Subway dewatering and recharge parameter measurement method and system based on intelligent sensing system

Abstract

The application relates to the field of intelligent measurement, and particularly discloses a subway dewatering and recharge parameter measurement method and system based on an intelligent sensing system. The method comprises the following steps: arranging flow, liquid level and water quality sensors according to different water-bearing rock group distribution at multiple well groups and stations along a subway dewatering and recharge project, synchronously collecting data, normalizing initial data according to precision levels, establishing fusion rules according to mechanical, seepage and well flow coupling models and geohydrological characteristics, integrating pretreatment data according to the rules, and generating a conflict-free fusion parameter set; comparing a normal working threshold of the sensor to troubleshoot faults, judging abnormal reasons according to crack clay permeation characteristics and model calculation values, and screening effective parameter data; outputting the effective parameters for dewatering and recharge scheme adjustment, inversely calibrating the fault sensor according to the effective parameters, and forming a closed-loop measurement process. The application can reduce engineering adjustment deviation caused by data problems, and guarantee stable progress of the subway dewatering and recharge project.

Description

Technical Field

[0001] This invention relates to the field of intelligent measurement, specifically to a method and system for measuring subway precipitation reinjection parameters based on an intelligent sensing system. Background Technology

[0002] Jinan Rail Transit Line 9 (Phase I), a major implementation project of Shandong Province in 2023 (Document No. 13 of Shandong Provincial Government

[2023] ), is the first metro line in the Jinan New East Railway Station area to traverse the concentrated outcrop area of ​​the Baiquan Spring Group, carrying the dual mission of upgrading urban transportation and protecting historical and cultural resources. Jinan's springs, as the "soul" of the city, bear significant economic, cultural, and ecological value. The project is only 1.5km from the Baiquan Spring Group (Lujia Station and Peijiaying Station). How to achieve "metro construction and spring water coexistence" during the project's advancement has become the core issue in resolving the contradiction between urban infrastructure construction and the protection of scarce ecological resources, and it also places extremely high demands on the accuracy and reliability of the project's hydrogeological parameter measurements.

[0003] From an engineering geological and hydrological perspective, the geological structure along the route is complex. Except for the Huashan area, which is covered by exposed gabbro, the rest of the area is covered by Quaternary strata, including fill, silty clay, clay, and gravelly soil. The underlying strata involve Cretaceous Qingshan Group gabbro, Ordovician limestone, marble, and Carboniferous-Permian mudstone and sandstone. The groundwater system exhibits a "multi-rock group, strong correlation" characteristic, divided into three types of aquifer groups along the route: loose rock porous aquifer group, clastic and igneous rock fracture aquifer group, and carbonate rock fracture karst aquifer group. Although each group is controlled by aquitards and forms a relatively independent water cycle, tectonic activity organically links them in the overall cycle, further increasing the complexity of hydrological parameter monitoring. Crucially, a unique "karst water top-supporting pore water" phenomenon exists along the route: underlying limestone karst water (single well yield > 230m³) 3 / d) The upward flow through the limestone skylight forms a hydraulic connection with the pore water of the upper Quaternary system. This process can easily lead to non-continuous fluctuations in water level and flow rate data, which can interfere with the judgment of the accuracy of parameters.

[0004] To accurately control the impact of the project on the spring water, seven well groups (numbered 1 to 7) were set up along the project route (using a regular hexagonal layout, with a karst water pumping well at the center and pore water pumping wells at the corners, along with supporting observation wells of the same depth) and seven monitoring points at stations including Xiaoqinghe East Station and Kaiyuan Road Station to Maozhuang Station. Simultaneous collection of three core parameters—flow rate, liquid level, and water quality—is required. However, traditional subway precipitation reinjection parameter measurement technology reveals significant shortcomings in this scenario: firstly, data fusion conflicts arise because flow rate, liquid level, and water quality sensors operate on different measurement principles, and data formats (such as flow rate units of m³) differ. 3The large differences in the units (e.g., flow rate ±2%, flow rate ±1cm) and accuracy levels (e.g., flow rate ±2%, flow rate ±1cm) make direct integration prone to logical contradictions. Secondly, it is difficult to determine the cause of anomalies. Phenomena such as sudden drops in flow rate cannot be quickly distinguished as sensor failure, changes in the permeability characteristics of fractured cohesive soil (the normal range for sudden changes in permeability coefficient is ±20% of the baseline value), or actual hydrological anomalies (e.g., karst top-back recharge), which can easily lead to misjudgments. Thirdly, the lack of a closed-loop process means that there is no reverse calibration mechanism based on effective parameters after sensor failure, making it difficult to continuously guarantee data accuracy. This may lead to deviations in the adjustment of precipitation recharge plans, threatening project safety and spring stability (e.g., during the continuous high-water season from 2021 to 2023, the rise in water level along the route further amplified the risk of parameter errors).

[0005] Against this backdrop, there is an urgent need for a set of intelligent measurement technologies adapted to the complex geological and hydrological conditions of Jinan Metro Line 9.

[0006] In the measurement technology of subway precipitation reinjection parameters, the intelligent sensing system needs to simultaneously collect flow rate (m³ / s). 3 Data types such as liquid level (m) and water quality (pollutant concentration) are easily fused due to the different measurement principles of different sensors, data formats, and accuracy levels. At the same time, it is difficult to determine the validity of the data. For example, if the liquid level suddenly drops, it is not possible to quickly distinguish whether it is due to sensor failure, changes in the permeability characteristics of fractured cohesive soil, or actual hydrological anomalies, which can easily lead to errors. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for measuring subway precipitation reinjection parameters based on an intelligent sensing system, as well as a system for measuring subway precipitation reinjection parameters based on an intelligent sensing system, to solve the problems mentioned in the background art.

[0008] This invention provides a method for measuring subway precipitation reinjection parameters based on an intelligent sensing system, comprising the following steps:

[0009] Along the subway rainwater recharge project, flow, liquid level and water quality sensors were deployed according to different water-bearing rock groups at multiple well groups and stations to collect data synchronously, record basic hydrogeological parameters and initially unify the data format to form an initial parameter dataset.

[0010] The initial data was normalized according to the accuracy level, and a filtering algorithm was used to remove interference noise from the construction environment. Non-continuous abnormal data were removed based on the characteristics of karst water top support replenishment, resulting in a preprocessed dataset.

[0011] Based on the coupled models of mechanics, seepage, and well flow, as well as geological and hydrological characteristics, fusion rules are established. Preprocessed data are then weighted and integrated according to these rules to generate a conflict-free fusion parameter set.

[0012] By comparing the sensor's normal operating threshold to troubleshoot the fault, and by determining the cause of the anomaly based on the permeability characteristics of fractured cohesive soil and the model calculation values, effective parameter data are selected.

[0013] The output of valid parameters is used to adjust the precipitation reinjection scheme, and the faulty sensor is calibrated in reverse based on the valid parameters, forming a closed-loop measurement process.

[0014] Furthermore, the flow rate, liquid level, and water quality sensors are deployed along the multiple well groups and stations along the subway dewatering recharge project, according to the distribution of different aquifer rock groups. Specifically, the following steps are taken: First, survey the well groups and stations along the subway dewatering recharge project to determine and differentiate the types of aquifer rock groups; then, deploy flow rate, liquid level, and water quality sensors at corresponding locations according to the rock group type, with different installation locations and densities for different rock groups; after installation, record the sensor location, the rock group it belongs to, and the surrounding strata; finally, select appropriate sensor types based on the subway construction environment and accuracy requirements.

[0015] Furthermore, the synchronous data acquisition, recording of basic hydrogeological parameters, and preliminary standardization of data format to form an initial parameter dataset are specifically as follows: each sensor aligns its acquisition time through a unified time synchronization module; flow rate, liquid level, and water quality data are acquired at corresponding frequencies; during acquisition, basic hydrogeological parameters such as stratigraphic type, permeability coefficient, and porosity at the monitoring point are recorded synchronously; the raw data is converted into standard data frames containing sensor ID, timestamp, parameter type identifier, numerical value and unit, and check digit in a unified format; and the data are summarized to form the initial parameter dataset.

[0016] Furthermore, the initial data normalization process according to accuracy level is specifically as follows: first, sort out the accuracy level of each sensor in the initial parameter dataset; determine a unified error range and map the error range of all sensor data to the same range; for flow rate, liquid level, and water quality data, calculate the error range according to their accuracy, and then map them to the unified range through linear transformation.

[0017] Furthermore, the filtering algorithm is used to remove construction environment interference noise, and non-continuous abnormal data are eliminated based on the characteristics of karst water top-support replenishment to obtain a preprocessed dataset. Specifically, the initial parameter data is processed using the Kalman filter algorithm, the characteristics of subway construction environment noise are incorporated when constructing the equation, and the filtering parameters are dynamically adjusted to filter interference; the hydrological dynamic characteristics of karst water top-support replenishment are analyzed, and the top-support replenishment range is determined based on geological surveys along the line; the filtered data is checked, non-continuous abnormal data are identified and eliminated, and the remaining data are summarized to form a preprocessed dataset.

[0018] Furthermore, the fusion rules established based on the coupled mechanical, seepage, and well flow models and geological and hydrological characteristics are as follows: First, a coupled mechanical-seepage-well flow model is constructed, incorporating the geological structure, rock group mechanical properties, seepage parameters, and well flow parameters along the subway dewatering recharge project route; then, based on the model output and on-site geological and hydrological characteristics, flow rate weights are set according to the measured flow range, liquid level weights are set according to the formation permeability coefficient, and water quality weights are set according to the correlation between water quality and flow rate and liquid level, with the sum of the three weights being 1; finally, the weights are dynamically corrected based on the real-time theoretical values ​​of the model, and the geological and hydrological parameters are re-evaluated and the fusion rules are updated when the deviation exceeds the limit.

[0019] Furthermore, the step of weighted integration of preprocessed data according to rules to generate a conflict-free fusion parameter set specifically involves: retrieving preprocessed data and corresponding basic hydrogeological parameters; determining the weights of flow rate, level, and water quality for each monitoring point based on the design flow range of the irrigation and drainage scheme, the formation permeability coefficient, and the correlation between water quality and flow rate / level, ensuring that the sum of the weights is 1; multiplying the preprocessed parameter data by their corresponding weights to obtain weighted contribution values, and integrating them to form the fusion parameters for each monitoring point; comparing the theoretical parameter values ​​of the coupling model, adjusting and retaining the fusion parameters that meet the requirements according to the deviation, and summarizing to generate a conflict-free fusion parameter set.

[0020] Furthermore, the method of troubleshooting by comparing the normal operating thresholds of the sensors specifically involves: first, determining the normal operating thresholds for each type of sensor; comparing real-time collected data with the corresponding thresholds at a fixed frequency and recording the sensor identifier and collection time for data exceeding the threshold; not directly determining a fault based on a single instance of exceeding the threshold, but continuously monitoring subsequent data and determining whether it is a fault or a momentary anomaly based on the number of occurrences; using the sensor installation location and the type of aquifer to assist in the judgment; and summarizing the sensor identifier, installation location, type of aquifer, specific data exceeding the threshold, and comparison records for the preliminary determination of the fault, for further investigation into the cause of the fault.

[0021] Furthermore, the process of determining the cause of anomalies based on the permeability characteristics of fractured cohesive soil and model calculations, and then selecting valid parameter data, specifically involves: retrieving the fused parameters after sensor fault diagnosis and the baseline value of the permeability coefficient of fractured cohesive soil at the corresponding monitoring point; determining the normal range of permeability coefficient mutations; if the flow rate or liquid level is abnormal, first checking the actual permeability coefficient, and if it is within the normal range, determining that the anomaly is due to a change in permeability characteristics; otherwise, retrieving the theoretical value from the mechanical-seepage-well flow coupling model; calculating the percentage difference between the parameter and the theoretical value, and if it is within the allowable range, it is considered valid hydrological anomaly data; if it exceeds the range, checking the basic hydrogeological parameters, sensor records, preprocessing procedures, and correcting and re-judging; summarizing the valid parameters and recording the causes of anomalies to form a valid parameter dataset.

[0022] Furthermore, the output valid parameters are used to adjust the precipitation reinjection scheme. Based on these valid parameters, faulty sensors are calibrated in reverse, forming a closed-loop measurement process. Specifically, the valid parameters are categorized by well group, station, and aquifer type; the precipitation reinjection scheme is adjusted according to the flow rate, liquid level, and station water quality of different rock groups; the faulty liquid level, flow rate, and water quality sensors are calibrated using the valid parameters; after calibration, the sensors are reconnected to the system, and data is collected, preprocessed, fused, and its validity is determined again; the coupling model parameters are updated after each process is completed; this achieves a closed loop of data acquisition, processing, determination, and calibration, ensuring that valid parameters guide the scheme and that subsequent data is accurate.

[0023] A subway precipitation reinjection parameter measurement system based on an intelligent sensing system includes: sensors, which are arranged in multiple well groups and stations along the subway precipitation reinjection project, and flow rate, liquid level and water quality sensors are arranged according to different water-bearing rock groups;

[0024] The data acquisition unit is used to align the acquisition time of each sensor through a unified time synchronization module, acquire data at the corresponding frequency to synchronously record basic hydrogeological parameters, and convert the raw data into standard data frames to form an initial parameter dataset.

[0025] The normalization processing unit is used to sort out the accuracy levels of each sensor in the initial data, determine a unified error range, and map the error range of different types of data to the same range through linear transformation.

[0026] The filtering unit is used to incorporate the noise characteristics of the construction environment using the Kalman filtering algorithm and dynamically adjust parameters to filter out interference noise.

[0027] The anomaly removal unit is used to analyze the characteristics of karst water backwater recharge, check the filtered data, identify and remove non-persistent anomalies to form a preprocessed dataset.

[0028] The fusion rule unit is used to construct mechanical seepage well flow. The coupled model combines geological and hydrological characteristics to set the weights of each parameter according to the measured flow range, formation permeability coefficient and water quality correlation, and dynamically corrects them.

[0029] The weighted integration unit is used to retrieve preprocessed data and hydrological parameters, calculate the weighted contribution value of each parameter according to the integration rules, and integrate them to form a fused parameter set.

[0030] The fault diagnosis unit is used to determine the normal operating threshold of each sensor, compare the real-time data with the threshold at a fixed frequency, record the situations that exceed the threshold, and determine the sensor fault after continuous monitoring.

[0031] The anomaly analysis unit is used to retrieve the fusion parameters and permeability coefficient baseline values ​​after fault diagnosis, check the actual permeability coefficient changes, calculate the difference percentage based on the theoretical values ​​of the coupling model, and filter valid parameter data.

[0032] The scheme adjustment unit is used to classify effective parameters according to well group sites and rock group types, and adjust the precipitation reinjection scheme based on flow rate, liquid level and water quality.

[0033] The sensor calibration unit is used to correct faulty level sensors based on valid parameters, adjust the sensitivity of flow sensors at zero point, and calibrate the detection range of water quality sensors.

[0034] The closed-loop update unit is used to reconnect the calibrated sensor to the system, and re-execute the acquisition preprocessing fusion and validity determination process. The coupling model parameters are updated every time the process is completed.

[0035] The beneficial effects of this invention are:

[0036] To address the fusion conflict caused by differences in data format and accuracy due to variations in measurement principles among different sensors, flow rate, level, and water quality sensors are deployed along the engineering route according to the distribution characteristics of aquifers. Data is collected synchronously and initially standardized in format. The initial data is then normalized according to accuracy level, and filtering algorithms are used to remove interference noise from the construction environment. Fusion rules are established based on mechanical, seepage, and well flow coupling models and geological and hydrological characteristics. Weighted integration of preprocessed data allows for smooth fusion of different data types under a unified error measurement standard and integration logic, effectively eliminating data fusion conflicts. The generated fusion parameter set accurately reflects the actual hydrological state of the monitoring points. For issues such as difficulty in determining data validity and the inability to quickly distinguish the causes of anomalies, which can easily lead to misjudgments, the fusion parameters are compared with the normal operating thresholds of the sensors for initial fault diagnosis. Then, the changing patterns of the permeability characteristics of fractured cohesive soil are used to determine whether the anomaly is caused by changes in formation characteristics. Difference analysis, combined with the calculated values ​​from the coupling model, can quickly and accurately distinguish whether the data anomaly is caused by sensor malfunction, changes in the permeability characteristics of fractured cohesive soil, or actual hydrological anomalies. This avoids previous misjudgments caused by the inability to clearly identify the cause of the anomaly, and the selected effective parameter data has high reliability. Simultaneously, the effective parameters are used to adjust the precipitation reinjection scheme, and the faulty sensors are calibrated in reverse based on the effective parameters. The calibrated sensors are then reconnected to the system to participate in data acquisition, and undergo preprocessing, fusion, and validity determination processes again, forming a closed loop of "acquisition, processing, determination, and calibration". This continuously improves the accuracy of subsequent data acquisition, ensuring that the effective parameters output each time can accurately guide the adjustment of the engineering scheme, reduce engineering adjustment deviations caused by data problems, and ensure the stable progress of the subway precipitation reinjection project. Attached Figure Description

[0037] Figure 1 A flowchart of a method for measuring subway precipitation reinjection parameters based on an intelligent sensing system, according to the present invention. Detailed Implementation

[0038] In its specific implementation, this application discloses a method for measuring subway precipitation reinjection parameters based on an intelligent sensing system, such as... Figure 1 The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system includes the following steps:

[0039] Along the subway rainwater recharge project, flow, liquid level and water quality sensors were deployed according to different water-bearing rock groups at multiple well groups and stations to collect data synchronously, record basic hydrogeological parameters and initially unify the data format to form an initial parameter dataset.

[0040] The initial data was normalized according to the accuracy level, and a filtering algorithm was used to remove interference noise from the construction environment. Non-continuous abnormal data were removed based on the characteristics of karst water top support replenishment, resulting in a preprocessed dataset.

[0041] Based on the coupled models of mechanics, seepage, and well flow, as well as geological and hydrological characteristics, fusion rules are established. Preprocessed data are then weighted and integrated according to these rules to generate a conflict-free fusion parameter set.

[0042] By comparing the sensor's normal operating threshold to troubleshoot the fault, and by determining the cause of the anomaly based on the permeability characteristics of fractured cohesive soil and the model calculation values, effective parameter data are selected.

[0043] The output of valid parameters is used to adjust the precipitation reinjection scheme, and the faulty sensor is calibrated in reverse based on the valid parameters, forming a closed-loop measurement process.

[0044] The plan involves deploying flow, level, and water quality sensors at multiple well groups and stations along the subway dewatering recharge project line, according to the distribution of different aquifer groups. Specifically, the implementation involves first surveying well groups 1 through 7 along the subway dewatering recharge project line, as well as stations such as Xiaoqinghe East Station and Peijiaying Station. This process clarifies the aquifer group types in each area, distinguishing the distribution range of loose porous aquifer groups and carbonate fractured karst aquifer groups.

[0045] For well groups 1 through 7, if the area where the well groups are located is a loose, porous, water-bearing rock group, sensors are installed in each monitoring well. Flow sensors are installed at the inlet and outlet of the monitoring wells to monitor the flow rate of the incoming and outgoing water. Level sensors are arranged at intervals along the well depth, with a spacing of 2 to 3 meters, to obtain level data at different depths. Water quality sensors are placed in locations within the monitoring wells where water flow is relatively smooth to ensure accurate water sample collection.

[0046] If the well group is located in an area of ​​carbonate rock fractured karst aquifer, the location of fracture development zones and karst channels should be determined through preliminary exploration. Flow sensors should be preferentially installed at the inlet and outlet of the karst channels to monitor the water flow within the channels. Level sensors should be strategically placed in densely fractured sections, one every 1.5 to 2.5 meters, to capture changes in water level in the fractured areas. Water quality sensors should be placed in the fracture development zones and within the karst channels to comprehensively collect water quality data from different areas.

[0047] At stations such as Xiaoqinghe East Station and Peijiaying Station, monitoring areas were divided according to the distribution of aquifers within the stations. In areas with loose, porous aquifers, the sensor density was relatively high, with each monitoring point spaced 3 to 5 meters apart. In areas with carbonate, fractured karst aquifers, sensors were deployed around fractures and karst channels to ensure coverage of the main hydrological activity areas.

[0048] After each sensor is installed, its specific location coordinates are recorded. The type of aquifer at the sensor's location, as well as the surrounding geological conditions, such as whether it is cohesive soil or gravelly soil, are also noted. Flow sensors are selected based on their vibration and electromagnetic interference resistance capabilities to suit the subway construction environment. Level sensors must achieve a measurement accuracy of ±1cm to meet the accuracy requirements of subsequent data processing. Water quality sensors are configured according to the types of pollutants of interest in the project to ensure accurate detection of pollutant concentrations.

[0049] The synchronized data acquisition records basic hydrogeological parameters and initially standardizes the data format to form an initial parameter dataset. Specifically, in implementation, each sensor uses a unified time synchronization module to align data acquisition times. The time synchronization accuracy is controlled within ±0.1 seconds. The flow sensor acquires instantaneous flow data every 5 minutes and simultaneously calculates the daily flow. The level sensor acquires data in real time and stores the recorded value every minute. The water quality sensor collects water samples hourly and analyzes pollutant concentrations.

[0050] During the data collection process, basic hydrogeological parameters at each monitoring point were recorded simultaneously. These included the stratigraphic type, such as silty clay, medium sand, and limestone. The permeability coefficient of cohesive soils was recorded, ranging from 0.467 to 12.274 m / d. The permeability coefficient of gravelly soils was recorded, for example, 40.325 m / d. Porosity, specific yield, and other parameters were also recorded.

[0051] The raw data collected undergoes initial format standardization. All data is converted to a standard data frame format. Each data frame contains a unique sensor identifier (ID), a timestamp of data acquisition accurate to milliseconds, parameter type identifiers (e.g., flow rate, liquid level, water quality), and parameter values ​​and their corresponding units (flow rate units are standardized to meters per second). 3 / d, with liquid level units standardized to meters (m) and pollutant concentration units standardized to mg / L. A checksum is added to the end of the data frame for preliminary verification of data transmission integrity. After the above processing, the initial parameter dataset is formed.

[0052] The initial data is normalized according to accuracy level. Specifically, this involves first identifying the accuracy levels of each sensor in the initial parameter dataset. The flow sensor accuracy is ±2%. The level sensor accuracy is ±1cm. The water quality sensor accuracy is determined based on the type of pollutant detected, such as ±0.5mg / L for a certain pollutant concentration. A unified error range is established, mapping the error range of all sensor data to the [-1, 1] interval. For flow data, the measured value is used as a benchmark, and the upper and lower limits of the error are calculated based on an accuracy range of ±2%, then linearly transformed to map the error value to [-1, 1]. For level data, based on an absolute accuracy of ±1cm, the error range of the actual measured value is calculated, and similarly linearly transformed to [-1, 1]. For water quality data, based on the corresponding pollutant concentration detection accuracy, such as ±0.5mg / L, the error range is calculated and transformed to a unified interval. This process ensures that sensor data of different accuracy levels are under the same error measurement standard, facilitating subsequent data fusion processing.

[0053] The original "accuracy level linear transformation normalization" technique is a static basic method. Its core is to map the error range through a fixed linear function. It has two key limitations: First, it does not take into account the dynamic scene characteristics of subway precipitation recharge (such as strong mechanical vibration during the excavation of the foundation pit, which can easily amplify the error of liquid level / flow rate data; the error shrinks during the stable recharge period, and the static linear transformation cannot adapt to the dynamic fluctuation of the error); Second, it does not quantify the discrete characteristics of sensor accuracy (the actual error of sensors of the same accuracy level is discretely distributed, and the linear transformation treats it as "homogeneous accuracy", which can easily mask the unreliability of highly discrete data).

[0054] Therefore, this application further proposes a "two-dimensional adaptive normalization method based on dynamic scene factors and accuracy variation coefficient". It quantifies the dispersion (variance coefficient) of sensor accuracy through a statistical model, dynamically adjusts the mapping interval during the construction phase, and then achieves differentiated normalization of accuracy confidence through nonlinear mapping. This method is suitable for scene dynamics and can distinguish data confidence.

[0055] The "two-dimensional adaptive normalization method based on dynamic scene factors and accuracy variation coefficient" includes:

[0056] Input two types of information: "static accuracy data" and "dynamic scene data" to ensure coverage of both the inherent characteristics of the sensor and the impact of the engineering scenario.

[0057] Static accuracy data: The accuracy level of each sensor in the initial parameter dataset (e.g., flow sensor). ±2%, liquid level sensor ±1cm, water quality sensor (±0.5 mg / L), sensor historical error sequence over the past 3 months ( Groups used to calculate accuracy dispersion and sensor measurement range. (like 0~5000m (0~30m) 3 / d,

[0058] Dynamic scene data: Current construction stage identifier (e.g., "excavation period", "recharge stabilization period", "pipeline commissioning period"), and scene error amplification factor corresponding to each stage. (Based on historical engineering data: Excavation period) Stable period Debugging period ), current sensor Group real-time measurement values (like Five sets of real-time data: 1200, 1215, 1180, 1220, 1195 m 3 / d).

[0059] Calculate the coefficient of variation (quantization accuracy dispersion) of sensor accuracy by using the sensor's historical error sequence. (Reflects the degree of dispersion of the error,) The larger the value, the higher the precision dispersion and the lower the data reliability. This is further combined with the scene error amplification factor. The dynamic precision coefficient of variation is obtained. (The impact of related scenarios on error).

[0060] Precision coefficient of variation The coefficient of variation for precision is calculated as the ratio of the standard deviation to the mean of the error series, using the following formula:

[0061] ;

[0062] in: Standard deviation of the sensor's historical error sequence (m is the amount of historical error data, (For the i-th group of historical errors); The mean of the historical error sequence of the sensor. (If the sensor has no systematic error, Take at this time To avoid a denominator of 0 (which does not affect the dispersion trend); Dimensionless parameter The precision dispersion is low. It indicates high precision and dispersion.

[0063] Example: Historical error sequence Calculated m m (Due to the large absolute value of the flow error,) The value is relatively high, so range normalization needs to be considered. 3 / d, 3 / d, then

[0064] Dynamic accuracy coefficient of variation calculate:

[0065] Introducing scene error amplification factor Correcting the impact of scene on precision dispersion:

[0066] ;

[0067] in: Scene error amplification factor (excavation period) Stable period Debugging period ); Reference range (uniformly set to 1000, used to eliminate the impact of range differences between different sensors). (the impact); The current measurement range of the sensor (e.g.) of Example: During the excavation period ( ),but This quantifies the dynamic accuracy dispersion of "excavation period + flow sensor".

[0068] Determine the dynamic error mapping interval (adapting to different scenarios and accuracy) based on Determine the dynamic error range The fixed range that replaces the original technology The larger the value (higher precision dispersion / stronger scene interference), the wider the range, ensuring coverage of 99.7% of the actual error; The smaller the value, the narrower the interval, thus avoiding invalid data occupying the interval range.

[0069] Calculation of dynamic error interval:

[0070] ;

[0071] ;

[0072] in:

[0073] Current sensor The mean of the original data set. (like of );

[0074] Confidence factor (taken as 3, based on normal distribution) Principles, ensure The error falls within the range.

[0075] Example: of , , ,but , To adapt to error fluctuations during the excavation period; if In a stable period ( ), The interval then shrinks to This better matches the error characteristics during the steady-state period.

[0076] Then, a two-dimensional adaptive normalization mapping is performed (to distinguish data confidence levels), using a non-linear Sigmoid mapping function (replacing the original linear function), that is, mapping the original data to a unified interval. Furthermore, the slope of the function can distinguish the confidence level of the data. The closer the error is to the center of the interval (high confidence level), the more concentrated the mapped value is around 0.5; the closer the error is to the edge of the interval (low confidence level), the closer the mapped value is to 0 or 1, which makes it easier to prioritize the use of high confidence data when fusion data in the future.

[0077] Then, a nonlinear normalization mapping is performed, assuming the error of the original data is... ( For the first (Based on the original data set), the normalized value is... for:

[0078] ;

[0079] in: : Slope adjustment parameter (set to 8, to control the steepness of the Sigmoid function and ensure that confidence differences can be effectively distinguished); : To reduce the error Intermediate variables that are linearly mapped to [0, 1]; : Natural constant (approximately 2.718).

[0080] Example: A set of raw data , Substituting into the formula, we get: intermediate variable = ,but (Close to 0.5, high confidence level); If a certain set of data (error (close to the edge of the interval), then the intermediate variable , (Closer to 1, lower confidence level), clearly distinguishing the reliability of the two types of data.

[0081] Then output the results, normalizing the dataset: Group normalized value (All fall within the uniform interval [0, 1], and implicitly contain data confidence information). Dynamic parameter set: the dynamic accuracy variation coefficient of each sensor. Dynamic error range (It can be directly used in subsequent data fusion stages as the basis for calculating "data confidence weights" and improve the reliability of fusion parameters).

[0082] The proposed method employs a filtering algorithm to remove construction environment interference noise. Based on the characteristics of karst water backflow, non-continuous abnormal data are eliminated to obtain a preprocessed dataset. Specifically, a Kalman filter algorithm is used to process the initial parameter data. Noise characteristics of the subway construction environment are incorporated when constructing the state equation and observation equation. Mechanical vibrations during construction generate low-frequency noise of 5 to 50 Hz, and electromagnetic interference manifests as instantaneous pulse signals. Initial filtering parameters are set based on these characteristics, and the process noise covariance matrix and measurement noise covariance matrix are dynamically adjusted according to the actual measured noise intensity. The filtering parameters are updated every 10 seconds, enabling the algorithm to adapt to noise changes in real time, filtering out irregularly fluctuating interference data, and making the flow rate, liquid level, and water quality data smoother.

[0083] The hydrological dynamics of pore water recharge from karst tops were analyzed. Top-pressure recharge is mostly triggered by a short-term increase in karst water pressure, manifested as a sudden rise in liquid level, typically lasting 10 to 30 minutes, without synchronous and stable changes in flow rate, and water quality indicators show no significant fluctuations. Based on the geological survey results along the route, the degree of karst development and the possible range of top-pressure recharge in the areas where each monitoring point is located were determined.

[0084] The filtered data is examined point by point. When the liquid level data shows a sudden rise or fall, the duration of the change is calculated. If the duration is within 10 to 30 minutes, and the flow rate data at the corresponding monitoring point does not change accordingly, the concentration of water pollutants is stable, and the area is within the possible range of karst water backwater replenishment, then it is determined to be non-continuous abnormal data. This type of data is removed from the dataset, and the remaining data is compiled to form a preprocessed dataset.

[0085] The proposed integration rules are based on a coupled model of mechanics, seepage, and well flow, as well as geological and hydrological characteristics. Specifically, in implementation, a coupled mechanical-seepage-well flow model is first constructed. This model incorporates geological structural parameters along the subway dewatering and recharge project route, including the thickness distribution and spatial arrangement of various aquifer groups. The model considers the mechanical properties of different rock groups, such as the compressive strength and deformation modulus of fractured cohesive soil. Simultaneously, seepage parameters, such as permeability coefficient and pore water pressure transmission characteristics, are integrated. Well flow calculations include parameters such as well pumping / injection rates and influence radii.

[0086] Based on the output of the coupled model, flow data fusion rules are established in conjunction with on-site geological and hydrological characteristics. Weighting benchmarks are determined according to the design flow range in the irrigation and drainage scheme. When the measured flow is within the design range, a higher weight is assigned, up to a maximum of 0.4. If the measured flow exceeds the design range but is within the allowable fluctuation range predicted by the model, the weight is reduced to 0.2 to 0.3. For flow data exceeding the model prediction range, the weight is further reduced to below 0.1.

[0087] For liquid level data, the weights are adjusted based on the permeability coefficient of the corresponding strata. For cohesive soils in loose porous aquifers, with permeability coefficients ranging from 0.467 to 12.274 m / d, the weight for liquid level data is set to 0.25 to 0.35. For gravelly soils with a permeability coefficient of 40.325 m / d, the weight for liquid level data is increased to 0.4 to 0.5. In carbonate fractured karst aquifers, the weight for liquid level data in densely fractured sections is set to 0.3 to 0.45 according to the degree of fracture development, while the weight for liquid level data within karst channels is set to 0.45 to 0.6.

[0088] The data fusion rules are based on groundwater cycle patterns, with correlation thresholds set accordingly. First, baseline pollutant concentrations at each monitoring point under normal operating conditions are determined through long-term monitoring. When the measured pollutant concentration is within ±10% of the baseline value, it is considered to have a good correlation with flow rate and liquid level data, and the weight is set to 0.2 to 0.3. If the concentration change exceeds ±10% but does not exceed ±20%, and is consistent with the flow rate change trend, the weight is adjusted to 0.15 to 0.25. When the concentration change exceeds ±20% or contradicts the flow rate change trend, the weight drops to below 0.1.

[0089] All weight settings must ensure that the sum of the weights for flow rate, liquid level, and water quality is 1. After the fusion rules are established, the weight coefficients are dynamically adjusted based on the theoretical values ​​calculated in real time by the coupled model. When the deviation between the theoretical value and the measured data is less than 3%, the current weights are maintained. When the deviation is between 3% and 5%, the corresponding parameter weights are fine-tuned according to the deviation ratio. When the deviation exceeds 5%, the geological and hydrological characteristic parameters are reassessed, and the fusion rules are updated.

[0090] The process involves weighted integration of preprocessed data according to rules to generate a conflict-free fusion parameter set. Specifically, this involves first retrieving flow rate, liquid level, and water quality data, along with corresponding hydrogeological parameters, from each monitoring point within the preprocessed dataset. For each monitoring point, the weight of the flow rate data is determined. If the measured flow rate is within the design flow rate range of the irrigation and dewatering scheme, the flow rate weight is 0.4. If the measured flow rate exceeds the design range but is within the allowable fluctuation range predicted by the mechanical seepage well flow coupling model, the flow rate weight is 0.2 to 0.3. If the measured flow rate exceeds the model prediction range, the flow rate weight is below 0.1. Next, the weight of the liquid level data is determined. If the monitoring point is located in a cohesive soil area of ​​a loose porous aquifer with a permeability coefficient of 0.467 to 12.274 m / d, the liquid level weight is 0.25 to 0.35. If it is located in a gravelly soil area of ​​a loose porous aquifer with a permeability coefficient of 40.325 m / d, the liquid level weight is 0.4 to 0.5. If the water level is located in a densely fractured section of a carbonate rock-type fractured karst aquifer, the liquid level weight is 0.3 to 0.45. If it is located within a karst channel of a carbonate rock-type fractured karst aquifer, the liquid level weight is 0.45 to 0.6. Next, determine the weight of the water quality data. First, obtain the baseline value of pollutant concentration under normal operating conditions for the monitoring point. If the measured pollutant concentration is within ±10% of the baseline value, the water quality weight is 0.2 to 0.3. If the measured pollutant concentration exceeds ±10% of the baseline value but does not exceed ±20%, and is consistent with the flow rate trend, the water quality weight is 0.15 to 0.25. If the measured pollutant concentration exceeds ±20% of the baseline value or contradicts the flow rate trend, the water quality weight is below 0.1. Confirm that the sum of the flow rate, liquid level, and water quality weights for each monitoring point is 1. Multiply the pre-processed flow rate data for each monitoring point by the corresponding flow rate weight to obtain the flow rate weighted contribution value. Multiply the pre-processed liquid level data by the corresponding liquid level weight to obtain the liquid level weighted contribution value. The pre-processed water quality data is multiplied by the corresponding water quality weight to obtain the water quality weighted contribution value. The flow-weighted contribution value, level-weighted contribution value, and water quality weighted contribution value of the monitoring point are integrated to form the fusion parameter for that monitoring point. After calculating the fusion parameters for all monitoring points, the results are compared with the theoretical parameter values ​​output by the mechanical seepage well flow coupling model. If the deviation between the fusion parameter and the theoretical parameter value for a monitoring point is less than 3%, the fusion parameter is retained. If the deviation is between 3% and 5%, the weight settings for that monitoring point are checked and fine-tuned before recalculating the fusion parameter. If the deviation exceeds 5%, the hydrogeological basic parameters and pre-processed data for that monitoring point are re-verified, corrected, and the fusion parameter is recalculated. All fusion parameters that meet the requirements are summarized to generate a conflict-free fusion parameter set.

[0091] The method of troubleshooting using the normal operating thresholds of the sensors involves, in practice, first defining the normal operating thresholds for each type of sensor. The normal operating threshold for flow sensors is set to 0 to 5000 cubic meters per day. Flow data exceeding this range is initially considered a potential sensor malfunction. The normal operating threshold for level sensors is set to a level fluctuation of no more than ±0.5 meters per hour. When the level data fluctuation exceeds 0.5 meters per hour, the level sensor is initially considered a potential malfunction. The normal operating threshold for water quality sensors needs to be determined in conjunction with the baseline pollutant concentration values ​​under normal operating conditions at each monitoring point. The normal operating threshold for water quality sensors is set to ±30% of this baseline value. When the pollutant concentration data collected by the water quality sensor exceeds this range, the water quality sensor is initially considered a potential malfunction.

[0092] In practice, this involves comparing real-time collected data with corresponding thresholds at a fixed frequency. Threshold comparisons are performed every five minutes for flow rate data, every minute for liquid level data, and every hour for water quality data. During comparisons, the sensor identifiers corresponding to data exceeding the thresholds are recorded. The collection time of data exceeding the thresholds is also recorded during comparisons.

[0093] In practice, a single instance of data exceeding the threshold is not directly considered a fault. Instead, the sensor continuously monitors the data collected in the next three data sessions. If two or more of the subsequent three data sessions exceed the normal operating threshold, a preliminary assessment of the sensor's fault is made. If only one of the subsequent three data sessions exceeds the threshold, or if all data return to within the normal threshold, it is considered a transient anomaly and not a sensor fault.

[0094] In practice, the judgment is aided by the sensor's installation location and the type of aquifer. Within karst channels of carbonate rock fractured karst aquifers, the flow sensor occasionally exceeds the threshold for short periods. If other similar sensors in the vicinity show normal data, five more data points are monitored. Only if three or more of these five data points exceed the normal threshold is the flow sensor preliminarily determined to be faulty.

[0095] Specifically, the implementation involves compiling information on sensors initially identified as faulty. This compilation includes: sensor identification; sensor installation location; type of aquifer where the sensor is located; specific data exceeding thresholds; and a record of the entire comparison process. This compiled information is used for further investigation into the cause of the sensor fault.

[0096] The process involves determining the cause of anomalies based on the permeability characteristics of fractured cohesive soil and model calculations, and then selecting valid parameter data. Specifically, this is done by first retrieving the fused parameter data after sensor fault diagnosis. Simultaneously, the baseline value of the permeability coefficient of fractured cohesive soil at the corresponding monitoring point is obtained. The baseline value of the permeability coefficient of fractured cohesive soil ranges from 0.467 to 12.274 m / d.

[0097] Define the normal range for abrupt changes in the permeability coefficient of the fractured cohesive soil at this monitoring point. The normal range for abrupt changes in the permeability coefficient is set to not exceed 20% of its baseline value.

[0098] Check for any anomalies in the flow rate or liquid level data in the fusion parameters. If the flow rate data deviates from the normal fluctuation range, or the liquid level data shows unexpected rises or falls, first check the actual permeability coefficient of the fractured cohesive soil at that monitoring point.

[0099] If the actual permeability coefficient change is within the normal range for abrupt changes, the cause of the data anomaly is determined to be a change in the permeability characteristics of fractured cohesive soil. If the actual permeability coefficient does not change, or the change exceeds the normal range for abrupt changes, the process proceeds to comparing it with the model calculation value.

[0100] Retrieve the theoretical calculation values ​​for the corresponding monitoring points from the coupled mechanics-seepage-well flow model. The theoretical calculation values ​​include theoretical flow rate and theoretical liquid level. Taking Xiaoqinghe East Station as an example, its theoretical dewatering capacity is 94.13 m³. 3 / d.

[0101] Calculate the difference between the flow rate data in the fusion parameters and the theoretical flow rate value. Calculate the difference between the liquid level data in the fusion parameters and the theoretical liquid level value.

[0102] Calculate the percentage of each difference relative to the corresponding theoretical value. If the percentage is within ±5%, the parameter data is considered valid hydrological anomaly data.

[0103] If the percentage exceeds ±5%, first check the basic hydrogeological parameter records for that monitoring point. Basic hydrogeological parameters include stratigraphic type and permeability coefficient, etc. If errors are found in the records, correct them and recalculate the percentage difference.

[0104] If the basic hydrogeological parameters are recorded accurately, review the sensor threshold comparison records for that monitoring point. Confirm whether there are any transient anomalies that were not detected previously. If any undetected transient anomalies are found, remove the data for that parameter.

[0105] If no transient anomalies are found, check the data preprocessing procedure for that monitoring point. Verify that the Kalman filter algorithm was applied correctly. Also check if non-continuous anomalies caused by karst water backflow were effectively removed.

[0106] If there are problems in the preprocessing, correct them and regenerate the fusion parameters. Then compare the new fusion parameters with the theoretically calculated values ​​of the model and recalculate the percentage difference.

[0107] If the new percentage difference is within ±5%, it is considered valid hydrological anomaly data. If it still exceeds ±5%, the accuracy of the geological and hydrological characteristic parameters of the area should be further checked. After updating the parameters, the data should be recalculated and reassessed.

[0108] All valid parameter data are compiled. The cause of any anomaly for each valid parameter data point is also recorded. Anomalies are categorized into two types: changes in the permeability characteristics of fractured cohesive soil and actual hydrological anomalies. This process ultimately forms the valid parameter dataset.

[0109] The output valid parameters are used to adjust the precipitation reinjection scheme. Based on these valid parameters, faulty sensors are calibrated in reverse, forming a closed-loop measurement process. Specifically, in implementation, the valid parameter data is first categorized according to well groups 1 through 7 and stations such as Xiaoqinghe East Station and Peijiaying Station. Each category is labeled with the aquifer type corresponding to the monitoring point. Aquifer types include loose rock porous aquifers and carbonate rock fractured karst aquifers.

[0110] Specifically, in implementation, for monitoring points in the loose rock porous aquifer areas of well groups 1 through 7, if the effective flow rate data is lower than the design flow rate value of the dewatering and irrigation scheme, the operating power of the pumping pump for that well group will be adjusted. If the effective flow rate data is higher than the design flow rate value of the dewatering and irrigation scheme, the operating frequency of the pumping pump will be appropriately reduced.

[0111] Specifically, in practice, for monitoring points in the carbonate rock fractured karst aquifer areas of well groups 1 through 7, if the effective liquid level data is higher than the theoretical liquid level value of the mechanical-flow coupling model, the reinjection volume for that well group will be reduced. If the effective liquid level data is lower than the theoretical liquid level value, the reinjection volume for that well group will be increased.

[0112] Specifically, in practice, if the effective flow data at the Xiaoqinghe East Station deviates from the theoretical discharge capacity of 94.13 m³, 3 / d. First, check the pipeline connections of the well groups within the site. After confirming there are no pipeline leaks, adjust the operating parameters of the pumping equipment to bring the actual flow rate closer to the theoretical drainage capacity.

[0113] Specifically, in practice, if the valid water quality data from Peijiaying station shows that the pollutant concentration exceeds the normal benchmark value, the filter components of the water purification device at that station will be replaced. Water quality data will then be collected again after replacement. It will be confirmed that the pollutant concentration has returned to the normal range.

[0114] Specifically, for the liquid level sensor initially determined to be faulty, the effective liquid level data of its corresponding monitoring point is retrieved. If the monitoring point is located in a cohesive soil area, the permeability coefficient of the cohesive soil is between 0.467 and 12.274 m / d. The zero point of the sensor is corrected based on the effective liquid level data. After correction, the sensor's acquisition accuracy is tested to ensure an accuracy of ±1 cm.

[0115] In practice, for flow sensors initially diagnosed as faulty, their collected data is compared with the effective flow data at the corresponding monitoring point. The percentage difference between the two is calculated. If the percentage difference exceeds ±2%, the sensor's sensitivity parameters are adjusted. Data is then collected again after adjustment until the percentage difference is controlled within ±2%.

[0116] Specifically, in practice, for water quality sensors initially diagnosed as faulty, the sensor's detection range is calibrated using the pollutant concentration benchmark value from the valid water quality data at the corresponding monitoring point. A water quality sample is then collected after calibration. The detection results are compared with the valid water quality data to ensure the deviation is within acceptable limits.

[0117] In practice, after calibrating the faulty sensors, these sensors are reconnected to the intelligent sensing system. The system then re-acquires data. The acquired data enters the preprocessing stage, where filtering and outlier data removal operations are performed again.

[0118] In practice, the preprocessed new data undergoes multi-source data fusion. The fusion process follows the original fusion rules, generating a new set of fusion parameters. The validity of this new set of parameters is then assessed, and a new round of valid parameter data is selected.

[0119] In practice, this involves updating the parameters of the mechanical-to-seepage well-flow coupling model after each data acquisition and calibration cycle. The updated parameters include the permeability coefficients and formation characteristics of each aquifer group, making the model more closely reflect the actual conditions at the current engineering site.

[0120] In practice, this involves a closed-loop measurement process—from data acquisition and processing to judgment and calibration—through such a cycle. This ensures that the effective parameters output each time accurately guide adjustments to the precipitation reinjection plan, while also guaranteeing the accuracy of subsequent data collection.

[0121] The present invention also provides a subway precipitation reinjection parameter measurement system based on an intelligent sensing system, comprising: sensors, which are arranged in multiple well groups and stations along the subway precipitation reinjection project, and flow rate, liquid level and water quality sensors are arranged according to different water-bearing rock groups;

[0122] The data acquisition unit is used to align the acquisition time of each sensor through a unified time synchronization module, acquire data at the corresponding frequency to synchronously record basic hydrogeological parameters, and convert the raw data into standard data frames to form an initial parameter dataset.

[0123] The normalization processing unit is used to sort out the accuracy levels of each sensor in the initial data, determine a unified error range, and map the error range of different types of data to the same range through linear transformation.

[0124] The filtering unit is used to incorporate the noise characteristics of the construction environment using the Kalman filtering algorithm and dynamically adjust parameters to filter out interference noise.

[0125] The anomaly removal unit is used to analyze the characteristics of karst water backwater recharge, check the filtered data, identify and remove non-persistent anomalies to form a preprocessed dataset.

[0126] The fusion rule unit is used to construct mechanical seepage well flow. The coupled model combines geological and hydrological characteristics to set the weights of each parameter according to the measured flow range, formation permeability coefficient and water quality correlation, and dynamically corrects them.

[0127] The weighted integration unit is used to retrieve preprocessed data and hydrological parameters, calculate the weighted contribution value of each parameter according to the integration rules, and integrate them to form a fused parameter set.

[0128] The fault diagnosis unit is used to determine the normal operating threshold of each sensor, compare the real-time data with the threshold at a fixed frequency, record the situations that exceed the threshold, and determine the sensor fault after continuous monitoring.

[0129] The anomaly analysis unit is used to retrieve the fusion parameters and permeability coefficient baseline values ​​after fault diagnosis, check the actual permeability coefficient changes, calculate the difference percentage based on the theoretical values ​​of the coupling model, and filter valid parameter data.

[0130] The scheme adjustment unit is used to classify effective parameters according to well group sites and rock group types, and adjust the precipitation reinjection scheme based on flow rate, liquid level and water quality.

[0131] The sensor calibration unit is used to correct faulty level sensors based on valid parameters, adjust the sensitivity of flow sensors at zero point, and calibrate the detection range of water quality sensors.

[0132] The closed-loop update unit is used to reconnect the calibrated sensor to the system, and re-execute the acquisition preprocessing fusion and validity determination process. The coupling model parameters are updated every time the process is completed.

[0133] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0134] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for measuring subway precipitation reinjection parameters based on an intelligent sensing system, characterized in that, Including the following steps: Along the subway rainwater recharge project, flow, liquid level and water quality sensors were deployed according to different water-bearing rock groups at multiple well groups and stations to collect data synchronously, record basic hydrogeological parameters and initially unify the data format to form an initial parameter dataset. The initial data was normalized according to the accuracy level, and a filtering algorithm was used to remove interference noise from the construction environment. Non-continuous abnormal data were removed based on the characteristics of karst water top support replenishment, resulting in a preprocessed dataset. A fusion rule is established based on the coupled mechanical, seepage, and wellflow models and geological and hydrological characteristics. Specifically, a coupled mechanical-seepage-wellflow model is first constructed, incorporating the geological structure, rock group mechanical properties, seepage parameters, and wellflow parameters along the subway dewatering recharge project route. Then, based on the model output and the on-site geological and hydrological characteristics, flow rate weights are set according to the measured flow range, liquid level weights are set according to the formation permeability coefficient, and water quality weights are set according to the correlation between water quality and flow rate and liquid level, with the sum of the three weights being 1. Finally, the weights are dynamically adjusted based on the real-time theoretical values ​​of the model, and the geological and hydrological parameters are re-evaluated and the fusion rule is updated when the deviation exceeds the limit. Preprocessed data is weighted and integrated according to rules to generate a conflict-free fusion parameter set; To troubleshoot, the sensors were compared against their normal operating thresholds. Based on the permeability characteristics of fractured cohesive soil and model calculations, the causes of anomalies were determined, and valid parameter data was selected. Specifically, the fused parameters after sensor troubleshooting and the baseline permeability coefficient of the corresponding monitoring point were retrieved. The normal range for abrupt changes in permeability coefficient was determined. If the flow rate or liquid level was abnormal, the actual permeability coefficient was checked first. If it was within the normal range, the anomaly was determined to be due to a change in permeability characteristics; otherwise, the theoretical value of the mechanical-seepage-well flow coupling model was retrieved. The percentage difference between the parameter and the theoretical value was calculated. If it was within the allowable range, it was considered valid hydrological anomaly data. If it exceeded the range, the basic hydrogeological parameters, sensor records, preprocessing procedures were checked, and the judgment was corrected and re-judged. Valid parameters were summarized, and the causes of anomalies were recorded to form a valid parameter dataset. The output of valid parameters is used to adjust the precipitation reinjection scheme, and the faulty sensor is calibrated in reverse based on the valid parameters, forming a closed-loop measurement process.

2. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The process involves deploying flow, level, and water quality sensors along multiple well groups and at various stations along the subway dewatering recharge project route, according to the distribution of different aquifer rock groups. Specifically, this involves: firstly, surveying the well groups and stations along the subway dewatering recharge project route to determine and differentiate the types of aquifer rock groups; secondly, deploying flow, level, and water quality sensors at corresponding locations according to the rock group type, with different sensor installation locations and densities for different rock groups; thirdly, recording the sensor locations, the rock groups they are located in, and the surrounding strata after installation; and finally, selecting appropriate sensor types based on the subway construction environment and accuracy requirements.

3. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The synchronously acquired data records basic hydrogeological parameters and initially unifies the data format to form an initial parameter dataset. Specifically, each sensor aligns its acquisition time through a unified time synchronization module. Collect flow rate, liquid level, and water quality data at the corresponding frequency; simultaneously record the basic hydrogeological parameters of the monitoring point, such as stratum type, permeability coefficient, and porosity; convert the raw data into standard data frames containing sensor ID, timestamp, parameter type identifier, numerical value and unit, and check digit in a unified format; and summarize them to form an initial parameter dataset.

4. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The initial data normalization process according to accuracy level is as follows: First, sort out the accuracy level of each sensor in the initial parameter dataset; determine a unified error range and map the error range of all sensor data to the same range; for flow rate, liquid level, and water quality data, calculate the error range according to their accuracy, and then map them to the unified range through linear transformation.

5. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The process involves using a filtering algorithm to remove construction environment noise and eliminating non-continuous abnormal data based on the characteristics of karst water backwater replenishment to obtain a preprocessed dataset. Specifically, the initial parameter data is processed using a Kalman filter algorithm, and the characteristics of subway construction environment noise are incorporated into the equations. The filtering parameters are dynamically adjusted to filter out interference. The hydrological dynamic characteristics of karst water backwater replenishment are analyzed, and the backwater replenishment range is determined based on geological surveys along the line. The filtered data is checked, and non-continuous abnormal data is identified and eliminated. The remaining data are then compiled to form the preprocessed dataset.

6. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The step of weighted integration of preprocessed data according to rules to generate a conflict-free fusion parameter set involves: retrieving preprocessed data and corresponding basic hydrogeological parameters; determining the weights of flow rate, level, and water quality for each monitoring point based on the design flow range of the irrigation and drainage scheme, the formation permeability coefficient, and the correlation between water quality and flow rate / level, ensuring that the sum of the weights is 1; multiplying the preprocessed parameter data by their corresponding weights to obtain weighted contribution values, and integrating them to form the fusion parameters for each monitoring point; comparing the theoretical parameter values ​​of the coupling model, adjusting and retaining the fusion parameters that meet the requirements according to the deviation, and summarizing to generate a conflict-free fusion parameter set.

7. The method for measuring subway precipitation reinjection parameters based on an intelligent sensing system according to claim 1, characterized in that, The method for troubleshooting by comparing the normal operating threshold of the sensors is as follows: First, determine the normal operating threshold of each type of sensor; compare the real-time collected data with the corresponding threshold at a fixed frequency and record the sensor identifier and collection time of the data exceeding the threshold; do not directly determine the fault for a single instance of exceeding the threshold, but continuously monitor subsequent data and determine whether it is a fault or an instantaneous anomaly based on the number of times; The sensor's installation location and the type of aquifer in which it is located are used to assist in the judgment; the sensor identification, installation location, type of aquifer in which it is located, specific data exceeding the threshold, and comparison records of the sensors that were initially determined to be faulty are summarized.

8. A system for measuring subway precipitation reinjection parameters based on an intelligent sensing system as described in claim 1, characterized in that, include: Sensors were deployed at multiple well groups and stations along the subway dewatering and recharge project, with flow rate, liquid level and water quality sensors arranged according to different aquifer groups. The data acquisition unit is used to align the acquisition time of each sensor through a unified time synchronization module, acquire data at the corresponding frequency to synchronously record basic hydrogeological parameters, and convert the raw data into standard data frames to form an initial parameter dataset. The normalization processing unit is used to sort out the accuracy levels of each sensor in the initial data, determine a unified error range, and map the error range of different types of data to the same range through linear transformation. The filtering unit is used to incorporate the noise characteristics of the construction environment using the Kalman filtering algorithm and dynamically adjust parameters to filter out interference noise. The anomaly removal unit is used to analyze the characteristics of karst water backwater recharge, check the filtered data, identify and remove non-persistent anomalies to form a preprocessed dataset. The fusion rule unit is used to construct mechanical seepage well flow. The coupled model combines geological and hydrological characteristics to set the weights of each parameter according to the measured flow range, formation permeability coefficient and water quality correlation, and dynamically corrects them. The weighted integration unit is used to retrieve preprocessed data and hydrological parameters, calculate the weighted contribution value of each parameter according to the integration rules, and integrate them to form a fused parameter set. The fault diagnosis unit is used to determine the normal operating threshold of each sensor, compare the real-time data with the threshold at a fixed frequency, record the situations that exceed the threshold, and determine the sensor fault after continuous monitoring. The anomaly analysis unit is used to retrieve the fusion parameters and permeability coefficient baseline values ​​after fault diagnosis, check the actual permeability coefficient changes, calculate the difference percentage based on the theoretical values ​​of the coupling model, and filter valid parameter data. The scheme adjustment unit is used to classify effective parameters according to well group sites and rock group types, and adjust the precipitation reinjection scheme based on flow rate, liquid level and water quality. The sensor calibration unit is used to correct faulty level sensors based on valid parameters, adjust the sensitivity of flow sensors at zero point, and calibrate the detection range of water quality sensors. The closed-loop update unit is used to reconnect the calibrated sensor to the system, and re-execute the acquisition preprocessing fusion and validity determination process. The coupling model parameters are updated every time the process is completed.

Images