Fishway hydrodynamic and fish passing efficiency monitoring method and system based on multi-source fusion

Abstract

The application provides a fishway hydrodynamic and fish passing efficiency monitoring method and system based on multi-source fusion, and belongs to the technical field of water conservancy projects and ecological monitoring. The method comprises the following steps: arranging a hydrodynamic sensing array and a biological behavior monitoring array at key sections of the fishway; collecting flow velocity vectors, turbulence characteristics, fish swimming speed, body length and trajectory coordinates in real time; performing spatiotemporal alignment and gridding pretreatment on multi-source heterogeneous data; calculating a hydrodynamic suitability index to quantify the matching degree of the flow velocity vector and the fish swimming speed; constructing a fish passing efficiency evolution model to calculate real-time comprehensive fish passing efficiency in combination with the spatiotemporal density of fish schools and the suitability index; and according to the time-varying characteristics of the efficiency, the wake-up and sleep time lengths of the collection nodes are dynamically adjusted by using multi-objective optimization. The system comprises a data collection layer, an edge computing layer, a data transmission layer, a model analysis layer and an application display layer. The application realizes quantitative coupling evaluation of hydrodynamics and biological behavior, and significantly reduces the power consumption of the monitoring system.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy engineering and ecological monitoring technology, specifically relating to a method and system for monitoring fishway hydrodynamics and fish passage efficiency based on multi-source fusion. Background Technology

[0002] While providing benefits such as flood control, power generation, and water supply, water conservancy projects also disrupt the longitudinal connectivity of natural river channels, posing a serious threat to the survival and reproduction of migratory aquatic organisms. Fishways are crucial engineering facilities for mitigating this obstruction and restoring the longitudinal connectivity of rivers. The efficiency of fish passage in a fishway depends not only on the rationality of its hydraulic structure design but also on the degree of dynamic adaptation between the hydrodynamic environment and fish swimming behavior during operation. Therefore, simultaneous monitoring of hydrodynamic parameters and fish behavior within the fishway, and evaluating the fish passage effect based on their quantitative relationship, is fundamental to achieving the scientific operation and management of fishways.

[0003] Some studies have attempted to monitor fishway conditions. For example, Chinese patent application CN202410512156.1 proposes a smart fishway transmission optimization system and method. Its core is to adjust the server's receiving logic based on the data acquisition cycle and transmission signal to reduce data transmission energy consumption. However, this approach only focuses on adjusting the data transmission strategy and does not address the intrinsic relationship between the hydrodynamic environment and fish behavior, nor does it analyze the question of "under what water flow conditions is fish passage most efficient." When the acquisition cycle is set improperly, it is easy to have fish dormant during periods of fish activity and to acquire data at high frequencies during periods without fish, resulting in insufficient timeliness and effectiveness of the monitoring data.

[0004] Chinese patent document CN202511526387.9 discloses an intelligent monitoring system for fishway passage based on multi-sensor fusion. This system utilizes underwater cameras, sonar, and environmental parameter sensors, combined with deep learning algorithms, to achieve real-time fish detection and behavior recognition. While the system is comprehensive at the perception level, its data fusion remains limited to target detection and species identification, lacking a mathematical model that quantitatively couples hydrodynamic and biomechanical parameters. Although it can count the number of fish passing through, it lacks a dynamic efficiency evaluation mechanism based on hydrodynamic-biological behavior coupling, and cannot quantify the degree to which water flow promotes or inhibits fish passage. Furthermore, the sensing units lack the ability to adaptively adjust the sampling frequency based on the probability of fish passage, generating a large amount of redundant data and energy consumption during fish-free periods, and potentially losing crucial details due to insufficient sampling frequency when fish are densely passing through. Its early warning function mainly targets abnormal behavior or numbers, and cannot perform predictive maintenance and scheduling based on hydrodynamic trends.

[0005] In summary, existing technologies have not yet achieved deep integration of hydrodynamic parameters and fish behavior at the mathematical model level, nor have they established an intelligent monitoring and dynamic control system driven by real-time fish passage efficiency index. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a method and system that can deeply integrate multi-source data, quantify fish passage efficiency in real time, and adaptively adjust collection resources accordingly, addressing the technical problems existing in fish passage monitoring technology, such as the separation between hydrodynamic monitoring and biological behavior monitoring, the lack of a quantitative efficiency evaluation model for hydrodynamic-biological coupling, and the high energy consumption and low data validity caused by rigid collection strategies.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for monitoring fishway hydrodynamics and fish passage efficiency based on multi-source fusion includes the following steps: Step S1: Deploy hydrodynamic sensor arrays and biological behavior monitoring arrays along key sections of the fishway to construct a physical monitoring network for the fishway. Step S2: Real-time acquisition of multi-source heterogeneous data, including three-dimensional velocity components characterizing water flow characteristics, eddy current and Reynolds stress characterizing water turbulence characteristics, and fish swimming speed, fish body length and swimming trajectory coordinates characterizing biological response. Step S3 involves preprocessing the collected multi-source heterogeneous data, including data cleaning, outlier removal, and spatiotemporal synchronization alignment, mapping data of different frequencies and coordinate systems to a unified spatiotemporal grid. Step S4: Based on the principles of fluid mechanics and the fish bioenergetics model, calculate the hydrodynamic suitability index of each monitoring grid point in the fishway. The hydrodynamic suitability index is used to quantify the degree of matching between the current flow velocity vector and the swimming speed of the fish. Step S5: Construct a fish passage efficiency evolution model based on multi-source data fusion, and calculate the real-time comprehensive fish passage efficiency of the fishway within a set time period by combining the spatiotemporal distribution density of fish swarms and the hydrodynamic suitability index. Step S6: Based on the time-varying characteristics of the real-time comprehensive fish passage efficiency, use a multi-objective optimization algorithm to dynamically adjust the wake-up cycle and sleep duration of the acquisition node.

[0008] This invention also provides a multi-source fusion-based fishway hydrodynamics and fish passage efficiency monitoring system, used to implement the above-mentioned multi-source fusion-based fishway hydrodynamics and fish passage efficiency monitoring method, the system comprising: The data acquisition layer consists of intelligent sensing nodes deployed inside the fishway. Each node integrates water physics sensors and biological sensors and has local data caching capabilities. The edge computing layer is deployed in the control cabinet at the fishway site. It receives the raw data from the acquisition layer and performs data cleaning, feature compression, and preliminary behavior recognition. The data transmission layer adopts a redundant transmission method that combines fiber optic wired communication with 5G / LoRa wireless communication to transmit the data processed by the edge computing layer to the cloud server. The model analysis layer, deployed on a cloud server, includes a data fusion module, a hydrodynamic suitability calculation module, a fish passage efficiency evolution module, and an optimization decision module. The data fusion module completes spatiotemporal registration and multi-source feature-level fusion. The hydrodynamic suitability calculation module outputs a hydrodynamic suitability index. The fish passage efficiency evolution module calculates the real-time comprehensive fish passage efficiency. The optimization decision module generates node hibernation and wake-up strategies based on the time-varying characteristics of efficiency and issues control commands. The application presentation layer is used to display 3D flow fields, fish heat maps, fish passage efficiency curves, early warning information, and scheduling suggestions to operation and maintenance personnel.

[0009] Compared with the prior art, the beneficial effects of the present invention are: First, it achieves mathematical coupling between hydraulic and biological data. By using the hydrodynamic suitability index and the fish passage efficiency evolution model, it quantitatively correlates the velocity field, turbulence intensity, and fish swimming speed and trajectory, breaking through the limitations of existing technologies where the two types of data only remain at the perception and recognition level, and enabling the quantification of the actual impact of water flow conditions on fish passage.

[0010] Second, a dynamic fish passage efficiency evaluation mechanism has been established. Instead of relying solely on fish counts, it combines fish density distribution, water flow suitability, and upstream migration probability to output real-time comprehensive fish passage efficiency. This can more accurately reflect the actual operation of the fishway and provide a quantitative basis for fishway operation scheduling.

[0011] Third, it enables on-demand intelligent scheduling of data collection resources. Based on the time-varying characteristics of fish-catching efficiency, it automatically adjusts the sleep and wake-up cycles of sensor nodes through multi-objective optimization, significantly reducing system energy consumption while ensuring complete capture of key information. This solves the problems of existing fixed-cycle data collection strategies that waste energy during inefficient periods and may undersample during efficient periods.

[0012] Fourth, it enhances the robustness of the monitoring system in complex field environments. Through measures such as dynamic weighted fusion of acoustic and optical signals, eddy current back-calculation for degradation, and hardware timestamp alignment, it effectively addresses adverse conditions such as turbid water, bubble interference, and temporary target loss, ensuring monitoring continuity and data reliability.

[0013] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the main process of the monitoring method of the present invention; Figure 2 This is a schematic diagram of the hierarchical structure of the monitoring system of the present invention; Figure 3 A schematic diagram illustrating the principle of hardware sleep and wake-up control for data acquisition nodes; Figure 4 This is a data flow diagram for multi-source data fusion and fish passage efficiency analysis; Figure 5 This is a velocity contour map around the fish in an embodiment of the present invention; Figure 6 This is a hydrostatic pressure cloud diagram around the fish in an embodiment of the present invention; Figure 7 This is a k-cloud diagram of the turbulent kinetic energy around the fish in this embodiment of the invention; Figure 8 This is a contour map showing the spatial distribution of the fishway hydrodynamic suitability index (HSI) in an embodiment of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0016] This embodiment provides a multi-source fusion-based fishway hydrodynamic and fish passage efficiency monitoring system, the overall architecture of which is as follows: Figure 2 As shown, the system is divided into five layers from bottom to top: data acquisition layer, edge computing layer, data transmission layer, model analysis layer, and application presentation layer. The optimization decision-making module in the model analysis layer sends control commands to the edge computing layer via dashed arrows, forming a control loop from environmental perception to policy feedback.

[0017] Specifically, the data acquisition layer consists of intelligent sensing nodes deployed along key sections of the fishway. The fishway is divided into inlet, pool, resting, and outlet sections according to its hydraulic structure. The selection of monitoring sections follows these principles: the inlet section assesses fish entry intentions; the pool section monitors mainstream and backflow structures; the resting pool evaluates fish recovery; and the outlet section counts successful upstream migrations. A typical deployment involves setting up monitoring sections at the inlet, every 10 pools, resting pools, and outlet. Each monitoring section has at least three layers of nodes (upper, middle, and lower) in the vertical direction to accommodate the water layer preferences of different fish species. The hydrodynamic sensing array includes an Acoustic Doppler Current Profiler (ADCP) and a Three-Dimensional Point Current Meter (ADV). The former acquires a wide-range velocity profile, while the latter captures microscale turbulence details. The biological behavior monitoring array employs a complementary configuration of an underwater high-definition camera and a dual-frequency identification sonar. In clear water or under supplemental lighting conditions, the camera is the primary sensor, while in turbid water or at night, the sonar is the primary sensor. The node integrates environmental sensors for water temperature, turbidity, and dissolved oxygen, and is equipped with a waterproof power supply and a backup battery. The intelligent sensing node features a streamlined, biomimetic design with an anti-biofouling coating. Internally, it integrates an inertial measurement unit (IMU) that automatically triggers a displacement alarm when it detects displacement or tilt exceeding 10 degrees due to water flow impact. A miniature stepper motor-driven scraper ring is installed in front of the camera lens. When the lens transmittance monitoring sensor detects a decrease in transmittance exceeding 15%, the scraper automatically starts cleaning to ensure the quality of visual data acquisition.

[0018] The edge computing layer is deployed within a control cabinet at the fishway site. Edge computing nodes can utilize embedded computing modules such as NVIDIA Jetson Nano to run lightweight deep learning models. Each node has general-purpose input / output (GPIO) pin control capabilities and a high-precision hardware clock. Nodes connect to MOSFET switches via GPIO pins to control the power supply circuits for the ADCP and underwater camera. An infrared sensor and optocoupler form a trigger circuit, whose output is connected to the external interrupt pin of the edge computing node for low-power monitoring and rapid wake-up.

[0019] The data transmission layer uses single-mode fiber as the backbone network to ensure high bandwidth and interference-resistant transmission. In areas where fiber optic laying is difficult, communication is achieved with surface buoys via underwater acoustic modems, and then the buoys transmit the data back to the server via the 5G network, forming redundant transmission links to ensure reliable data upload to the cloud.

[0020] The model analysis layer runs on a cloud server and includes a built-in data fusion engine, hydrodynamic suitability calculation module, fish passage efficiency evolution module, and optimization decision-making module. The data fusion engine performs spatiotemporal registration and multi-source feature-level fusion, employing a Kalman filter algorithm for joint tracking of sonar and visual data. When insufficient lighting causes the confidence level of visual data to fall below a set threshold, the proportion of sonar data in the fusion weights is automatically increased. The weight adjustment formula is as follows: ,in This serves as an environmental compensation factor. The hydrodynamic suitability calculation module outputs a hydrodynamic suitability index based on fluid mechanics and bioenergetics models. The fish passage efficiency evolution module combines fish population distribution with suitability calculations to comprehensively analyze fish passage efficiency in real time. The optimization decision module generates node dormancy and wake-up strategies based on the time-varying characteristics of efficiency and sends control commands to the edge. The optimization decision module is also equipped with a self-learning unit, which uses a Long Short-Term Memory (LSTM) network to perform online rolling parameter correction on the fish passage efficiency evolution model based on historical fish passage efficiency data for different seasons and fish species, and generates an optimized operation strategy table based on the hydrological forecast for the next 24 hours.

[0021] The application presentation layer provides a visual interface for displaying 3D flow fields, fish heat maps, fish passage efficiency curves, early warning information, and intelligent scheduling suggestions.

[0022] Figure 1 The main flowchart of the fishway hydrodynamics and fish passage efficiency monitoring method based on multi-source fusion, executed by the above system, is shown, specifically including the following steps: Step S1: Construct a physical monitoring network for fish passages; Based on the fishway design drawings and historical flow field simulation data, the specific monitoring section locations for the inlet section, pool section, resting pool, and outlet section were determined, and nodes were deployed in three layers (upper, middle, and lower) vertically. ADCP and ADV were deployed at the edges of the main flow core area and the backflow area behind the baffle to obtain average velocity profiles and local turbulence characteristics. A hydrodynamic sensor array using a combination of acoustic Doppler current profilers and point current meters was deployed to capture the backflow structure and main flow velocity within the fishway pool. Underwater high-definition cameras and dual-frequency identification sonar were installed on the pool sidewalls or baffles to cover the main upstream channels of fish; cameras were used primarily in clear water areas, while sonar was used primarily in turbid water areas. The biological behavior monitoring array used a combination of dual-frequency identification sonar and underwater supplemental high-definition cameras, utilizing edge computing terminals to perform fish contour extraction and target tracking locally, uploading only feature vectors to the server. All sensors were connected to the edge computing nodes via waterproof cables, and a hybrid network of fiber optic backbone and 5G / LoRa wireless was constructed for data transmission, equipped with waterproof power supplies and backup batteries. according to Figure 3The control circuit shown connects the ADCP power supply line to the first MOSFET switch controlled by GPIO_12 and the camera power supply line to the second MOSFET switch controlled by GPIO_13. The optocoupler output of the infrared sensor is connected to the INT_0 interrupt pin. Each edge node has a built-in temperature-compensated crystal oscillator (TCXO) and uses a Precision Time Protocol (PTP) to achieve full network clock synchronization with the main control server, with clock deviation controlled within 1 microsecond. After the physical network is built, each sensor node is in standby mode, providing the hardware foundation for subsequent multi-source heterogeneous data acquisition. It also completes the pre-embedding of the control circuit, providing the physical conditions for achieving microsecond-level hardware sleep and wake-up control in step S6.

[0023] Step S2: Real-time acquisition of multi-source heterogeneous data; The hydrodynamic sensing unit acquires instantaneous three-dimensional velocity components of the water body in the longitudinal, lateral, and vertical directions at frequencies ranging from 25Hz to 100Hz. , , Simultaneously, the vorticity characterizing the water flow rotation trend and the Reynolds stress characterizing the turbulent shearing effect are calculated. The biological monitoring unit acquires underwater video streams and sonar echo images, and runs the YOLOv8 target detection algorithm through edge computing nodes to initially extract the bounding box, center coordinates, and body length of the fish. The fish swimming speed is calculated by analyzing displacement changes between video frames or the Doppler effect of the sonar, while continuously recording the fish's position sequence in three-dimensional space to generate motion trajectory coordinates. The method for determining the fish swimming speed vector includes: extracting the displacement change of the fish's center point through image sequence analysis, calculating the instantaneous velocity based on the acquisition time interval; and simultaneously calculating the instantaneous velocity based on the fish's body length. Using regression models Estimate the critical swimming speed for this fish species. Where, The critical swimming speed of the target fish species is expressed in m / s, representing the maximum sustainable swimming speed that the fish can maintain. This refers to the length of the fish, in meters (m). These are empirical coefficients, calibrated through annular water tank experiments; This is an empirical index, calibrated through an annular flume experiment. The critical swimming speed is used to determine whether the current flow rate exceeds the physiological limits of fish. In addition, environmental auxiliary sensors simultaneously record water turbidity (NTU value) and bubble content parameters. These environmental parameters will be used for dynamic adjustment of subsequent data fusion weights, but do not directly participate in the calculation of fish passage efficiency. The acquired multi-source data exhibits significant heterogeneity; hydrodynamic data consists of high-frequency scalar and vector fields, while biological data consists of relatively low-frequency image and sonar streams, and the coordinate systems are different, making direct joint calculation impossible. Standardization processing is required in step S3.

[0024] To suppress parameter extraction bias caused by turbid water and air bubbles in the field, after data acquisition and before entering step S3 preprocessing, dynamic weighted fusion of acoustic and optical data and filtering of abnormal water flow data are performed.

[0025] The logic of dynamic weighted fusion of sound and light is as follows: set a turbidity threshold. As one specific implementation method, The value can be 50 NTU. When the measured turbidity... At that time, if the visual data is deemed to have high confidence, the weights of the fish swimming speed and trajectory parameters obtained from visual recognition will be adjusted accordingly. Set to 0.8, sonar parameter weights Set to 0.2; when Or, if the turbidity change rate increases sharply, meaning the increase in turbidity per unit time exceeds a preset change rate, which can be set according to the on-site water environment, for example, set to an increase of 20 NTU per minute, and the visual data is deemed unreliable, the system will automatically adjust the settings. It dropped to 0.1. The improvement to 0.9 primarily relies on sonar trajectory extraction. The fused fish motion parameters are as follows: The position coordinates are weighted in the same way; where, For visual recognition of fish movement parameters, the fish swimming speed vector is calculated by target detection and tracking algorithms from image sequences captured by an underwater camera. Fish movement parameters identified by sonar are obtained by Doppler analysis of echo data collected by dual-frequency sonar to calculate the fish swimming velocity vector. Abnormal flow data filtering targets noise generated by ADCP in areas of high bubble density. The raw ADCP data includes the echo intensity correlation coefficient for each measurement point. Set the correlation coefficient threshold. For example, counting 64. When a certain measuring point If the velocity vector at a given point is below a certain threshold, and the gradient between the velocity vector at that point and the average velocity of adjacent grid points exceeds a preset multiple (e.g., 3 times), the measuring point is marked as a bubble noise point. Lagrange interpolation is then performed using the effective velocity values ​​of neighboring grids to generate replacement values, ensuring the continuity of the flow field data. The velocity data and fish movement data processed in the above way are used as inputs for subsequent fusion.

[0026] In addition, this method also includes an abnormal water flow interference identification step: by comparing historical normal hydrodynamic data with real-time acquired data, the isolated forest algorithm is used to detect whether there are abnormal eddies or instantaneous high pressure. If an abnormality is identified and the fish density is not zero at this time, the high-frequency acquisition mode is immediately triggered and the acquisition frequency is increased to 5 to 10 times the base frequency.

[0027] Step S3: Data spatiotemporal alignment and gridding preprocessing; First, data cleaning is performed, using... Isolated outliers in the flow velocity data are removed using criteria or interquartile range (IVR) methods. Image and sonar data are deblurred, denoised, and contrast-enhanced to filter out false targets caused by floating objects. Time synchronization is then performed, using the PTP protocol to align the data streams from distributed sensors to the same millisecond-level time base. For high-frequency flow velocity data, the average value within each frame's video exposure time window is taken to match different source data in the time dimension.

[0028] The specific hardware implementation of time synchronization is as follows: Figure 3 As shown in the lower section: When an edge computing node receives sensor data, it directly reads the value of the underlying hardware counter and latches this value into the header of the data packet to achieve hardware timestamp marking, avoiding time delays at the operating system layer. For the inherent response delays of different sensors, a delay lookup table is pre-stored in the firmware; when packaging data, a microsecond-level delay compensation value is obtained from the lookup table based on the current sensor operating mode to correct the timestamp.

[0029] In terms of spatial registration, a global three-dimensional physical coordinate system for the fishway is established. Using a pre-calibrated coordinate transformation matrix, the pixel coordinates in the camera images and the polar coordinates of the sonar are uniformly transformed to the physical coordinate system. The fishway space is discretized into 3D grid cells with a side length of 0.5 meters. Velocity measurement points and fish positions are mapped to their respective grid cells, generating a unified spatiotemporal grid dataset where each grid cell simultaneously contains hydrodynamic attributes (velocity vector, turbulence intensity) and biological attributes (presence of fish, fish swimming speed). Thus, the co-occurrence of aquatic environmental parameters and biological parameters is achieved within each grid cell, providing a standardized data structure for calculating the hydrodynamic suitability index in step S4.

[0030] Step S4: Calculate the hydrodynamic suitability index; The system uses an analysis window of 10 seconds to extract the average velocity vector at the center of each grid cell. and turbulence intensity Based on the fish length identified in step S2 The system queries the built-in fish swimming ability database to determine the optimal swimming speed range, critical swimming speed, and endurance swimming speed threshold for the target fish species under the current water temperature conditions. The database stores the critical swimming speed estimation formulas based on body length for each fish species. The hydrodynamic suitability index (HSI) is calculated using the following formula: ; In the formula: --time ,Location The hydrodynamic suitability index at a location ranges from 0 to 1, with a higher value indicating greater suitability. --time ,Location The vector of water flow velocity at a given location, in m / s; --time The observed swimming velocity vector of the target fish, relative to the fish's body length Related, unit is m / s; —The Euclidean norm, used to calculate the magnitude of the difference between the flow velocity vector and the swim velocity vector; —A parameter for fish tolerance to differences in flow velocity, expressed in m / s, which is related to the habits of fish species. The smaller the value, the more stringent the requirements for flow velocity matching. --time ,Location The turbulence intensity at the point is dimensionless. —The critical turbulence intensity threshold for the target fish species, which is dimensionless. If this value is exceeded, the fish will find it difficult to maintain a normal swimming posture. —Weighting coefficient, used to balance the contributions of velocity fitness and turbulence fitness, with a value range of 0.6 to 0.8.

[0031] Among them, turbulence intensity The calculation formula is: ; In the formula: —The fluctuating component of the longitudinal instantaneous velocity, that is, the difference between the longitudinal instantaneous velocity and the longitudinal average velocity, in m / s; —The fluctuating component of the transverse instantaneous velocity, that is, the difference between the transverse instantaneous velocity and the transverse average velocity, in m / s; —The fluctuating component of the vertical instantaneous velocity, that is, the difference between the vertical instantaneous velocity and the vertical average velocity, in m / s; —The average velocity modulus, which is the combined modulus of the average velocity components in three orthogonal directions, is expressed in m / s.

[0032] , The values ​​were pre-calibrated based on the target fish species through flume experiments. The experiments used a variable-slope circulating flume equipped with a particle image velocimetry (PIV) system to simultaneously measure the flow field. At least 30 healthy individuals of each fish species were selected for repeated experiments. Instantaneous flow field parameters were recorded when individuals exhibited sideslip, stopped swimming, or retreated behavior under various flow velocities and turbulent conditions. The parameter range was determined using a 95% confidence interval. Table 1 lists the recommended parameter values ​​for several typical fish species calibrated through experiments. When the calculated HSI value was below 0.3, the system automatically marked the area as an inefficient fish passage zone.

[0033] Table 1. Recommended values ​​for hydrodynamic suitability parameters for typical fish species; To further verify the feasibility of the calculation method for the hydrodynamic suitability index described in this step, this embodiment uses numerical simulation to calculate the flow field of the fishway and substitutes the flow field results into the HSI calculation formula to generate a spatial distribution contour map.

[0034] Flow field calculations were performed using ANSYS Workbench 2023 R2 and ANSYS Fluent 2023 R2. A two-dimensional steady-state calculation was used for the local flow around the fish body, and the SST k-ω model was selected for the turbulence model. The inlet was set as a velocity inlet, and the average incoming flow velocity at the fishway inlet section was set to 0.20 m / s. The calculation results output velocity modulus, static pressure, and turbulent kinetic energy k, which were used to identify the local acceleration region, low-velocity wake region, pressure change region, and turbulent disturbance region around the fish body.

[0035] Figure 5 The diagram shows the velocity contour around the fish's body. The fluid splits near the head, with localized velocity increases on both sides of the fish. A distinct low-velocity wake forms behind the fish, gradually recovering downstream. The velocity field illustrates the spatial differences in flow velocity at various locations, which is consistent with the HSI formula. Provide basic data. Figure 6 The static pressure cloud map around the fish shows that a relatively high-pressure zone is formed at the upstream end of the fish, while the pressure decreases on both sides and the tail region. The downstream pressure gradually recovers. This result corresponds to the flow splitting, acceleration and wake structure in the velocity cloud map, indicating that the numerical model can reflect the interaction between the fish and the surrounding flow field. Figure 7 The k-cloud diagram of turbulent kinetic energy around the fish is shown. The turbulent kinetic energy is lower in the upstream and distant areas, and increases significantly from the posterior edge of the fish to the downstream tail track, with a maximum value of approximately 4.14 × 10⁻⁶. - The result of 3 m² / s² indicates that the velocity fluctuations and flow disturbances in the wake region are stronger. This result is converted into the local turbulence intensity TI, which is then used as the turbulence fitness input in the HSI formula.

[0036] The HSI spatial distribution calculation selected three consecutive chambers, using a three-chamber vertical slit fishway model. Each chamber was 15 m long and 3 m wide, with a total calculation section length of 45 m. The average inflow velocity at the fishway inlet section was set to 0.20 m / s. The fishway mesh velocity and turbulence data exported from Fluent were used as input, employing the same grass carp parameters as the previous example (σ = 0.32 m / s). , Body length 0.5 m Substitute the values ​​into the HSI formula grid by grid for calculation. Figure 8The diagram shows the spatial distribution contour map of the hydrodynamic suitability index of the fish passage based on the grass carp example. In the diagram, HSI≥0.7 is marked as a high suitability zone, 0.3≤HSI<0.7 is marked as a transition zone, and HSI<0.3 is marked as an inefficient fish passage zone. The contour lines change with the flow field near the pool chamber, partition and vertical slit, which can intuitively identify areas that are conducive to the continuous swimming of grass carp as well as areas with large differences in flow velocity or turbulent disturbance.

[0037] The simulation results above, together with the aforementioned single-point example, demonstrate that the hydrodynamic suitability index calculation method described in this step is feasible and can achieve a coupled evaluation of fishway hydrodynamic conditions and fish swimming ability.

[0038] Based on the above formula, the HSI of each grid can be calculated to generate a three-dimensional distribution map of the hydrodynamic suitability of the entire fishway, dividing the area within the fishway into a highly suitable zone, a transition zone, and an inefficient fish passage zone.

[0039] Step S5: Construct a fish passage efficiency evolution model and calculate the real-time comprehensive fish passage efficiency; Based on the unified spatiotemporal grid data from step S3, the fish density within each grid is calculated. And the probability of fish choosing to continue swimming upstream. The state transition probability The transition probability matrix, determined by the Markov chain model, is composed of the fish swarm distribution at the previous time step and the hydrodynamic suitability index at the current time step, expressed as follows: ; In the formula: --time ,Location The probability of the fish group choosing to continue swimming upstream, ranging from 0 to 1; —The steepness factor of the curve reflects the sensitivity of fish to changes in environmental adaptability. The larger Even a tiny decrease can lead to a sharp drop in the probability of tracing back; — The half-saturation suitability coefficient, which is the characteristic suitability value when the probability is 0.5, preferably 0.5.

[0040] To further improve adaptability to different fish species, when the camera identifies the main fish passing by as a specific species, it can... Adjustments are made. Taking grass carp as an example, its current-following preference coefficient is introduced. The corrected probability of uptracing is: ; In the formula: —Modified timing for specific fish species ,Location The probability of the fish group choosing to continue swimming upstream, ranging from 0 to 1; —The general state transition probability calculated from the above formula; —The flow direction preference coefficient specific to a particular fish species is an empirical constant; —The angle between the direction of water flow and the axis of the fish's body, in degrees or radians; — Angle The value of the sine function. When The probability increases when approaching 180 degrees (i.e., the peak flow).

[0041] Set time period (For example, 1 hour) serves as the statistical window to comprehensively analyze fish passage efficiency in real time. Defined as the ratio of the effective upstream passage volume to the number of fish entering the fishway entrance during this period, it is expressed in the following continuous form: ; In the formula: --time The real-time comprehensive fish passage efficiency is dimensionless and is quantified by integral form as the ratio of the fish passage volume to the total entry volume under suitable hydrodynamic conditions per unit time. —Fish density distribution function, representing time interval. ,Location Fish density at the location; —Total length of the fishway, in meters; — Statistical time window length, in seconds; —Fish entry rate at the fishway entrance, the number of fish entering the fishway entrance per unit time; — Correction factor for side passageways or resting pools; —The number of side passages; ——No. Weighting factors for each side channel; —Entering the The probability of a side passage; Entering the The rate of change of the probability of each side channel over time.

[0042] Under normal operating conditions, the fish density in this formula and state transition probability The real-time integrated fish passage efficiency is determined directly from the real-time measurement results of combined visual and sonar tracking, and then substituted into the above formula to calculate the overall fish passage efficiency. When both visual and sonar data fail to capture a valid target trajectory for multiple consecutive frames (e.g., N=30 frames, approximately 1 second), making it impossible to obtain the fish density through direct measurement, the system triggers a degradation fallback mechanism. The degradation mode utilizes the principle that fish movement locally alters the vorticity of the flow field, calculating the local vorticity of the current grid. Average vorticity of background flow field deviation Then, based on the mapping relationship established in advance through the water tank experiment... By deducing the possible estimated density of the lingering fish population, This is used to replace the direct density value in the above formula, ensuring that the fish-passing efficiency can still be continuously output even when the target is briefly lost.

[0043] Figure 4 The complete data stream for the above multi-source data fusion and fish passage efficiency analysis is shown: ADCP data and echo intensity correlation coefficients from the data source layer on the left. The three-dimensional velocity vector field, turbulence characteristics and Reynolds stress, and fish swimming velocity / trajectory coordinates are preprocessed from hydrodynamic data and biological behavior data, respectively, and then converged into a feature-level multi-source data fusion engine to achieve spatiotemporal gridding. These data are then input into the hydrodynamic suitability index calculation model and the fish swarm movement continuity and fish passage probability evolution model, respectively, and finally output the real-time comprehensive fish passage efficiency. The data is then fed into the multi-objective optimization and dormancy decision module, forming the feedback loop shown by the dashed line.

[0044] Step S6: Dynamic sleep and wake-up resource scheduling.

[0045] Real-time calculation of fish passage efficiency output in step S5 gradient of change The short-term variance is used to characterize the activity level and fluctuation of fish activity. A multi-objective optimization cost function is constructed. By minimizing Determine the optimal hibernation period : ; In the formula: —The cost function aims to minimize ; —The length of the sleep period set by the system, in seconds; —The system during its sleep period Energy consumption function within; —Energy consumption weighting coefficient, used to adjust the proportion of energy consumption in the cost function; —Information loss weighting coefficient, used to adjust the proportion of information loss terms in the cost function; --time Fishing efficiency; —The absolute value of the rate of change of fish passage efficiency over time; —Indicator function, when the fish crossing efficiency Exceeding the preset key threshold The value is 1 if the condition is met, and 0 otherwise. —Preset key threshold for fish passage efficiency; —The current moment; — The integral variable represents time.

[0046] Fish passage efficiency When the price remains at a high level or is in a rapid upward phase, increase This optimization process tends to shorten sleep duration or even enable full-time data collection to prevent information loss. This is especially important during late nights or dry seasons. When it remains at a low and stable level for an extended period, it increases. The relative weights of the optimization results result in a longer sleep duration, thus reducing the overall power consumption of the system.

[0047] To further enhance foresight, the optimized decision-making module also incorporates a Long Short-Term Memory (LSTM) prediction unit. This unit utilizes historical fish passage efficiency sequences and hydrological environment forecast data to predict fish passage efficiency over a future period. Trend. When the prediction results show... When a significant increase is expected in the short term, even if the current efficiency is low, the weight of the information loss term in the cost function will be automatically adjusted, the dormancy window will be shortened or the system will not be dormant for the time being, to ensure that the system is in full-efficiency collection state when the fish arrive.

[0048] When the optimization module determines that deep sleep is required, the edge computing node executes the following... Figure 3 The hardware control flow is as follows: The CPU writes a low level to the GPIO_12 and GPIO_13 registers, driving the corresponding MOSFET switches to turn off, physically cutting off the power supply circuits for the ADCP and camera, reducing their power consumption to near zero. When the preset wake-up time is reached or the infrared sensor detects fish activity, the INT_0 pin generates a rising edge interrupt signal, triggering the hardware interrupt service routine. The interrupt service routine pulls GPIO_12 and GPIO_13 high within microseconds, turning on the MOSFETs and powering the sensor back on. The edge node loads the pre-stored fast startup configuration script, skipping some self-test steps, completing a warm start and resuming data acquisition within milliseconds.

[0049] When encountering abnormal operating conditions such as a sharp increase in inflow during the flood season, the system can automatically enter emergency mode. ADCP monitoring detected that the flow velocity in a certain chamber momentarily exceeded the design value, and the vortex volume increased dramatically. Simultaneously, the vision system detected fish lingering at the inlet, a sharp drop in entry rate, and fish already in the fishway exhibiting bottom-hugging swimming behavior. Formula calculations show that, due to the increased difference in flow velocity, The value is generally below 0.1 throughout the entire river section. At this time, the system does not trigger hibernation, but instead increases the sampling frequency to 10Hz and outputs an alarm message, indicating that the current hydrodynamic conditions have seriously affected fish passage. At the same time, it automatically generates scheduling suggestions, suggesting that the dam's flood discharge gate be appropriately adjusted to improve the flow pattern at the fish passage inlet.

[0050] The following example demonstrates the online monitoring of a vertical slotted fishway using the above method and system, illustrating the calculation process of each step more clearly.

[0051] The fishway is approximately 600 meters long, with a designed water level difference of 25 meters. Monitoring nodes are deployed at the inlet, every 10 compartments, the resting pool, and the outlet. Assume that at a certain monitoring moment, at a grid point within the fishway (denoted as position x), the ADCP measures the water flow velocity vector. m / s, and simultaneously, video tracking yielded the swimming velocity vector of a grass carp with a body length of 0.5m at that point. According to Table 1, grass carp... Take 0.32 m / s, Take 0.12, The value is set to 0.7; the turbulence intensity at this grid point is also measured. This exceeds the critical threshold for grass carp. Substituting... Formula calculation: The modulus of the velocity difference is: .

[0052] The index term is: .

[0053] The turbulence term is: .

[0054] but .

[0055] This indicates that although the flow velocity difference at this grid point is acceptable, the turbulence has exceeded the tolerance range of grass carp, resulting in a low overall suitability.

[0056] After one hour of continuous statistics, the system recorded a total of 150 fish entering the system. The numerator obtained by integrating the formula is 72 (equivalent to the number of fish that successfully passed through), which represents the real-time overall fish passage efficiency for that hour. Meanwhile, the system detected a significant increase in fish dwell time in the resting pool area. Combined with the HSI distribution map, it was determined that the high turbulence in the main flow area forced fish to utilize the resting pool more. It was recommended that operators adjust the outflow rate or regulate the fishway baffle opening to improve the main flow pattern. This is especially important during late-night hours. To maintain a low and stable level, the system optimizes the cost function through multi-objective optimization. The optimized output has a longer sleep time, with only the entrance infrared sensor remaining on duty. Once fish activity is detected, the entire system is woken up via an interrupt signal to enter high-frequency acquisition mode.

[0057] The above examples verify that the method and system proposed in this invention can quantitatively couple water flow conditions with fish behavior, and provide quantifiable efficiency evaluations and operational recommendations accordingly.

[0058] Finally, it should be noted that although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, those skilled in the art can make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring fishway hydrodynamics and fish passage efficiency based on multi-source fusion, characterized in that, Includes the following steps: Step S1: Deploy hydrodynamic sensor arrays and biological behavior monitoring arrays along key sections of the fishway to construct a physical monitoring network for the fishway. Step S2: Real-time acquisition of multi-source heterogeneous data, including three-dimensional velocity components characterizing water flow characteristics, eddy current and Reynolds stress characterizing water turbulence characteristics, and fish swimming speed, fish body length and swimming trajectory coordinates characterizing biological response. Step S3 involves preprocessing the collected multi-source heterogeneous data, including data cleaning, outlier removal, and spatiotemporal synchronization alignment, mapping data of different frequencies and coordinate systems to a unified spatiotemporal grid. Step S4: Based on the principles of fluid mechanics and the bioenergetics model of fish, calculate the hydrodynamic suitability index of each monitoring grid point in the fishway. The hydrodynamic suitability index is used to quantify the degree of matching between the current flow velocity vector and the swimming speed of the fish. Step S5: Construct a fish passage efficiency evolution model based on multi-source data fusion, and calculate the real-time comprehensive fish passage efficiency of the fishway within a set time period by combining the spatiotemporal distribution density of fish swarms and the hydrodynamic suitability index. Step S6: Based on the time-varying characteristics of the real-time comprehensive fish passage efficiency, use a multi-objective optimization algorithm to dynamically adjust the wake-up cycle and sleep duration of the acquisition node.

2. The method according to claim 1, characterized in that, The turbulence characteristics mentioned in step S2 include turbulence intensity. The turbulence intensity The calculation formula is: ； In the formula, For a moment ,Location The turbulence intensity at the point is dimensionless. This is the fluctuating component of the longitudinal instantaneous velocity, which is the difference between the longitudinal instantaneous velocity and the longitudinal average velocity, in m / s. It represents the fluctuating component of the transverse instantaneous velocity, that is, the difference between the transverse instantaneous velocity and the transverse average velocity, in m / s; The vertical instantaneous velocity is the fluctuating component, which is the difference between the vertical instantaneous velocity and the vertical average velocity, and the unit is m / s. The mean velocity modulus is the combined modulus of the mean velocity components in the three orthogonal directions, and its unit is m / s.

3. The method according to claim 2, characterized in that, The formula for calculating the hydrodynamic suitability index (HSI) in step S4 is as follows: ； In the formula, For a moment ,Location The hydrodynamic suitability index at a location ranges from 0 to 1, with a higher value indicating greater suitability. For a moment ,Location The vector of water flow velocity at a given location, in m / s; For a moment The observed swimming velocity vector of the target fish, relative to the fish's body length Related, unit is m / s; It is the Euclidean norm, used to calculate the magnitude of the difference between the flow velocity vector and the swim velocity vector; This is a parameter representing the tolerance of fish to differences in flow velocity, expressed in m / s. It is related to the habits of fish species, and the smaller the value, the more stringent the requirements for flow velocity matching. The critical turbulence intensity threshold for the target fish species is dimensionless; beyond this value, fish cannot maintain a normal swimming posture. This is a weighting coefficient used to balance the contributions of velocity fitness and turbulence fitness, with a value ranging from 0.6 to 0.

8.

4. The method according to claim 3, characterized in that, The probability of the fish choosing to continue swimming upstream in step S5. The transition probability matrix, determined by the Markov chain model, is composed of the fish swarm distribution at the previous time step and the hydrodynamic suitability index at the current time step, expressed as follows: ； In the formula, For a moment ,Location The probability of the fish group choosing to continue swimming upstream, ranging from 0 to 1; The steepness factor of the curve reflects the sensitivity of fish to changes in environmental adaptability. The larger Even a tiny decrease can lead to a sharp drop in the probability of tracing back; This is the half-saturation fitness coefficient, which is the characteristic fitness value when the probability is 0.

5.

5. The method according to claim 4, characterized in that, The real-time comprehensive fish passage efficiency mentioned in step S5 The calculation formula is: ； In the formula, For a moment Real-time comprehensive fish passage efficiency, dimensionless; Let be the fish density distribution function, and represent time . ,Location Fish density at the location; This represents the total length of the fishway, in meters (m). The statistical time window length is expressed in seconds. The fish entry rate at the fishway entrance is the number of fish entering the fishway entrance per unit time. This is a correction factor for the side passageway or resting pool for fish. This refers to the number of side passages; For the first Weighting factors for each side channel; To enter the The probability of a side passage; To enter the The rate of change of the probability of each side channel over time.

6. The method according to claim 5, characterized in that, The dynamic adjustment of the wake-up cycle and sleep duration of the acquisition node in step S6 is achieved by minimizing the following cost function. Determine the optimal hibernation period : ； In the formula, Let be the cost function, and the objective is to minimize . ; The length of the sleep period set for the system, in seconds; For the system during sleep period Energy consumption function within; This is the energy consumption weighting coefficient, used to adjust the proportion of the energy consumption term in the cost function; This is the information loss weighting coefficient, used to adjust the proportion of the information loss term in the cost function; For a moment Fishing efficiency; It represents the absolute value of the rate of change of fish passage efficiency over time; As an indicator function, when the fish crossing efficiency Exceeding the preset key threshold The value is 1 if the condition is met, and 0 otherwise. The preset key threshold for fish passage efficiency; The current moment; Let be the integral variable, representing time.

7. The method according to claim 1, characterized in that, The hydrodynamic sensing array described in step S1 uses a combination of acoustic Doppler current profiler and point current meter to capture the structure of the reflux zone and the flow velocity of the main flow zone in the fishway pool. The biological behavior monitoring array uses a combination of dual-frequency identification sonar and underwater supplemental lighting high-definition camera. It uses an edge computing terminal to complete the fish contour extraction and target tracking locally, and only uploads the feature vector to the server.

8. The method according to claim 1, characterized in that, The method for determining the fish's swimming velocity vector in step S2 includes: extracting the displacement change of the fish's center point through image sequence analysis, and calculating the instantaneous velocity by combining the acquisition time interval; simultaneously, based on the fish's length... Using regression models Estimate the critical swimming speed of the fish species; among which, The critical swimming speed of the target fish species is expressed in m / s, representing the maximum sustainable swimming speed that the fish can maintain. This is an empirical coefficient; It is an experience index.

9. The method according to claim 1, characterized in that, It also includes an abnormal water flow interference identification step: by comparing historical normal hydrodynamic data with real-time acquired data, the isolated forest algorithm is used to detect whether there are abnormal eddies or instantaneous high pressure. If an abnormality is identified and the fish density is not zero at this time, the high-frequency acquisition mode is immediately triggered and the acquisition frequency is increased to 5 to 10 times the base frequency.

10. A fishway hydrodynamics and fish passage efficiency monitoring system based on multi-source fusion, characterized in that, The system is used to implement the multi-source fusion-based method for monitoring fishway hydrodynamics and fish passage efficiency as described in any one of claims 1 to 9, the system comprising: The data acquisition layer consists of intelligent sensing nodes deployed inside the fishway. Each node integrates water physics sensors and biological sensors and has local data caching capabilities. The edge computing layer is deployed in the control cabinet at the fishway site. It receives the raw data from the acquisition layer and performs data cleaning, feature compression, and preliminary behavior recognition. The data transmission layer adopts a redundant transmission method that combines fiber optic wired communication with 5G / LoRa wireless communication to transmit the data processed by the edge computing layer to the cloud server. The model analysis layer, deployed on a cloud server, includes a data fusion module, a hydrodynamic suitability calculation module, a fish passage efficiency evolution module, and an optimization decision module. The data fusion module completes spatiotemporal registration and multi-source feature-level fusion. The hydrodynamic suitability calculation module outputs a hydrodynamic suitability index. The fish passage efficiency evolution module calculates the real-time comprehensive fish passage efficiency. The optimization decision module generates node hibernation and wake-up strategies based on the time-varying characteristics of efficiency and issues control commands. The application presentation layer is used to display 3D flow fields, fish heat maps, fish passage efficiency curves, early warning information, and scheduling suggestions to operation and maintenance personnel.

Images