Hydrometric station flow measurement method and system

Abstract

The invention discloses a hydrometric station flow measurement method and system. Flow calculation model parameters are dynamically regulated and controlled based on real-time water level changes so as to compensate the influence of water level fluctuation on flow measurement precision. The sampling point layout is optimized according to the flow velocity distribution characteristics, and sampling data are collected and corrected; the sampling frequency is intelligently adjusted according to the suspended sediment concentration spatial-temporal variation characteristic so as to improve the sediment transportation amount estimation precision; and the ultrasonic or ADCP measurement speed is compensated and regulated according to the temperature and air pressure change so as to reduce the system deviation caused by environmental factors. A total-factor dynamic compensation system is constructed, the limitation of a traditional static mode is broken through through self-adaptive adjustment of a model and a sampling strategy driven by parameters such as the water level, the flow speed and suspended sediment, and the complex water regimen is accurately adapted. And temperature and pressure compensation and equipment self-correction are fused, so that the environmental robustness is enhanced, and stable operation under wide-range variation is ensured. Fine flow prediction is achieved through an intelligent algorithm, a monitoring-analysis-response closed loop is formed, and the initiative and accuracy of drainage basin water safety prevention and control are improved.

Description

Technical Field

[0001] This invention relates to the field of hydrological monitoring technology, specifically to methods and systems for measuring flow at hydrological stations. Background Technology

[0002] The hydrological station flow measurement method and system is a technical system that comprehensively utilizes modern measurement technology and data processing algorithms to accurately monitor key hydrological parameters such as river flow, velocity, and sediment transport. However, this method still faces many challenges: First, the accuracy of the flow calculation model decreases due to real-time fluctuations in water level, requiring research on how to dynamically adjust model parameters to adapt to changing hydrological conditions; second, the data representativeness is insufficient due to unreasonable sampling point layout, requiring optimization of the layout to better reflect the velocity distribution characteristics; third, the spatiotemporal variation of suspended sediment concentration affects the sediment transport estimation error, requiring intelligent adjustment of the sampling frequency based on its characteristics; fourth, environmental factors (such as temperature and air pressure) can introduce system biases into ultrasonic or ADCP velocimetry equipment, requiring the design of compensation and control mechanisms to address this; and fifth, the flow measurement lag problem during sudden floods necessitates prediction of short-term flow trends based on historical data patterns to improve the system's ability to cope with extreme hydrological events. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a hydrological station flow measurement method and system that features high measurement accuracy, intelligent control, and a compensation mechanism.

[0004] This invention provides a method for measuring flow at a hydrological station, comprising: The parameters of the flow calculation model are dynamically adjusted based on real-time water level changes to compensate for the impact of water level fluctuations on flow measurement accuracy. Optimize the layout of sampling points based on flow velocity distribution characteristics and collect and correct the sampling data; Intelligent adjustment of sampling frequency based on the spatiotemporal variation characteristics of suspended sediment concentration to improve the accuracy of sediment transport estimation; The speed of ultrasonic or ADCP measurements is compensated and adjusted according to changes in temperature and air pressure to reduce system bias caused by environmental factors.

[0005] According to one embodiment, the parameters of the flow rate calculation model based on real-time water level changes further include: Obtain the real-time water level value h; The adjustment coefficients for the flow calculation model parameters are determined using the following formula. ,in It is the historical average water level value, and α is the regulation rate constant; An alarm is triggered when K < 8, indicating that the model parameters need to be recalibrated. If the water level fluctuation amplitude Δh > 2m, then reduce the model parameter sensitivity threshold to half of the initial value.

[0006] According to one embodiment, the optimized sampling point layout further includes: Analysis of historical flow velocity distribution characteristics reveals the mainstream and boundary regions; Define the parameters for dividing the main flow region and the boundary as R = max(v) / median(v), where v is the velocity distribution; When R>5, increase the density of sampling points in the mainstream region; For each sub-region i, the area si and the corresponding weight wi satisfy si×wi≤10% of the total area to ensure global balance.

[0007] According to one embodiment, the intelligent adjustment of the sampling frequency further includes: The spatiotemporal coefficient of variation CV = σ / μ was obtained from historical suspended sediment concentration records, where σ is the standard deviation and μ is the mean. The sampling frequency will be doubled when CV > 3. The new optimal sample size N is calculated using Bayesian estimation: N = [log(PriorErr)log(PostErr)] × γ, where γ is the learning rate. For the interval where the sediment content Cj exceeds the threshold of 30 mg / L, a priority encrypted sample extraction strategy is adopted.

[0008] According to one embodiment, the speed compensation control for ultrasonic measurement further includes: Real-time monitoring of temperature T℃ and air pressure P (hPa) is used to calculate the compensation ratio factor through the environmental function En=βT+γP; When the En value deviates from the set range [10, +10], the measurement result Vadj is corrected as Vadj = Vo / (1 + θEn), where θ is the negative correlation adjustment coefficient. Adding a speed error prediction unit to ADCP-type equipment to provide early warning of system deviations; If more than 5 large deviations accumulate within an hour, a self-correction process will be triggered.

[0009] According to one embodiment, the step of predicting short-term traffic trends further includes: Retrieve 7 consecutive days of data from the historical database as the reference time series Dt; The target water volume for the next time period t+Δt is predicted using the exponential smoothing algorithm: Qt=αDt+(1-α)Dt1; where α is the attenuation coefficient, usually taking values ​​[2,6]. If, based on the actual water conditions, extreme precipitation is predicted to occur in the next two days with a probability P(t>2days)>95%, then an emergency correction item will be added to the estimated flow rate. Ensure that the response cycle of the entire forecasting system is controlled at the minute level to respond quickly to sudden flooding events.

[0010] According to one embodiment, the dynamic control process additionally includes a stability verification step: The cumulative model error Eacc is calculated every two hours. If Eacc ≥ threshold (set tolerance limit), the adjustment mode will be automatically restarted; The next control interval is adjusted using the recursive feedback control principle, where dt = dt × δ, and δ is the feedback strength correction factor. Keep total computing resources below the server performance load limit Load_max.

[0011] According to one embodiment, the optimization of the flow velocity distribution characteristics is further refined to a non-uniform sampling mode: For different partitions k, the sampling density dk∝|Ak / ΣA| is set, where Ak represents the effective flow area of ​​the k-th region; A fuzzy logic controller is introduced to determine whether to activate the secondary local precision measurement module; the condition is F(x,y)=1 if and only the current sampled data residual Set accuracy level And x and y belong to the critical domain; Ensure that the point cloud coverage rate in the key area is not lower than the specified baseline r0; To prevent any two adjacent partitions from having a sampling gap smaller than the preset minimum interval min_gap, the assumption of spatial independence is guaranteed.

[0012] According to one embodiment, the ultrasonic velocity correction mechanism is further improved by taking into account the influence of water temperature as follows: Set the temperature gradient dT / dL and evaluate the significance of the change along the propagation path; specifically, use the equation Δc~f(dT)L to express the relationship between the propagation velocity difference and the path length; f is determined by laboratory calibration; If ΔT crosses the critical value of layering Tthreash=5°C, then the depth subdivision processing method is activated to correct each layer segment separately. The multi-level correction results are integrated and a comprehensive velocity vector matrix is ​​generated for the final calculation of Vres=(∑WiVi) / (∑Wi), where Wi represents the weight allocation; Regularly update environmental parameters and calibrate the database to maintain the long-term effectiveness of the model.

[0013] In one embodiment, machine learning components were incorporated to assist in the process of more accurately assessing future traffic trends: First, select the core characteristic variable set X={xi,i≤n} that affects the flow; n≥5 dimensions are sufficient to depict a complete causal relationship picture. Training a sequence prediction model Ypred=g(X) based on a long short-term memory neural network; where the g function captures long-term memory association features, which is helpful for capturing special scenarios such as floods; Comparing the simulation results, the output probability distribution P(y|x) > λ, with a confidence interval λ generally chosen between 8 and 9 to ensure reliability; Incorporating this high-precision prediction scheme into the overall scheduling plan allows for proactive protective or resource deployment measures to reduce potential risks and losses.

[0014] According to one embodiment, the following new constraints are added to enhance the intelligence level of suspended sediment concentration collection frequency: Based on historical experience, a particle size distribution curve p(d) is constructed, where d is the particle diameter. The dominant transport scale d_major is identified and set as one of the basic decision indicators. When monitoring detects that d > 3 times the median within a specific time period, the frequency of temporarily increasing the detection intensity is triggered, f = f * χ, χ >= 2 times the base frequency; The overall quality delivery efficiency is evaluated using a mathematical expectation framework: Eff = sum(w_i*C(i)), where the weight of w_i depends on the importance of stage i; at the same time, it is ensured that Eff ≥ the pre-set optimal target. The final output is a complete and adjusted cycle task schedule for the execution end to refer to and execute, thereby improving overall efficiency and balance.

[0015] Another aspect of the present invention provides a hydrological station measurement system applicable to the hydrological station flow measurement method described above, the hydrological station measurement system comprising: The control unit is used to dynamically adjust the parameters of the flow calculation model based on real-time water level changes to compensate for the impact of water level fluctuations on flow measurement accuracy. The acquisition unit is used to optimize the layout of sampling points according to the flow velocity distribution characteristics and to acquire and correct the sampling data; The calculation unit is used to intelligently adjust the sampling frequency according to the spatiotemporal variation characteristics of suspended sediment concentration to improve the accuracy of sediment transport estimation. The compensation unit is used to compensate and adjust the ultrasonic or ADCP measurement speed according to changes in temperature and air pressure to reduce system deviations caused by environmental factors.

[0016] Beneficial Effects: This invention constructs a dynamic compensation system covering all elements including water level, flow velocity, suspended sediment, and environmental parameters. Through adaptive adjustment of flow model parameters driven by real-time water level fluctuations, intelligent sampling layout optimization based on flow velocity distribution characteristics, and dynamic switching of sampling strategies guided by spatiotemporal variations in suspended sediment concentration, it overcomes the limitations of static parameters and fixed sampling modes in traditional flow measurement methods. For complex scenarios such as sudden water level changes, uneven flow fields, and intense sediment transport, the system can automatically identify key areas and time periods, dynamically enhance data acquisition density and model calculation accuracy, and ensure that the measurement results of core parameters such as flow rate and sediment transport always closely match the actual evolution of hydrological conditions. This provides key technical support for upgrading hydrological monitoring from "fixed-time, fixed-point" to "real-time intelligent." By integrating a temperature and pressure compensation model with an equipment self-calibration mechanism, the system effectively eliminates the systematic interference of environmental variables such as temperature and air pressure on ultrasonic / ADCP measurements, ensuring stable operation of the monitoring equipment under wide environmental changes. Simultaneously, the introduction of recursive feedback control and dynamic allocation of computing resources enables the system to adjust the control cycle in real time based on data processing load and model errors, maintaining efficient and stable operation under high-concurrency data processing and extreme hydrological responses. This avoids the measurement deviation amplification or functional failure problems caused by drastic environmental changes or data overload in traditional systems, making it particularly suitable for long-term continuous monitoring in complex scenarios such as remote watersheds and unmanned stations, significantly reducing manual intervention costs and equipment maintenance pressure. By deeply integrating long short-term memory neural networks with traditional time series algorithms, the system achieves refined prediction of flow trends and early warning of extreme hydrological conditions. Based on historical data and real-time hydrological dynamics, the system can dynamically generate multi-timescale flow prediction schemes and automatically trigger emergency correction mechanisms for disaster scenarios such as extreme precipitation, providing minute-level response decision-making basis for flood control scheduling and water resource allocation. Combined with end-to-end stability testing and self-correction functions, the system effectively ensures the reliability of prediction results and the continuity of system operation, forming an integrated technology chain of "real-time monitoring—intelligent analysis—rapid response." This significantly enhances the initiative and accuracy of watershed water security prevention and control, providing a solid technical barrier for responding to extreme hydrological events under the background of climate change. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart illustrating the steps of one embodiment of the hydrological station flow measurement method of this application; Figure 2 This is a schematic diagram of the functional modules of one embodiment of the hydrological station flow measurement system of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0019] As attached Figure 1 As shown, the hydrological station flow measurement method of the present invention includes the following steps: S1. Dynamically adjust the parameters of the flow calculation model based on real-time water level changes to compensate for the impact of water level fluctuations on flow measurement accuracy. S2. Optimize the layout of sampling points based on the flow velocity distribution characteristics and collect and correct the sampling data; S3. Intelligently adjust the sampling frequency based on the spatiotemporal variation characteristics of suspended sediment concentration to improve the accuracy of sediment transport estimation. S4. The ultrasonic or ADCP measurement speed is compensated and adjusted according to changes in temperature and air pressure to reduce system deviations caused by environmental factors.

[0020] In one embodiment, the impact of water level fluctuations on flow measurement accuracy is compensated by dynamically adjusting the parameters of the flow calculation model based on real-time water level changes. Specifically, a series of sensors installed on the river cross-section acquire real-time water level values, which are then compared with predetermined historical averages or empirical functions. Once a deviation exceeding a threshold is detected (such as a sharp rise in water level during floods), the correlation coefficients in the flow equation, such as the Manning roughness coefficient and riverbed slope parameters, are immediately updated to make the newly generated results closer to actual conditions rather than a single value from the original fixed model. Taking a river as an example, if the river's flow is low under normal conditions, its model may use a lower roughness parameter setting; however, when torrential rains during the rainy season cause widespread rises, it needs to be adjusted to a higher roughness level to match the characteristics of turbulent flow, ensuring that even under extreme conditions, the output accurately reflects the true flow situation.

[0021] Secondly, optimizing the sampling point layout based on velocity distribution characteristics and collecting and correcting the sampling data involves using Doppler technology measurement equipment, such as the Acoustic Doppler Current Profiler (ADCP), to assess the magnitude and direction of fluid flow velocity at different depths within the target cross-sectional area. These readings can visually reveal which specific areas have higher or lower velocity ranges, allowing for a redesign of the monitoring network structure—increasing the deployment density near high-velocity areas while appropriately reducing repeated observations in slow transition areas. This not only saves energy but also maintains the quality of a sufficiently representative dataset. For example, assuming a wide river channel includes a main channel with relatively calm tributary waters on both sides, in the initial exploration phase, five nodes might be evenly deployed to cover the entire area; subsequent in-depth learning reveals a significant vortex phenomenon in the main channel, contributing approximately 90% of the overall effect. In this case, it can be decided to concentrate the four most sensitive areas in the first part, leaving the rest as background references, avoiding wasting resources on less significant areas. The final comprehensive statistical data should include clean values ​​that have undergone rigorous screening and correction to support a more reliable and stable sediment concentration calculation process in the next step.

[0022] Furthermore, regarding the third point, which considers the challenges posed by the large time span and spatial variations: how can we use known patterns to determine whether it's necessary to increase the sampling test interval frequency during certain key periods to achieve better estimation performance? The approach here primarily relies on mathematical modeling analysis to construct a predictive framework, combined with constraints based on physical and chemical principles to complete an automated decision-making mechanism. For example, by organizing and discovering common patterns or trend lines from a massive database of sample points accumulated over many years, a priori knowledge system is formed. This knowledge is then input into a machine learning platform, trained into a mature and usable version, and then fed back to the application site for real-time monitoring and operation guidance. For instance, in a northern seasonal reservoir project, it was found that every spring during the thawing season, large-scale mud clumps would detach and float away with the river, showing a significant increase in this process. The system would then suggest increasing the detection frequency within this interval to capture a sufficient number of event instances, providing detailed evidence for total calculation, rather than using a uniform standard throughout the year, which would result in significant information loss and a reduced accuracy score. This dynamic adaptive capability makes the entire solution more flexible and efficient, saving human resources while greatly improving responsiveness and handling efficiency at critical moments, avoiding the missed occurrence of any unexpected emergencies.

[0023] Finally, the discussion concerns the accurate representation of instruments at the hardware level under the influence of environmental factors. Specifically, it addresses the issue of temperature and air pressure fluctuations causing changes in the characteristics of the transmission medium, leading to error accumulation. This necessitates the introduction of external supplementary correction values ​​to eliminate adverse effects and interference, until the expected target requirements are met. Specific approaches typically include two important aspects: first, establishing detailed meteorological element collection and recording archives to promptly obtain the temperature and humidity conditions around the current work site; second, utilizing fundamental physics knowledge to develop formulas and algorithms to solve for the difference relationship, achieving automatic adjustment functions to meet high-standard requirements and maximize the value and efficiency of the expected output. For example, in modern water conservancy projects, ultrasonic detectors are frequently used as auxiliary tools for liquid level tracking and measurement. The core process is inevitably affected by changes in the sound transmission path due to the law of thermal expansion and contraction. Therefore, contingency plans must be prepared in advance, and appropriate gain or attenuation values ​​must be inserted as needed to balance the relationship between the two, preventing major errors that could damage the overall image and reputation.

[0024] The above details each step of the hydrological station flow measurement method and system, and points out the specific application methods and examples of each step in problem-solving, aiming to ensure accurate and efficient measurement and management in various complex real-world environments.

[0025] Next, the parameters of the flow calculation model based on real-time water level changes of the present invention will be further described. The entire process includes four main steps: obtaining real-time water level values, calculating the adjustment coefficient K and determining the alarm triggering conditions, analyzing water level fluctuations and adjusting the sensitivity threshold, and an example illustration.

[0026] Obtaining the real-time water level value h refers to directly collecting the current water level data from the hydrological station's flow measurement system or sensors. This step ensures that all subsequent operations are based on the most accurate hydrological conditions.

[0027] According to the formula The dynamic adjustment coefficient K used to adjust the parameters of the flow calculation model is calculated. Where h is the real-time water level value. This represents the historical average water level value, and α is a regulation rate constant, typically set as a positive real number with a recommended optimal value of 0.5-1.2. The meaning of this formula is based on the deviation... This formula expresses the magnitude of the change in the current water level relative to the historical benchmark and reflects it exponentially in the model calibration intensity. This formula is designed to quickly adapt to new conditions when hydrological status changes significantly, ensuring measurement accuracy.

[0028] When K < 8, an alarm is triggered to indicate that the model parameters need to be recalibrated. Setting a critical value of 8 is to account for the possibility that the model may deviate too much from reality when the K value is too low; therefore, this method provides an early warning so that manual checks or parameter optimization can be performed.

[0029] If the real-time water level fluctuation Δh > 1.2m, reduce the model parameter sensitivity threshold to half of the initial value. This design aims to address the potential for misjudgments caused by extreme water level changes, reduce over-adjustment issues caused by fluctuations within a small range, and thus maintain more stable operation.

[0030] For example, in one embodiment, suppose a hydrological station records hydrological data following a period of sustained high rainfall. Specifically, the system samples at h=15m, while the historical average is... And it is known that α = 1.0. At this point... Therefore, K = exp(1*5) = 148.4. Since this result is greater than 8, the system operates normally without issuing an alarm signal. However, a few days later, the rainfall suddenly intensified, causing the water level to rise. The fluctuation Δh exceeded the 1.2-meter limit, so the sensitivity threshold was reduced by half to prevent oversensitivity from affecting the final measurement results. In this way, by continuously and automatically correcting relevant parameters based on the water situation, the accuracy of flow estimation can be improved while ensuring the robustness of the system.

[0031] Next, the optimized sampling point layout of the present invention will be further described. This method includes several steps to improve the accuracy and efficiency of flow measurement at hydrological stations: First, historical velocity distribution characteristics are analyzed to determine the main flow zone and boundary zone; second, a partitioning parameter R = max(v) / median(v) is defined, and the quantitative difference between the main flow zone and the boundary zone is calculated using this formula; third, when the parameter R is greater than a specific threshold of 5, dense sampling points are added within the main flow zone; fourth, the product of the area si of each sub-region i and its weight wi does not exceed 10% of the total area, achieving globally balanced sampling.

[0032] The significance of the above steps lies in the rational allocation of sampling resources. The first step aims to identify high-interest areas (mainstream) and low-interest areas (boundary areas) in the flow characteristics. By analyzing historical velocity data, the behavioral characteristics of water flow at different locations can be discovered. For example, the mainstream generally exhibits larger and more uniform velocity variations, while the boundary area may tend to have lower values ​​or be highly volatile. The second step introduces the parameter R, which represents the ratio of the maximum velocity value max(v) in a certain area to the median(v) of all velocities in that area. The range of v is usually positive and depends on the actual scenario. The optimal threshold for R is set to be greater than 5 because this represents a significant degree of velocity dispersion, requiring more attention. In the third step, if R exceeds the set threshold, the sampling density is concentrated in the main flow area, which helps to capture highly dynamic data characteristics. The final step stipulates that the product of the area and weight of a single sub-region is limited to one-tenth of the total system area. This is to avoid some local areas excessively occupying resources and affecting global fairness.

[0033] In one embodiment, the application of these steps is specifically illustrated. For example, for a hydrological monitoring station along the Yangtze River, the average daily flow velocity at various points over many years is collected as historical baseline data for analysis. The results show that the flow velocity in the middle section of a certain river segment is generally higher than that at the edges, thus defining it as a clear main flow zone and boundary zone. Next, the R value is calculated to reach approximately 6.2 on a specific day (maximum flow velocity approximately 3 m / s, median value approximately 0.48 m / s), meeting the conditions for increased sampling density in this main flow zone. Furthermore, the water segment is divided into five smaller regions, each assigned a corresponding wi according to its flow proportion. The largest wi corresponds to the central region, and si multiplied by its relative importance does not exceed one-tenth of the total monitored area, thus achieving the goal of scientifically and accurately arranging the entire network.

[0034] Next, the intelligent adjustment of sampling frequency according to the present invention will be further described. This method specifically includes the following steps, one by one, to achieve the goal of optimizing the sampling strategy based on historical data and dynamic changes. The first step is to calculate the coefficient of variation (CV) by analyzing historical suspended sediment concentration records. The formula is defined as the ratio of the standard deviation σ to the mean μ, where σ characterizes the dispersion of the sample, and μ is the mean of the historical samples. The coefficient of variation (CV) is an important parameter for measuring the uncertainty of a random variable. If the CV exceeds a preset threshold (3 in this case), the sampling frequency should be doubled to accommodate the increased measurement accuracy requirements under greater variability.

[0035] The second step involves dynamically determining the optimal sample size N using Bayesian estimation. The core principle is to combine the prior error (PriorErr) and the posterior error (PostErr) and ablate the influence weights using a specific factor γ to derive the new sample size. The learning rate γ is recommended to be between 0 and 1, with an optimal value generally close to 0.8 to ensure a moderate adjustment that avoids overfitting or excessive sluggishness. This process embodies the core concept of stepwise derivation from the known to the more precise unknown, aiming to maximize the representativeness of the results with the minimum sampling size.

[0036] Next is the operation part of prioritizing the detection area. The third step clearly states that for any case where the sediment content Cj exceeds the established baseline (mentioned in this article as 30mg / L), localized and intensified collection of key control target point data will be carried out to ensure the overall distribution while paying attention to the coverage of extreme values.

[0037] For example, in one embodiment, it can be assumed that recent periodic observation data from a hydrological station shows that the historical average sediment concentration in a specific river section reaches 20 mg / L with significant fluctuations (e.g., a maximum of 150 mg / L). In this case, the system calculates a CV (volume shift) that is higher than a critical value, deciding to double the frequency to obtain more information. Then, it utilizes previously accumulated correlation training to derive a more suitable γ parameter value and recalculates the necessary total number of samples to better reflect the characteristics of that period. For special zones with excessive sediment concentration, a separate refinement scheme is initiated to increase the number of sampling intervals, achieving higher accuracy. This method comprehensively plans sampling behavior by combining historical background, current needs, and temporal characteristics, ultimately achieving automated and intelligent control and meeting the requirements of high-quality fluid dynamics monitoring.

[0038] Next, the speed compensation control of ultrasonic measurement according to the present invention will be described.

[0039] Real-time monitoring of temperature (T°C) and air pressure (P, hPa) is used to assess the impact of environmental changes. In this step, real-time temperature and pressure data are used as key variables to calculate the environmental function subsequently used for compensation. Environmental changes are quantified using the formula En = βT + γP, where T represents temperature (degrees Celsius), P represents air pressure (hPa), and β and γ are constants related to equipment performance, representing the set values ​​for the weighting of temperature and air pressure responses, respectively. Their optimal range is typically set to β = 0.15 / °C and γ = -0.02 / hPa. The design of these parameters ensures that the formula comprehensively captures the actual impact of the environment on ultrasonic transmission speed under different conditions.

[0040] Subsequently, the accuracy of velocity measurement is optimized by judging the deviation of the calculated result En and adjusting the compensation ratio factor. Specifically, if the calculated En value exceeds the range [10, +10], the actual adjusted measurement result is calculated based on the correction model Vadj = Vo / (1 + θEn). In this formula, Vo represents the uncompensated initial velocity reading, Vadj is the actual flow rate value obtained after compensation, and θ is the negative correlation adjustment coefficient (the optimal value is approximately 0.002 to 0.004). The purpose of this formula is to utilize the mathematical logic of negative correlation to reduce the adverse bias of excessively high or low En on the final measurement result.

[0041] Furthermore, a predictive unit is introduced for early warning in measurement devices using the ADCP model. This step aims to capture potential deviation trends during formal operation through internal error prediction functions and feed back corresponding information to the control module for rapid intervention. For example, in situations with high water flow velocity and sudden changes in the external environment, this predictive mechanism will detect potential large error signs first and respond promptly.

[0042] If frequent exceedances occur within a single statistical hour, an additional correction sequence is triggered. If the above predictive measures still fail to prevent large error accumulation from reaching the threshold condition—that is, more than five significant deviations are detected within an hour—the entire measurement system automatically enters a deeper overall self-calibration phase, including reassessing the baseline environmental parameters, comparing historical calibration records, and further optimizing the current compensation parameter settings.

[0043] In one embodiment, specifically after deploying such a system at a hydrological station, a heavy rainfall event caused the water temperature to plummet to 8°C while the atmospheric pressure rapidly increased to approximately 1026 hPa. Based on the aforementioned principle, the system measured the temperature and pressure values ​​in real time and calculated En ≈ -5.78. Since this was within the effective tolerance range, no immediate correction was needed. However, the prediction function detected that the expected increase in En over the next half hour could cause significant drift, thus triggering a warning. Subsequently, after three consecutive sampling cycles, it was confirmed that the cumulative deviation exceeded the tolerance limit, initiating a full verification process, and ultimately restoring the system to a stable state.

[0044] Next, the steps for predicting short-term flow trends according to the present invention are described. The steps for predicting short-term flow trends mainly include the following steps: retrieving data for 7 consecutive days from a historical database as a reference time series; using an exponential smoothing algorithm for prediction; adjusting the estimated value based on actual hydrological information and adding an emergency correction term; and ensuring that the response cycle is at the minute level.

[0045] Specifically, the system first retrieves seven consecutive days of data from a historical database as a reference time series. This step involves selecting recent historical flow data as a baseline input to capture the trend characteristics of the current hydrological situation. For example, in one embodiment, the system might retrieve hourly flow values ​​for a river over the past week from the database as the basis for predicting the target water volume for the next period.

[0046] Then, an exponential smoothing algorithm is applied for flow forecasting. The target water volume Qt = αDt + (1-α)Dt1 in the formula is used to estimate the target flow for a future period. The parameter α is the attenuation coefficient, typically ranging from [0.2, 0.6], and preferably 0.4. Here, α controls the weighting of new and old data; higher weights tend to be more sensitive to recent data, while lower weights refer more to historical data. Therefore, this formula can integrate the influence of current and historical data for smoothing to obtain more accurate flow forecast results.

[0047] Furthermore, the decision to add an emergency correction term should be based on the actual water situation. If the probability analysis of precipitation indicates that the probability of extreme precipitation in the next two days, P(t>2days), is greater than 95%, this indicates a high probability of a major rainfall event. Therefore, an emergency correction term needs to be added to the prediction model to compensate for sudden increases caused by large-scale rainfall that cannot be reflected in conventional predictions. In a specific example, if the system detects that a rainstorm is about to occur in a certain watershed, and its intensity and coverage exceed normal fluctuations, a fixed percentage or fixed range will be superimposed on the original flow prediction result to improve the accuracy of the warning water level estimation.

[0048] The final step is to ensure the system's response efficiency remains within an extremely short time span, i.e., a response rate on the order of minutes, to ensure that sudden flooding can be quickly monitored and warned, thereby reducing the negative impact of disaster risks. In practical operation, the system architecture must fully consider computational performance optimization design so that freshly collected data can immediately participate in the latest prediction cycle, while ensuring that overall operational stability and accuracy are simultaneously maintained.

[0049] Next, the dynamic control process of this invention includes an additional stability verification step: First, the key steps in the dynamic control are listed, including: calculating the cumulative model error Eacc every two hours; automatically restarting the control mode if Eacc ≥ threshold; using the recursive feedback control principle to adjust the next control interval dt = dt × δ; and ensuring that the total computing resources do not exceed the server performance load limit Load_max. Each step and its meaning are described in detail below.

[0050] The first step is to calculate the cumulative model error Eacc every two hours. In this stage, the cumulative model error Eacc refers to the result of accumulating all individual error terms since the last check according to a specific formula. This setting allows for periodic monitoring of whether the overall model's performance deviates from the initial target state. This two-hour interval is chosen because it is sufficient to reflect the cumulative effect over a period of time without consuming excessive system resources for real-time monitoring.

[0051] The second step is to trigger a reset mechanism if the accumulated error exceeds the tolerance threshold. The tolerance threshold is a preset value used to measure the maximum acceptable range of error. Its specific value is determined according to different application scenarios, and is usually set within a reasonable accuracy range (e.g., ±1.5%). When the condition is met, stability is restored by triggering the system to return to the default or optimized initial parameter values. This is a safeguard mechanism to prevent system failure caused by continuous bias.

[0052] The third step involves applying recursive feedback control to adjust the control interval. Specifically, the correction factor δ in the formula dt=dt×δ is used to dynamically change the time difference between two consecutive monitoring sessions. Here, dt represents the original time gap length, and δ is generally within the range of 0.9 to 1.1 to ensure that the rhythm is not drastically altered while allowing for fine-tuning to improve the fit. This definition helps to promptly shorten the assessment frequency after abnormal fluctuations are detected until the system returns to a stable trajectory, and then subsequently extend the time span again to save computational resources and maintain efficiency.

[0053] The final step is to ensure that the monitoring computational load level is below the server's capacity limit, Load_max. This constraint aims to balance the trade-off between accuracy and practicality, minimizing excessive hardware usage while maintaining accurate results, thereby extending the server's lifespan and reducing potential risks.

[0054] Specifically, in one embodiment, for example, for the flow measurement method at a hydrological station, the system collects sensor reading errors every two hours and compares them with historical average deviations to determine Eacc. If Eacc exceeds a previously calibrated threshold value, such as a cumulative water volume difference of 2 mm / hour, a comprehensive review and initial configuration reset are immediately performed to eliminate the possibility of long-term small deviations accumulating into large errors. Simultaneously, the original three-hour interval is reduced to 2.7 hours using a formula (dt=dt×δ=3×0.9) for more frequent monitoring until the overall situation is confirmed to be back to normal. Then, the interval is expanded again to an economically effective standard range. Throughout the entire process, the server workload is strictly controlled, always staying below the safe threshold to ensure that Load_max will not cause a system crash.

[0055] Next, the optimization of the flow velocity distribution characteristics of the present invention will be further refined to a non-uniform sampling mode.

[0056] For different zones k, a sampling density dk ∝ |Ak / ΣA| is set. This formula represents the allocation of sampling density based on the proportion of effective flow area within each zone. Here, Ak is the effective flow area of ​​the k-th zone, representing the actual movable area of ​​water flow within the zone; ΣA is the sum of the total effective flow areas of all zones. Allocating sampling points through this proportional relationship ensures that sampling resources are more inclined towards areas with larger water flow, thereby improving sampling efficiency and accuracy. This non-uniform setting optimizes resource allocation based on hydrological characteristics to reduce overall redundancy. Typically, ΣA should cover the effective area within all measurement areas, and when the number of areas is large, an approximate value can be calculated through numerical integration. Theoretically, a larger dk value indicates a higher importance for the zone, but it is necessary to balance the weight allocation between different zones to avoid excessive local bias.

[0057] A fuzzy logic controller is introduced to determine whether to activate the secondary local fine measurement module, under the condition that F(x,y)=1 if and only the current sampled data residual Set accuracy level Meanwhile, x and y belong to the critical region. Here, Sq represents the standard error of the existing sampled data. This is a predefined accuracy benchmark threshold (e.g., 0.05 m / s for a specific hydrological station scenario). Its range should be set in conjunction with measurement accuracy requirements; under complex conditions, a value between 0.03 and 0.07 m / s is recommended to balance accuracy and economy. The fuzzy controller determines whether to perform local fine-tuning remeasurement by evaluating the residuals and the spatial coordinates of the target point to compensate for deficiencies in the original sampling. This strategy ensures supplementary measurements are performed at locations with large errors or highly dynamic flow fields, avoiding unnecessary resource waste. In one embodiment, when a specific location on a critical monitoring section is detected to have an error Sq reaching 0.12 m / s due to local eddies (exceeding the set threshold of 0.07 m / s), a secondary fine-tuning mechanism is immediately triggered to improve data quality.

[0058] Furthermore, the point cloud coverage within key areas must be ensured to be no less than a specified baseline r0. Here, r0 specifies a percentage value (e.g., an 85% regional coverage standard) to guarantee sufficient measurement point density in the core flow field to support subsequent data analysis. This requirement aims to ensure basic data integrity is maintained even when local sampling intensities are low. If the actual coverage is lower than r0, the operating parameters of adjacent sampling units must be automatically adjusted to increase the measurement density until the standard is met. For example, in a river cross-section monitoring case, r0 is set to 90%. If the actual coverage in a key area is only 78%, a local additional measurement procedure will be initiated to meet the coverage requirement.

[0059] Simultaneously, it is necessary to ensure that the sampling gap between any two adjacent partitions is less than the preset minimum interval min_gap to avoid excessively tight spatial correlation that could undermine the statistical independence assumption. The specific value of min_gap depends on the performance of the specific instrument (e.g., 1 meter or more). If the interval is too small, it may lead to strong data correlation and loss of independence; if the interval is too large, it may fail to capture flow field details. In a specific hydrological station case, if the initial spacing between partitions A and B is set at 0.5 meters, it will be adjusted to meet the minimum spacing requirement of 1 meter before subsequent operations.

[0060] Through the above steps and their reasonable settings, the process of efficient and accurate water flow measurement under non-uniform conditions is optimized.

[0061] Next, the improved ultrasonic velocity correction mechanism of the present invention, taking into account the influence of water temperature, is described as follows: The process includes several steps. First, the operation and purpose of each step are clarified. Then, the content of each step is explained in detail, and further illustration is provided through a practical example.

[0062] The first step is to set the temperature gradient dT / dL and assess the significance of the change along the propagation path. In this process, the formula Δc ~ f(dT)L is used to quantify the relationship between the difference in propagation speed and the product of the path length, where f is a coefficient calibrated in the laboratory, Δc represents the deviation in ultrasonic propagation speed, dT is the temperature change between two points, and L is the propagation distance, typically ranging from a few meters to tens of meters. The formula shows that the effect of temperature change on the sound speed along the propagation path is linear. Since the temperature gradient directly affects the degree of change in sound wave propagation speed, the initial purpose of this formula is to accurately assess the error value caused by water temperature per unit length.

[0063] The second step involves initiating depth subdivision processing when ΔT crosses a specific stratification threshold, Tthreash = 5°C. The goal of this step is to implement multi-level refinement corrections for water bodies exhibiting significant temperature stratification. For example, when low-temperature density currents occur at the bottom of rivers or lakes, and the temperature difference between the top water temperature and the bottom stratified regions exceeds 5°C, the measurements for each temperature stratification segment must be corrected separately to ensure the final results more closely reflect reality.

[0064] The third step integrates the multi-level corrected data and generates a comprehensive velocity vector matrix for subsequent calculations using the formula Vres=(∑WiVi) / (∑Wi), where Wi represents the weighting coefficient and Vi represents the velocity vector value after each level of correction. This weighted allocation method optimizes the overall flow measurement accuracy. This weighting design is based on considerations of the reliability and importance of different regions; for example, data from shallow and deep water areas may have different meanings for total flow calculation, thus assigning them appropriate weights can better improve calculation accuracy.

[0065] Finally, the calibration database, which includes environmental parameters such as seasonal variations and geographical features, needs to be updated regularly to ensure the model operates efficiently over long periods. This calibration data will be continuously adjusted and improved over time to adapt to new natural conditions or hardware characteristics.

[0066] In one embodiment, a complete flow measurement system is deployed at the hydrological station to monitor the overall hydrological conditions of a wide river section. If on-site observations reveal that the surface water temperature is high (approximately 24°C), while the deep water temperature is only 18°C ​​(i.e., the temperature difference is greater than or equal to the stratification limit (5°C), the system immediately activates the multi-layer segmentation module. Specifically, the detection points are divided into two groups based on vertical stratification. Each stratum is then recalibrated using the aforementioned equation Δc~f(dT)L before data fusion is used to establish the final velocity distribution model. The overall water velocity is then estimated using a weighted average algorithm. The backend database is regularly maintained based on this type of on-site data collection to continuously improve and refine the logical framework, thereby achieving a more ideal long-term performance.

[0067] Next, the invention incorporates machine learning components to assist in the more accurate assessment of future flow trends: First, a set of core feature variables affecting flow, X = {xi, i ≤ n}, is selected, where n represents the number of core features and is required to be ≥ 5, to ensure that the features can fully characterize the complete causal relationship picture of the hydrological system. The core of this step lies in determining the set of input features, which are key factors closely related to river flow. These factors may include, but are not limited to, precipitation, upstream reservoir water level changes, evaporation rate, and historical data from different time windows in the time series.

[0068] Based on the selected set of core feature variables, a sequence prediction model Ypred=g(X) based on a Long Short-Term Memory (LSTM) neural network is constructed and trained. Here, g represents a nonlinear function mapping relationship captured by the LSTM model, which transforms the input sequence into an output sequence through a learning process involving multiple hidden layers, thereby effectively capturing long-term memory-related features. In this formula, Ypred refers to the target flow value predicted for future times. LSTM is chosen because this architecture is particularly suitable for handling time-dependent data streams, and can more accurately reflect the long-term evolution patterns in special scenarios such as floods.

[0069] By comparing the predicted results with the actual data, the confidence interval of the probability distribution P(y|x)>λ is output. Here, x is the known observed sample input dataset, and y is the corresponding unknown but estimated target variable value. The parameter λ is generally located in the range [0.8, 0.9] to provide a reasonable basis for judging reliability. When the model's output falls within a certain specified high confidence level, its predictive quality can be relied upon and thus used in the decision support process.

[0070] High-precision prediction mechanisms are integrated into the overall resource scheduling plan, enabling proactive implementation of defense strategies or optimized allocation of necessary response resources based on anticipated potential problems, effectively mitigating the negative impacts of natural disasters such as floods. For example, in one embodiment, a hydrological station can use the aforementioned method to detect a potentially abnormally high water flow peak in the near future, immediately issuing an early warning to surrounding communities and recommending the activation of additional drainage pumps to reduce the risk of damage to downstream residential areas. Specifically, in a simulation test experiment, a model trained using detailed meteorological records sampled every minute over the past ten years successfully predicted a significant increase in water volume dynamics in the 72 hours preceding a severe flood three months prior, demonstrating the considerable practical feasibility and potential for improvement of this algorithm system.

[0071] Next, the newly added constraints on the intelligent level of the enhanced suspended sediment concentration acquisition frequency of the present invention are described as follows: First, the steps are outlined. A particle size distribution curve is constructed based on historical experience, and the dominant transport scale is determined. For situations where particle size exceeds the normal threshold under specific conditions, the detection frequency is temporarily increased. The quality transfer efficiency is evaluated using a mathematical expectation framework to ensure the overall effect is not lower than the set target value. Finally, an adjusted periodic task scheduling plan is developed to optimize overall execution efficiency.

[0072] Specifically, a particle size distribution curve p(d) is established based on historical experience and data analysis, where d represents the particle diameter, ranging from μm (e.g., 0.1 μm to 1000 μm). Analyzing this distribution curve identifies the dominant transport scale ddominant, which serves as one of the fundamental decision-making indicators to determine the contribution of the main particle size range to sediment transport, thereby ensuring that the sampling process focuses on key scale regions. In one embodiment, if a hydrological station finds that the main particle distribution of a sediment sample is concentrated in the 50-80 μm range, this is designated as the dominant transport scale ddominant.

[0073] When monitoring detects that particle size exceeds 3 times the median within a specific time period, an operation frequency adjustment mechanism is triggered. Specifically, f = f * χ, where f is the base detection frequency and χ is the amplification factor, which must satisfy χ ≥ 2, indicating that the frequency needs to be increased to twice or more than the base frequency to more accurately capture dynamic changes under abnormal conditions. This setting is based on a rapid response mechanism, increasing the sampling intensity to ensure data integrity when particulate matter concentration is abnormal or water quality conditions change abruptly. For example, a sudden increase in river flow may carry large-diameter particles; immediately increasing the sampling rate helps to reflect the real-time state of sediment movement in the river channel.

[0074] The overall quality transfer efficiency is evaluated using a mathematical expectation framework. Eff represents the comprehensive efficiency value, calculated as Eff = sum(w_i * C(i)), where w_i is the weight value for the i-th stage, typically set within the range [0,1], and C(i) corresponds to the actual sediment concentration (g / L or kg / m³) at different detection stages. By rationally allocating the weights w_i, the importance of different measurement segments is highlighted. Simultaneously, the efficiency value Eff is ensured to be no less than the optimal target (with units consistent with Eff), guaranteeing the feasibility of the scheme. For example, the preset target can be set at over 90% of the detection data conforming to the overall trend prediction range to achieve a high-precision objective. The introduction of this formula makes the data acquisition scheme more scientific and systematic.

[0075] Finally, a complete, adjusted version of the cycle task plan is output for the executor to refer to and run, thereby effectively coordinating resource allocation and achieving global optimization. For example, intelligent algorithms can generate specific daily and hourly detection schedules and intensity parameters, which are then sent to monitoring equipment for control, improving long-term stability and reducing energy consumption costs. This enables closed-loop management of the entire chain from strategy formulation to actual implementation.

[0076] The hydrological station flow measurement method and system of this invention comprehensively improves the accuracy and real-time response capability of flow measurement and sediment transport at hydrological stations from multiple perspectives. By comprehensively considering changes in environmental factors and fluid dynamic characteristics, a specific scheme for correction and optimization of real-time measurement data is proposed. The technical issues are elaborated below: 1. To address the impact of real-time water level changes on the parameters of the flow calculation model, this invention constructs a dynamic adjustment algorithm based on data collected by a water level sensor. This algorithm can automatically adjust key parameters involved in the flow calculation model, such as the flow depth and bottom roughness, according to the magnitude and rate of the current water level change, ensuring that the model output more closely reflects the actual dynamic conditions of water flow. Compared to traditional fixed-parameter methods, this approach can significantly reduce the accumulation of errors caused by water level fluctuations.

[0077] 2. To address the optimization of sampling point layout, this invention introduces an intelligent layout control strategy based on flow velocity distribution characteristics. First, the flow velocity gradient pattern within the river channel (e.g., the division into high, medium, and low velocity zones) is obtained through preliminary experiments or historical data. Then, the spatial distribution of sampling points is designed based on the actual geometry of the target flow measurement area. Furthermore, a real-time feedback module is added to automatically adjust the positions of some sampling points when specific anomalies occur, avoiding the problem of missing key flow velocity zones and thus effectively improving data representativeness.

[0078] 3. In the field of sediment transport estimation, this invention proposes a novel intelligent control method that sets appropriate sampling frequency rules under different conditions based on the temporal and spatial variability of historical suspended sediment concentrations. For example, a lower frequency is maintained when the water flow is relatively stable, while the detection density is increased during periods of sudden increase in sediment load to ensure that the particle transport process is fully captured. This flexible approach significantly reduces estimation bias caused by missing samples.

[0079] 4. To address velocity measurement errors caused by environmental interference, this method specifically compensates for common external conditions such as temperature and air pressure. By correcting for ultrasonic wave propagation time delay and pressure changes encountered during ADCP instrument operation, a correction factor is generated after fitting the corresponding physical relationship using mathematical formulas and directly applied to the measurement results, greatly improving the accuracy and reliability of the output data.

[0080] 5. Considering the importance of short-term forecasting and the specific needs of potential application scenarios, machine learning algorithms were further integrated to analyze the massive amounts of monitoring records accumulated over a long period. After training, the prediction module can better reflect the possible changes in traffic flow under seasonal trends and the influence of sudden weather events, and issue early warnings so that staff can take appropriate preventive measures or initiate additional enhanced detection processes.

[0081] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A method for measuring flow at a hydrological station, characterized in that, include: The parameters of the flow calculation model are dynamically adjusted based on real-time water level changes to compensate for the impact of water level fluctuations on flow measurement accuracy. Optimize the layout of sampling points based on flow velocity distribution characteristics and collect and correct the sampling data; Intelligent adjustment of sampling frequency based on the spatiotemporal variation characteristics of suspended sediment concentration to improve the accuracy of sediment transport estimation; The speed of ultrasonic or ADCP measurements is compensated and adjusted according to changes in temperature and air pressure to reduce system bias caused by environmental factors.

2. The hydrological station flow measurement method according to claim 1, characterized in that, The parameters of the flow calculation model based on real-time water level changes for dynamic control further include: Obtain the real-time water level value h; The adjustment coefficients for the flow calculation model parameters are determined using the following formula. ,in It is the historical average water level value, and α is the regulation rate constant; An alarm is triggered when K < 8, indicating that the model parameters need to be recalibrated. If the water level fluctuation amplitude Δh > 1.2m, then reduce the model parameter sensitivity threshold to half of the initial value.

3. The hydrological station flow measurement method according to claim 1, characterized in that, The optimized sampling point layout further includes: Analysis of historical flow velocity distribution characteristics reveals the mainstream and boundary regions; Define the parameters for dividing the main flow region and the boundary as R = max(v) / median(v), where v is the velocity distribution; When R>5, increase the density of sampling points in the mainstream region; For each sub-region i, the area si and the corresponding weight wi satisfy si×wi≤10% of the total area to ensure global balance.

4. The hydrological station flow measurement method according to claim 1, characterized in that, The intelligent adjustment of the sampling frequency further includes: The spatiotemporal coefficient of variation CV = σ / μ was obtained from historical suspended sediment concentration records, where σ is the standard deviation and μ is the mean. The sampling frequency will be doubled when CV > 3. The new optimal sample size N is calculated using Bayesian estimation: N = [log(PriorErr)log(PostErr)] × γ, where γ is the learning rate. For the interval where the sediment content Cj exceeds the threshold of 30 mg / L, a priority encrypted sample extraction strategy is adopted.

5. The hydrological station flow measurement method according to claim 1, characterized in that, Speed ​​compensation control for ultrasonic measurements also includes: Real-time monitoring of temperature T℃ and air pressure P (hPa) is used to calculate the compensation ratio factor through the environmental function En=βT+γP; When the En value deviates from the set range [10, +10], the measurement result Vadj is corrected as Vadj = Vo / (1 + θEn), where θ is the negative correlation adjustment coefficient. Adding a speed error prediction unit to ADCP-type equipment to provide early warning of system deviations; If more than 5 large deviations accumulate within an hour, a self-correction process will be triggered.

6. The hydrological station flow measurement method according to claim 5, characterized in that, It also includes predicting short-term traffic trends, the specific steps of which include: Retrieve 7 consecutive days of data from the historical database as the reference time series Dt; The target water volume for the next time period t+Δt is predicted using the exponential smoothing algorithm: Qt=αDt+(1-α)Dt1; where α is the attenuation coefficient, typically taking values ​​[2, 0.6]. If, based on the actual water conditions, extreme precipitation is predicted to occur in the next two days with a probability P(t>2days)>95%, then an emergency correction item will be added to the estimated flow rate. Ensure that the response cycle of the entire forecasting system is controlled at the minute level to respond quickly to sudden flooding events.

7. The hydrological station flow measurement method according to claim 1, characterized in that, An additional stability check step is added to the dynamic control process: The cumulative model error Eacc is calculated every two hours. If Eacc ≥ threshold, i.e., the set tolerance limit is met, the adjustment mode will be automatically restarted. The next control interval is adjusted using the recursive feedback control principle, where dt = dt × δ, and δ is the feedback strength correction factor. The total computing resources shall not exceed the server performance load limit Load_max.

8. The hydrological station flow measurement method according to claim 1, characterized in that, The optimization of the velocity distribution characteristics is further refined to a non-uniform sampling mode: For different partitions k, the sampling density dk∝|Ak / ΣA| is set, where Ak represents the effective flow area of ​​the k-th region; A fuzzy logic controller is introduced to determine whether to activate the secondary local precision measurement module; the condition is F(x,y)=1 if and only the current sampled data residual Set accuracy level And x and y belong to the critical domain; Ensure that the point cloud coverage rate in the key area is not lower than the specified baseline r0; To prevent any two adjacent partitions from having a sampling gap smaller than the preset minimum interval min_gap, the assumption of spatial independence is guaranteed.

9. The hydrological station flow measurement method according to claim 8, characterized in that, The further improvement of the ultrasonic velocity correction mechanism considering the influence of water temperature specifically includes: Set the temperature gradient dT / dL and evaluate the significance of the change along the propagation path; specifically, use the equation Δc~f(dT)L to express the relationship between the propagation velocity difference and the path length; f is determined by laboratory calibration; If ΔT crosses the critical value of layering Tthreash=5°C, then the depth subdivision processing method is activated to correct each layer segment separately. The multi-level correction results are integrated and a comprehensive velocity vector matrix is ​​generated for the final calculation of Vres=(∑WiVi) / (∑Wi), where Wi represents the weight allocation; Regularly update environmental parameters and calibrate the database to maintain the long-term effectiveness of the model.

10. A hydrological station measurement system, which can be applied to the hydrological station flow measurement method according to any one of claims 1 to 9, characterized in that, The hydrological station measurement system includes: The control unit is used to dynamically adjust the parameters of the flow calculation model based on real-time water level changes to compensate for the impact of water level fluctuations on flow measurement accuracy. The acquisition unit is used to optimize the layout of sampling points according to the flow velocity distribution characteristics and to acquire and correct the sampling data; The calculation unit is used to intelligently adjust the sampling frequency according to the spatiotemporal variation characteristics of suspended sediment concentration to improve the accuracy of sediment transport estimation. The compensation unit is used to compensate and adjust the ultrasonic or ADCP measurement speed according to changes in temperature and air pressure to reduce system deviations caused by environmental factors.

Images