steel wire rope, and lower the lead fish. A lead fish system, which is used to collect the flow velocity information of the actual river depth point, the deviation angle between the steel wire rope and the lead fish in real time, and perform power output according to the predicted compensation power uploaded by the information processing module, and determine whether the lead fish is stable at the point of the depth of the river to be detected. An information processing module, which is connected to the winch system and the lead fish system, is used to receive the information of the flow velocity at the actual river depth point and the deviation angle between the steel wire rope and the lead fish in real time, analyze the characteristics and relevance of the information, select a prediction model according to the analysis results and train it, use the trained prediction model to predict the flow velocity at the ideal point, and calculate the predicted compensation power at the ideal point according to the power compensation model.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. In the river flow measurement lead fish control method and system, according to the changes of multi-dimensional data such as the fluctuation range of the deviation angle and the flow velocity spectrum characteristics, a suitable prediction model is dynamically selected. This way of flexibly selecting a model according to the real-time data characteristics can more accurately adapt to different river flow measurement scenarios, and has a greater improvement in prediction accuracy compared with the prior art that fixedly uses a certain model.

[0014] 2. In the river flow measurement lead fish control method and system, the compensation power is calculated by the power compensation model, and the lead fish position is adjusted in combination with real-time feedback to achieve the control of the lead fish to be stable at the ideal point. This intelligent compensation and closed-loop feedback control mechanism improves the accuracy and stability of the lead fish positioning under complex water flow conditions, and is an innovative improvement on the traditional simpler lead fish control method. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the method flow of the present invention; Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all secondary embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] At present, the lead fish control technology has problems such as low manual positioning accuracy (relying on naked eye observation and empirical estimation, and being affected by the environment and terrain, resulting in significant horizontal deviation), insufficient dynamic stability (relying only on static balance, lacking active compensation mechanism, large deviation angle and difficult recovery in rapid flow environment), and lack of automatic feedback mechanism (open-loop control cannot respond to water flow changes in real time, and the effective data collection rate is low). These problems lead to low flow measurement accuracy and poor efficiency, making it difficult to adapt to the monitoring needs of complex river environments. For this reason, please see Figure 1 As shown, one of the purposes of the present invention is to provide a method for controlling a river flow measuring lead fish, the method comprising the following steps: S1. Move the lead fish horizontally to the top of the river depth point to be tested by the winch system, calculate the distance from the lead fish to the river depth point to be tested, record it as L, and release the steel rope to lower the lead fish to the river depth point to be tested; S2. During the lowering of the lead fish, the offset angle between the steel rope and the lead fish and the lowering size of the steel rope are detected by the angle sensor built into the lead fish, and the offset angle and the lowering size are uploaded to the information processing module. When the lowering size of the steel rope is L, the river depth point where the lead fish is actually fixed is recorded as the actual point, and the river depth point to be detected is recorded as the ideal point; S3. According to the offset angle, the power output by the power module of the lead fish is used to move the lead fish within the allowable error range of the offset angle. This position is regarded as the position of the lead fish at the ideal point. Meanwhile, the flow velocity information of the lead fish at the ideal point is collected and input into the information processing module. The information processing module selects a prediction model based on the analysis result of the flow velocity information, uses the prediction model to predict the flow velocity at the ideal point, and calculates the predicted compensation power at the ideal point according to the power compensation model; S4. Control the power module of the lead fish to output according to the predicted compensation power, so that the lead fish is stable within the error range of the ideal point, and the real-time flow velocity of the current position is collected through the ADCP probe of the lead fish.

[0018] In this method for controlling the lead fish for river flow measurement, according to the changes in multi-dimensional data such as the fluctuation range of the offset angle and the spectral characteristics of the flow velocity, a suitable prediction model is dynamically selected. This way of flexibly selecting a model according to the real-time data characteristics can more accurately adapt to different river flow measurement scenarios. Compared with the existing technology that fixedly uses a certain model, there is a significant improvement in the prediction accuracy. At the same time, the compensation power is calculated through the power compensation model, and the position of the lead fish is adjusted in combination with the real-time feedback to achieve the control of the lead fish being stable at the ideal point. This intelligent compensation and closed-loop feedback control mechanism improves the accuracy and stability of the lead fish positioning under complex water flow conditions, and is an innovative improvement on the traditional and relatively simple lead fish control method.

[0019] For river flow measurement, the lead fish needs to be accurately moved directly above the point to be detected. Manual operation is difficult to stably control the position of the lead fish in a complex water flow environment, and it has low efficiency and large errors. High-precision horizontal positioning needs to be achieved through mechanized and automated devices. Therefore, in step S1, the winch system includes a motor, a drum, a transmission device, a control circuit, and a control terminal. The two ends of the steel rope are respectively connected to the drum and the lead fish. The control terminal issues an instruction to the control circuit, so that the transmission device drives the drum to horizontally move the lead fish directly above the point of the river depth to be detected. Through electromechanical control and coordinate conversion, the automatic calibration of the horizontal position of the lead fish is realized, and the error can be controlled within the centimeter level, which is significantly better than manual operation, ensuring that the subsequent lowering process starts from directly above the "ideal point" and laying a foundation for accurate flow measurement. At the same time, the accurate horizontal positioning provides reliable initial conditions for the offset angle detection in step S2 and the power compensation model in step S3, reducing the compensation error caused by the initial position deviation, forming a closed-loop optimization of "positioning → detection → compensation", and ultimately improving the accuracy of the flow measurement data; In traditional river flow measurement, the lowering of the lead fish often relies on manual visual inspection, mechanical limit, or simple encoder counting, which has the following defects: First, the position is ambiguous, and it is impossible to accurately determine whether the lead fish is directly above the target point, and it can only be adjusted based on experience; Second, the length of the steel rope released is estimated manually or by a fixed formula without combining real-time position feedback; Third, the initial position deviation leads to an increase in the subsequent dynamic compensation load, and it may even be impossible to correct it to the ideal point. Therefore, in step S1, through the GPS positioning system and the coordinate information of the depth points of the river to be detected, the vertical distance between the current position of the lead fish and the point to be detected is calculated to determine L. After determining the distance L, an instruction is sent from the control terminal to the control circuit to drive the reel to rotate by the motor, release the steel rope, and the lead fish is lowered to the depth point of the river to be detected under the action of gravity; Through GPS positioning + coordinate calculation + automatic control, the "extensive lowering" is transformed into "precision positioning lowering", which solves the core pain points of the traditional method from the technical path, belongs to the key improvement of the lead fish control process, and has significant innovation and practical value.

[0020] Considering that the water flow impact causes a certain deviation angle between the lead fish and the steel rope, and the length of the steel rope released has been calculated in advance according to the ideal point, changing the length of the steel rope released may instead cause uncontrollability. Therefore, according to the deviation angle between the steel rope and the lead fish, the lead fish is moved within the allowable error range of the deviation angle through the power output by the power module of the lead fish, and it is regarded as the position of the lead fish at the ideal point.

[0021] Due to the dynamic and uncertain nature of the river environment (such as flow velocity changes and water flow impact), there may be a complex non-linear relationship between the deviation angle of the lead fish and the flow velocity. If the lead fish is to stably collect the flow velocity at the ideal point, the lead fish needs additional power to resist the changes caused by the river environment. At the same time, the amount of real-time collected data is large and the low-latency processing requirements need to be met. If the original data is directly processed, it may lead to overloading of computing resources or the model output not adapting to the actual environmental changes. Therefore, it is necessary to efficiently manage, extract features, and adapt the model to the data through a hierarchical information processing module to achieve precise prediction and dynamic compensation of the lead fish's motion state. Therefore, in S3, the information processing module includes a data access unit, a data processing unit, a model adaptation training unit, and a dynamic compensation model, where The data access unit is used to receive in real time the flow velocity information collected by the lead fish and the offset angle information between the steel wire rope and the lead fish. When the offset angle exceeds the allowable error range of the offset angle (e.g., -5° ≤ allowable error range of the offset angle ≤ 5°) (indicating strong water flow interference), the caching mechanism is triggered to activate the buffer area to store subsequent data. The buffer area adopts a circular queue structure with a maximum capacity of a preset number of data groups (e.g., 1000 data groups; by setting the caching mechanism, only the dynamic data when the water flow interference is significant is stored, reducing invalid data (such as redundant data when the lead fish is steadily lowered)), reducing the consumption of computing resources. At the same time, the circular queue ensures the "freshness" of the cached data and avoids the influence of historical stale data on the real-time nature of model training; Considering that the originally collected flow velocity and offset angle data may contain noise (such as sensor errors and water flow pulsation interference), directly using them for the analysis of the preset model will reduce the accuracy of feature extraction; moreover, the lead fish offset and flow velocity changes have temporal correlation (e.g., the offset at a certain moment may affect the flow velocity at the next moment), and it is necessary to capture the dynamic characteristics within different time windows. Therefore, the data processing unit, which is connected to the data access unit, is used to extract and clean the information in the buffer area. The data cleaning steps include mean removal, standardization (Z-score), and outlier repair (such as neighboring value interpolation) to ensure the stable distribution of the data input into the model. Then, based on the sliding window technique, feature extraction is performed on the information after data cleaning. Specifically, the window size (e.g., 100 ms) and the step size (e.g., 50 ms) are set, and the cached data is divided overlappingly to ensure the continuity of feature extraction; for each window, the fluctuation range of the offset angle (the difference between the maximum and minimum offset angles within the window (maxθ - minθ), quantifying the swaying degree of the lead fish impacted by the water flow) and the spectral characteristics of the flow velocity (performing a fast Fourier transform (FFT) on the flow velocity signal collected by the ADCP to extract the amplitudes and phases of the main frequency components, reflecting the periodic fluctuations of the water flow (such as vortex frequency)) are obtained, and the change amplitudes of the two in each window are statistically calculated to form the analysis result and upload it; and the fluctuation range and spectral characteristics directly correspond to the lead fish movement state and water flow characteristics, providing interpretable inputs for the model. Moreover, the sliding window dynamically captures signal mutations within a short time (such as sudden water flow impacts) and is more sensitive than global statistical features; Due to the significant spatio-temporal heterogeneity of river channel flow (such as the alternation of rapids / slow flows, tidal cycles, and sudden flood peaks), during the process of the lead fish collecting the flow velocity at the ideal point, the fluctuation range of the offset angle and the spectral characteristics of the flow velocity may exhibit non-linear and non-stationary characteristics. Traditional fixed models (such as single linear regression) cannot cover the complex mapping relationships in multiple scenarios, resulting in the lead fish being impacted by the water flow when collecting the flow velocity at the ideal point, causing its own position to be unstable and resulting in errors in the collected flow velocity information. Therefore, the model adaptation training unit is connected to the data processing unit. The model adaptation training unit includes a prediction model library. According to the uploaded analysis results, an appropriate prediction model is selected from the prediction model library; the offset angle and flow velocity between the lead fish and the steel wire at the actual point are input into the prediction model, and the predicted flow velocity at the actual point is output and uploaded to the dynamic compensation model; The prediction model library includes models of linear regression, random forest, LSTM, and support vector regression. Since the requirements of different models for data formats and feature types vary significantly (for example, LSTM requires time series input, and random forest depends on independent samples), the historical database labels the historical data with corresponding tags according to the model selection conditions. Each type of model is trained with the historical data with the corresponding tags in the historical database. By labeling the historical data according to the model selection conditions, it can be ensured that the corresponding model is trained with the corresponding historical data; Historical data marking: Mark the historical data based on the model selection conditions (such as the change amplitude threshold). For example: when the fluctuation range of the offset angle < 0.2 and the change amplitude of the spectral characteristics of the flow velocity < 0.2, it is marked as "suitable for linear regression"; when the change amplitude of one of them ≥ 0.5 and the characteristics of the other are synchronously abnormal, it is marked as "suitable for random forest".

[0022] Model training: Each type of model is only trained with the corresponding marked historical data. For example: the linear regression model is only trained with the data marked as "suitable for linear regression"; the LSTM model is only trained with the data marked as "time series pattern".

[0023] Calculation of predicted flow velocity: Input the offset angle and flow velocity at the actual point into the trained prediction model, and the model outputs the predicted flow velocity at the actual point (This predicted flow velocity actually changes according to the changes in the offset angle and flow velocity at the actual point and is not a fixed value), and the specific process is as follows: When the offset angle and flow velocity of the actual point are input, the "predicted flow velocity of the actual point" output by the prediction model. This "predicted flow velocity of the actual point" is essentially the flow velocity in the theoretically non-offset state. The difference between this predicted value and the true flow velocity of the actual point is the "deviation correction value". For example, if the true flow velocity of the actual point v_actual = 2 m / s and the model predicts v_ideal = 2.5 m / s, then the deviation Δv = 0.5 m / s, indicating that the power module of the lead fish needs to be adjusted to increase the flow velocity collected by the lead fish by 0.5 m / s under the condition of meeting the offset angle.

[0024] The model selection conditions are as follows: Based on the fluctuation range of the offset angle and the change amplitude of the flow velocity spectrum characteristics, when the change amplitudes of both are < the preset value (e.g., < 0.2) and stable, a linear regression model is selected, such as in a slow-flow environment where the change amplitudes of each characteristic are less than 0.2; when the change amplitude of one of them ≥ the preset value (e.g., ≥ 0.5), and the other also shows an abnormal change synchronously (e.g., the mutual information value between characteristics ≥ 0.4 (there is a non-linear interaction, such as the coupling of sudden flow velocity change and offset angle)), a random forest model is selected, such as when the sudden increase in flow velocity causes the offset angle to exceed the limit synchronously; when the analysis results show that the changes in each window have a temporal pattern, an LSTM model is selected, such as when the tide causes the flow velocity to peak every 6 hours; when the linear regression model, random forest model, and LSTM model are not satisfied, a support vector regression model is selected; At the same time, considering that the river channel water flow environment has dynamic mutation characteristics (such as sudden ship wakes, short-term heavy rainfall causing a sudden increase in flow velocity), the characteristics within the sliding window may show a non-stationary pattern of being stable in the front and changing in the back (such as slow flow in the first half and a sudden increase in the offset angle due to water flow impact in the second half). The traditional overall window processing only reflects the average characteristics within the window and cannot capture the trend change of the data in the second half, resulting in the model selection lagging behind the actual scenario change (such as the overall characteristics of the current window still conform to linear regression, but the data in the second half has shown mutation characteristics suitable for the random forest. If not predicted in advance, switching the model in the next window will cause control delay). Therefore, according to the uploaded analysis results, and in accordance with the model selection conditions, a suitable prediction model is selected from the prediction model library as the model to be used for the current window. At the same time, the data within the current window is divided into a front sub-interval (accounting for 60%) and a back sub-interval (accounting for 40%) according to the time period. The data in the back sub-interval overlaps with the first 20% of the data in the next window to ensure that the characteristics in the back sub-interval directly reflect the initial state of the next window. The change amplitudes of the offset angle fluctuation range and the flow velocity spectrum characteristics in the front sub-interval and the back sub-interval are calculated respectively, and the mutation coefficients of the three in the front sub-interval and the back sub-interval are calculated according to the change amplitudes. , ,where, represents the relative change amplitude of the th feature in the back sub-interval relative to the front sub-interval, represents the eigenvalues, representing the eigenvalues of the latter sub-interval, representing the larger value of the eigenvalues of the front and rear sub-intervals, representing the minimum value, which is an adjustable parameter and usually takes , to avoid invalid calculation caused by a zero denominator. If the change amplitudes of the two in the latter sub-interval meet the corresponding conditions in the model selection conditions, the corresponding model will be used for prediction in the next window; if the change amplitudes of the two in the latter sub-interval do not meet any of the model selection conditions, but the mutation coefficient (shows a trend close to the set threshold of a certain model selection condition, is the preset tolerance range), for example, the change amplitude of the offset angle fluctuation range is 0.4, and the mutation coefficient is 0.4. The model selection condition of the random forest model is that the change amplitude of one of the two is ≥0.5. Although the change amplitude does not reach the model selection condition of the random forest model, there is a close trend. Therefore, the pre-activated state of the random forest model will still be triggered, and this model will be preferentially used for initialization in the next window (such as pre-loading model parameters in advance to reduce calculation latency); Input the offset angle and flow velocity information of the actual point into the trained prediction model, output the predicted flow velocity of the actual point, and upload it to the dynamic compensation model; The dynamic compensation model, connected to the model adaptation training unit, is used to receive the predicted flow velocity information and calculate the compensation power. The dynamic compensation model takes the deviation amount of the steel rope offset angle and the flow velocity prediction error value in the historical data as the core training features, and conducts iterative learning in combination with the position calibration effect after actual power adjustment. The specific principle is as follows: From the dimension of physical force characteristics: Deviation amount of offset angle (Δθ): The difference between the actual offset angle θ and the ideal point offset angle θ0 (θ0 = 0°) in the historical data, reflecting the deviation between the current force state of the lead fish and the ideal equilibrium state. For example: Δθ = θ - θ0 = 5°, indicating that the lead fish needs to overcome the water flow impact force corresponding to a 5° offset; Flow velocity prediction error (Δv): The difference between the predicted ideal point flow velocity videal and the measured flow velocity vreal of the actual point, that is, Δv = videal - vreal. This error is essentially a direct manifestation of insufficient or excessive dynamic compensation. For example, when Δv = 0.3m / s, additional power is required to increase the flow velocity by 0.3m / s.

[0025] From the dimension of dynamic adjustment effect: Dynamic response data: After each application of the compensation power F (including the value and direction of the force) in the historical record, the change in the offset angle Δθ of the lead fish (e.g., Δθ changes from 5° to 3°) and the convergence rate of the flow velocity error Δv (e.g., the time for 0.3 m / s to 0.1 m / s) form a mapping pair of "power input - state change" (F, [Δθ, Δv]); Based on the above two dimensions, the dynamic compensation model is iteratively trained through the deviation of the wire rope offset angle, the flow velocity prediction error value in the historical data, and the position calibration effect after actual power adjustment until a trained dynamic compensation model is obtained. Then, by receiving the predicted flow velocity information, the value and direction of the compensation power F are calculated.

[0026] When the lead fish is impacted by the water flow in the river, it will generate an offset. Relying solely on lowering the wire rope cannot accurately locate the ideal point (the depth point to be detected). It is necessary to adjust the power output in real time through closed-loop feedback control to offset the influence of the water flow force and ensure that the lead fish is stable at the target position, avoiding measurement data deviation caused by continuous offset. Therefore, in step S4, by inputting the compensation power parameters into the control system of the lead fish power module, the power module drives the propulsion device of the lead fish to generate thrust according to the instructions issued by the control system to adjust the position of the lead fish. For example, the angle sensor collects the offset angle at a frequency of 100 Hz and transmits it to the control system in real time through the SPI bus to form a deviation signal with the zero angle of the ideal point. The control system uses the PID control algorithm to calculate the compensation power output according to the deviation signal. This output signal drives the propulsion device (such as a micro propeller with a rotational speed range of 0 - 5000 RPM) to generate a reverse thrust to offset the lateral force of the water flow on the lead fish. At the same time, the angle sensor continuously collects the offset angle of the lead fish and feeds it back to the control system to continuously compare the difference between the real-time angle and the zero value; when the offset angle decreases to within the preset error tolerance range (e.g., the allowable range is -5° to 5°, determined according to the measurement accuracy of the ADCP probe to ensure that the lateral displacement < 5 cm), and the stable holding time ≥ the preset value (e.g., the stable holding time ≥ 5S to avoid misjudgment due to instantaneous water flow fluctuations), it is determined that the lead fish is stable at the ideal point, and the control system triggers the ADCP probe to start measuring the real-time flow velocity at the ideal point. The measurement data is transmitted to the data processing unit through the communication module of the lead fish for storage. At the same time, the control system monitors the offset angle of the lead fish in real time. If the fluctuation value of the offset angle ≤ the preset value (e.g., the fluctuation value of the offset angle ≤ 0.5°), the record is confirmed to be valid; if the fluctuation value of the offset angle > 0.5° during the determination process, the re-calibration process is immediately triggered, the current power output is stopped, the lead fish briefly enters the "free drift" state, three windows of characteristic data are re-collected, the corresponding model is matched through the model adaptation training unit, and a new round of power adjustment is started until the stable condition is met again to trigger the ADCP probe to record; Please refer to Figure 2As shown, the second object of the present invention is to provide a control system for a river flow measurement lead fish, which is used to implement the above-mentioned river flow measurement lead fish control method, and includes: a winch system, a lead fish system, and an information processing module. The winch system is used to move the lead fish directly above the depth point of the river to be detected, calculate the distance from the lead fish to the depth point of the river to be detected, the lowering dimension of the steel wire rope, and lower the lead fish. The lead fish system is used to collect the flow velocity information of the actual river depth point, the offset angle between the steel wire rope and the lead fish in real time, and perform power output according to the predicted compensation power uploaded by the information processing module, and determine whether the lead fish is stable at the depth point of the river to be detected. The information processing module is connected to the winch system and the lead fish system, and is used to receive the information of the flow velocity of the actual river depth point and the offset angle between the steel wire rope and the lead fish in real time, analyze the characteristics and relevance of the information, select a prediction model according to the analysis result and train it, use the trained prediction model to predict the flow velocity of the ideal point, and calculate the predicted compensation power of the ideal point according to the power compensation model.

[0027] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A control method for a river flow measurement lead fish, characterized in that, It includes the following steps: S1. Through the winch system, first horizontally move the lead fish to directly above the river depth point to be detected, then calculate the distance from the lead fish to the river depth point to be detected, denoted as L. By releasing the steel cable, lower the lead fish to the river depth point to be detected; S2. During the process of lowering the lead fish, use the angle sensor built in the lead fish to detect the deviation angle between the steel cable and the lead fish and the lowering dimension of the steel cable, and upload the deviation angle and the lowering dimension to the information processing module. Until the lowering dimension of the steel cable is L, record the actual river depth point where the lead fish is fixed as the actual point, and the river depth point to be detected as the ideal point; S3. According to the deviation angle, use the power output by the power module of the lead fish to move the lead fish within the allowable error range of the deviation angle, which is regarded as the position of the lead fish at the ideal point. At the same time, collect the flow velocity information of the lead fish at the ideal point and input it to the information processing module. The information processing module selects a prediction model based on the analysis result of the flow velocity information, uses the prediction model to predict the flow velocity at the ideal point, and calculates the predicted compensation power at the ideal point according to the power compensation model; S4. Control the power module of the lead fish to output according to the predicted compensation power, so that the lead fish is stable within the error range of the ideal point, and collect the real-time flow velocity at the current position through the ADCP probe of the lead fish.

2. The method for controlling a river flow measuring lead fish according to claim 1, characterized in that: In S1, the winch system includes a motor, a drum, a transmission device, a control circuit and a control terminal. Both ends of the steel cable are respectively connected to the drum and the lead fish. The control terminal issues an instruction to the control circuit to drive the transmission device to drive the drum to horizontally move the lead fish to directly above the river depth point to be detected.

3. The method for controlling a river flow measuring lead fish according to claim 2, characterized in that: In S1, through the GPS positioning system and the coordinate information of the river depth point to be detected, determine L by calculating the vertical distance between the current position of the lead fish and the point to be detected. After determining the distance L, the control terminal issues an instruction to the control circuit to drive the motor to drive the drum to rotate, release the steel cable, and the lead fish is lowered to the river depth point to be detected under the action of gravity.

4. The channel flow measurement lead fish control method according to claim 1, characterized in that: In S3, the information processing module includes a data access unit, a data processing unit, a model adaptation training unit and a power compensation model. Among them, The data access unit is used to receive in real time the flow velocity information collected by the lead fish and the deviation angle information between the steel cable and the lead fish. When the deviation angle exceeds the allowable error range of the deviation angle, trigger the cache mechanism, and the subsequent data is stored in the cache area. The cache area adopts a circular queue structure, and the maximum capacity is a preset group of data; The data processing unit, connected to the data access unit, is used to extract and clean the information in the cache area, and perform feature extraction on the information after data cleaning based on the sliding window technology. The fluctuation range of the deviation angle and the spectral characteristics of the flow velocity are obtained for each window, and the change amplitude of the two in each window is statistically calculated to form an analysis result and upload it; The model adaptation training unit, connected to the data processing unit, the model adaptation training unit includes a prediction model library. According to the uploaded analysis result, select an adapted prediction model from the prediction model library; input the deviation angle and flow velocity information between the lead fish and the steel cable at the actual point into the prediction model, output the predicted flow velocity at the ideal point, and upload it to the power compensation model; The power compensation model is connected to the model adaptation training unit, and is used to receive the predicted flow velocity information and calculate the compensation power.

5. The channel flow measurement lead fish control method according to claim 4, characterized in that: The prediction model library includes models of linear regression, random forest, LSTM, and support vector regression. The historical database labels the corresponding tags for the historical data according to the model selection conditions, and each type of model is trained with the historical data of the corresponding tags in the historical database. The model selection conditions are as follows: based on the fluctuation range of the offset angle and the change amplitude of the flow velocity spectrum characteristics, when the change amplitudes of both are < preset values and stable, select the linear regression model; when the change amplitude of one of them ≥ preset value and the other also shows abnormal changes synchronously, select the random forest model; when the analysis results show that the changes of each window have temporal regularity, select the LSTM model; when the linear regression model, random forest model, and LSTM model are not satisfied, select the support vector regression model.

6. The method for controlling a river flow measuring lead fish according to claim 5, characterized in that: According to the uploaded analysis results and in accordance with the model selection conditions, select an appropriate prediction model from the prediction model library as the model to be used for the current window. At the same time, divide the data within the current window into a front sub-interval and a back sub-interval according to time periods, and calculate the change amplitudes of the offset angle fluctuation range and the flow velocity spectrum characteristics for both the front sub-interval and the back sub-interval respectively, and calculate the mutation coefficients of the three for the front sub-interval and the back sub-interval according to the change amplitudes. , if the change amplitudes of both in the back sub-interval meet the corresponding conditions in the model selection conditions, then use the corresponding model for prediction in the next window; if the change amplitudes of both in the back sub-interval do not meet any of the model selection conditions, but the mutation coefficient is within the preset tolerance range, then trigger the pre-activated state of the corresponding model and use the corresponding model for prediction in the next window.

7. The method for controlling a river flow measuring lead fish according to claim 1, characterized in that: In S4, by inputting the compensation power parameters into the control system of the lead fish power module, the power module drives the propulsion device of the lead fish to generate thrust according to the instructions issued by the control system to adjust the position of the lead fish. At the same time, the angle sensor continuously collects the offset angle of the lead fish and feeds it back to the control system, and continuously compares the difference between the real-time angle and the zero value. When the offset angle decreases to within the preset error tolerance range and the stable holding time ≥ preset value, it is determined that the lead fish is stable at the ideal point, and the control system triggers the ADCP probe to start, measures the real-time flow velocity at the ideal point, and the measurement data is transmitted to the data processing unit through the communication module of the lead fish for storage. At the same time, the control system monitors the offset angle of the lead fish in real time. If the fluctuation value of the offset angle ≤ preset value, the record is confirmed to be valid; if an abnormality occurs, immediately readjust the power output until the stable condition is met and then trigger the ADCP probe to record again.

8. A control system for a river flow measurement lead fish, the control system for the river flow measurement lead fish is used to implement the river flow measurement lead fish control method according to any one of claims 1-7, and is characterized in that, It includes: A winch system, a lead fish system, and an information processing module. The winch system is used to move the lead fish directly above the depth point of the river to be detected, calculate the distance from the lead fish to the depth point of the river to be detected, the lowering size of the steel wire rope, and lower the lead fish. The lead fish system is used to collect the flow velocity information of the actual river depth point, the offset angle between the steel wire rope and the lead fish in real time, and perform power output according to the predicted compensation power uploaded by the information processing module, and determine whether the lead fish is stable at the depth point of the river to be detected. The information processing module is connected to the winch system and the lead fish system, and is used to receive the information of the flow velocity at the actual river depth point and the offset angle between the steel wire rope and the lead fish in real time, analyze the characteristics and correlation of the information, select and train a prediction model according to the analysis results, predict the flow velocity at the ideal point using the trained prediction model, and calculate the predicted compensation power at the ideal point according to the power compensation model.

Images