Hydrological condition monitoring fault early warning method and system based on artificial intelligence

Abstract

The invention relates to the technical field of fault warning, in particular to a hydrological condition monitoring fault early warning method and system based on artificial intelligence, and the method comprises the steps: building a monitoring value linkage model between a calibration unit and each hydrological monitoring unit; obtaining a calibration monitoring value measured by the calibration unit, and calculating a prediction detection value of each hydrological monitoring unit based on the calibration monitoring value and a monitoring value linkage model; calculating a deviation value between the predicted detection value and a real-time monitoring parameter actually detected by each hydrological monitoring unit so as to obtain historical monitoring data of the target hydrological monitoring unit and the closest hydrological monitoring unit of the same type; acquiring data fluctuation characteristics based on the change degree of the historical monitoring data; and the data fluctuation characteristics of the current monitoring unit and the closest hydrological monitoring unit of the same type are compared to judge that the target monitoring unit fails, so that the fault condition of the monitoring units can be judged more accurately.

Description

Technical Field

[0001] The present invention relates to the field of fault alarm technology, and in particular to an artificial intelligence-based hydrological condition monitoring fault early warning method and system. Background Art

[0002] Hydrological monitoring involves comprehensive observation and analysis of the physical, chemical and biological characteristics of rivers, lakes, reservoirs and groundwater. It is an important means to understand the status of water resources, ensure flood control safety, promote agricultural irrigation, assist in the planning and operation management of water conservancy projects, and protect the environment. This process includes determining the velocity and volume of water flow to calculate flow, assessing water quality (such as pollutant concentration, dissolved oxygen, pH value and temperature) to ensure compliance with environmental quality standards, using water level gauges to monitor water level changes to warn of flood risks, recording precipitation and its distribution to grasp the dynamics of water resource recharge in the basin, measuring water surface evaporation rate to study regional water resource cycles, and paying attention to the water level and water quality of groundwater resources.

[0003] Existing hydrological monitoring requires the installation of multiple monitoring units in the monitoring area. Due to the complex outdoor environment, there are various influences such as temperature, humidity, atmospheric pressure, and animal interference, which makes the monitoring unit failure probability higher. The existing method of judging whether a monitoring unit is faulty by the extreme value of the monitoring data of a single monitoring unit is relatively simple and cannot accurately judge the situation where the error of the monitoring unit becomes larger but does not reach the extreme value, thus posing a monitoring risk. Summary of the invention

[0004] The purpose of the present invention is to provide a hydrological condition monitoring fault warning method and system based on artificial intelligence, so as to be able to judge the error situation of the monitoring unit in combination with reference data and historical data, so as to more accurately judge the fault situation of the monitoring unit.

[0005] To achieve the above-mentioned object, in a first aspect, the present invention provides a hydrological situation monitoring fault early warning method based on artificial intelligence, comprising establishing a monitoring numerical linkage model between a calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters, wherein the environmental parameters include temperature, humidity, and altitude data, and the calibration unit is arranged between multiple hydrological monitoring units;

[0006] Obtaining calibration monitoring values ​​measured by the calibration unit entering the monitoring environment at preset time intervals, and calculating predicted detection values ​​of each hydrological monitoring unit based on the calibration monitoring values ​​and the monitoring numerical linkage model;

[0007] Calculate the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, and determine whether the deviation is within the predetermined target range. When the deviation is not within the target range, obtain the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type;

[0008] Based on the degree of change of historical monitoring data, a first data fluctuation characteristic of the target monitoring unit and a second data fluctuation characteristic of the closest hydrological monitoring unit of the same type are established;

[0009] Compare the first data fluctuation feature of the current monitoring unit with the second data fluctuation feature of the nearest hydrological monitoring unit of the same type, and if the difference value exceeds a preset threshold, determine that the target monitoring unit is faulty;

[0010] Obtain characteristic information of faulty hydrological monitoring units and notify relevant personnel;

[0011] Classify the fault types of the hydrological monitoring units based on the characteristic information;

[0012] Match the replacement time period of the hydrological monitoring unit according to the fault type.

[0013] Among them, the types of hydrological monitoring units include rain gauges, flow meters and water quality analyzers.

[0014] The specific steps of establishing a monitoring numerical linkage model between the calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters include:

[0015] Calculate the placement position of the calibration unit based on the position information of all hydrological monitoring units;

[0016] After placing the calibration unit at the placement position, collecting calibration data at preset time intervals;

[0017] Obtain monitoring data of all hydrological monitoring units with the same timestamp as the calibration data;

[0018] The relationship between calibration data and monitoring data is obtained by using polynomial regression analysis, and a monitoring numerical linkage model is obtained.

[0019] The specific steps of using polynomial regression analysis to obtain the relationship between calibration data and monitoring data and to obtain a monitoring numerical linkage model include:

[0020] Obtain temperature, humidity, and atmospheric pressure data and use them as input features to predict the difference between calibration data and monitoring data;

[0021] Split the data into training and testing sets;

[0022] Use the linear regression algorithm to fit a polynomial regression model on the training set;

[0023] The test set is used to test and adjust the polynomial regression model to obtain the monitoring numerical linkage model.

[0024] The specific steps of collecting calibration data at preset time intervals after placing the calibration unit at the placement position include:

[0025] Set a calibration cycle;

[0026] Allow the calibration unit to run in the actual environment for a preset period of time before collecting calibration data.

[0027] The specific steps of calculating the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, judging whether the deviation is within a predetermined target range, and obtaining the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type when the deviation is not within the target range include:

[0028] Obtain predicted detection values ​​and real-time monitoring parameters at corresponding time points;

[0029] Remove outliers and missing values ​​from real-time monitoring parameters, and remove predicted detection values ​​at corresponding time points;

[0030] Calculate the deviation between the predicted detection value and the real-time monitoring parameter at the corresponding time point;

[0031] When the deviation value is not within the target range, the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type are obtained.

[0032] The specific steps of obtaining characteristic information of the faulty hydrological monitoring unit and notifying relevant personnel include:

[0033] Extracting basic information of the faulty hydrological monitoring unit from a database, the basic information including equipment model, installation location, operation time, and maintenance history;

[0034] Generate notification information based on basic information;

[0035] Use multiple communication channels to deliver notification information to relevant processing personnel.

[0036] The specific steps of classifying the fault types of the hydrological monitoring unit based on the characteristic information include:

[0037] Collect historical operating status data from hydrological monitoring units, including data under normal operating conditions as well as data under known fault conditions;

[0038] Extract fault classification features based on historical working status data to obtain a training data set, wherein the fault classification features include temperature, humidity, flow rate, and pressure;

[0039] The support vector machine model is trained using the training data set to obtain a discrimination model;

[0040] The fault type is obtained based on the discrimination model and the fault classification feature data of the currently detected hydrological monitoring unit.

[0041] In a second aspect, the present invention further provides a hydrological situation monitoring fault warning system, including a calibration unit, a plurality of hydrological monitoring units, a linkage model generation module, a detection value prediction module, a first deviation calculation module, a data feature calculation module, a second deviation calculation module and a fault alarm module;

[0042] The linkage model generation module is used to establish a monitoring numerical linkage model between the calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters, wherein the environmental parameters include temperature, humidity, and altitude data, and the calibration unit is arranged between multiple hydrological monitoring units;

[0043] The detection value prediction module is used to obtain the calibration monitoring value measured by the calibration unit entering the monitoring environment every preset time period, and calculate the predicted detection value of each hydrological monitoring unit based on the calibration monitoring value and the monitoring value linkage model;

[0044] The first deviation calculation module is used to calculate the deviation value between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, and determine whether the deviation value is within a predetermined target range. When the deviation value is not within the target range, the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type are obtained;

[0045] The data feature calculation module is used to establish a first data fluctuation feature of a target monitoring unit and a second data fluctuation feature of a hydrological monitoring unit of the same type that is closest to the target monitoring unit based on the degree of change of historical monitoring data;

[0046] The second deviation calculation module is used to compare the first data fluctuation characteristic of the current monitoring unit with the second data fluctuation characteristic of the nearest hydrological monitoring unit of the same type, and if the difference value exceeds a preset threshold, it is determined that the target monitoring unit is faulty;

[0047] The fault alarm module is used to obtain characteristic information of the faulty hydrological monitoring unit and notify relevant personnel.

[0048] The artificial intelligence-based hydrological monitoring fault warning method and system of the present invention establishes a linkage model connecting the calibration unit and multiple distributed hydrological monitoring units under set environmental conditions (such as specific temperature, humidity and altitude). This model allows the values ​​between various monitoring points to be correlated with each other, so as to more accurately reflect the hydrological changes in the area. The calibration unit is strategically placed between different hydrological monitoring units to ensure the consistency and accuracy of the entire monitoring network.

[0049] The calibration unit enters the actual monitoring environment at preset time intervals to conduct measurements and record these calibration monitoring values. Subsequently, the monitoring value linkage model established in advance is combined with the latest calibration monitoring values ​​to calculate the expected detection value of each hydrological monitoring unit. This process allows us to predict the theoretical readings of each monitoring point, providing a basis for subsequent deviation analysis.

[0050] The predicted detection value calculated above is compared with the actual monitoring data reported in real time by the hydrological monitoring unit, and the deviation between the two is calculated. If the deviation of a monitoring unit is found to exceed the pre-defined target range, it indicates that there may be an abnormal situation. At this time, the system will further analyze and establish the data fluctuation characteristics of each by obtaining the historical monitoring data of the problem monitoring unit and its neighboring monitoring units of the same type.

[0051] By comparing the first data fluctuation feature of the target monitoring unit with the second data fluctuation feature of its nearest neighbor unit of the same type, the difference between the two can be evaluated more carefully. If the difference exceeds the set threshold, it can be reasonably suspected that the target monitoring unit has a fault. This step effectively improves the accuracy of fault judgment and avoids false alarms due to abnormal data at a single time point.

[0052] Once a monitoring unit is confirmed to have failed, the system will automatically collect the relevant characteristic information of the unit (such as location, model, fault type, etc.) and immediately notify the relevant maintenance personnel so that they can respond quickly and take necessary repair measures. In this way, the error of the monitoring unit can be judged in combination with reference data and historical data to more accurately judge the fault of the monitoring unit, which not only helps to improve the stability of the hydrological monitoring system, but also greatly reduces the negative impact of faults and ensures the smooth development of water resources management. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 It is a flow chart of a hydrological condition monitoring fault early warning method based on artificial intelligence of the present invention.

[0055] Figure 2 It is a flow chart of the present invention for establishing a monitoring numerical linkage model between a calibration unit and various hydrological monitoring units under the condition of preset environmental parameters.

[0056] Figure 3 The present invention uses polynomial regression analysis to obtain the relationship between calibration data and monitoring data, and obtains a flow chart of a monitoring numerical linkage model.

[0057] Figure 4 It is a flow chart of collecting calibration data at preset time intervals after the calibration unit is placed at a placement position according to the present invention.

[0058] Figure 5 It is a flowchart of the present invention for calculating the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, judging whether the deviation value is within a predetermined target range, and obtaining the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type when the deviation value is not within the target range.

[0059] Figure 6 It is a flow chart of the present invention for obtaining characteristic information of a faulty hydrological monitoring unit and notifying relevant personnel.

[0060] Figure 7 It is a flow chart of classifying the fault types of the hydrological monitoring unit based on the characteristic information of the present invention.

[0061] Figure 8 It is a flow chart of matching the replacement time period of the hydrological monitoring unit according to the fault type of the present invention.

[0062] Fig. 9 It is a structural diagram of a hydrological situation monitoring fault early warning system of the present invention.

[0063] Calibration unit 101 , hydrological monitoring unit 102 , linkage model generation module 103 , detection value prediction module 104 , first deviation calculation module 105 , data feature calculation module 106 , second deviation calculation module 107 , fault alarm module 108 . DETAILED DESCRIPTION

[0064] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0065] First embodiment

[0066] See also Figures 1 to 8 The present invention provides a hydrological condition monitoring fault early warning method based on artificial intelligence, comprising:

[0067] S101 establishes a monitoring numerical linkage model between a calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters, wherein the environmental parameters include temperature, humidity, and altitude data, and the calibration unit is arranged between the plurality of hydrological monitoring units;

[0068] Under the conditions of preset environmental parameters (such as temperature, humidity, and altitude data), a monitoring numerical linkage model between the calibration unit and multiple hydrological monitoring units is established. These hydrological monitoring units include but are not limited to rain gauges, flow meters, and water quality analyzers, which are distributed in different geographical locations and are used to collect various hydrological data. The calibration unit will be strategically placed between these monitoring units to ensure that it can represent the environmental conditions of most monitoring points.

[0069] The specific steps include:

[0070] S201 calculates the placement position of the calibration unit based on the position information of all hydrological monitoring units;

[0071] Collect the precise geographic coordinates (latitude and longitude) of all hydrological monitoring units and convert them into coordinates (x,y) in a plane rectangular coordinate system. If high-precision calculations are required, the effect of the earth's curvature should be considered.

[0072] Choose an initial guess position as the starting position of the calibration unit. This can be the position of any hydrological monitoring unit or the average of all hydrological monitoring unit coordinates.

[0073] Use Weiszfeld's algorithm to find the optimal position. Specifically, set the initial estimate P_0P0 to be the position of the calibration unit.

[0074] For each iteration k, update the position P k The new estimate P k+1 , the formula is as follows:

[0075]

[0076] Here P i represents the location of the i-th hydrological monitoring unit, ‖P k -P i ‖ indicates that from the current estimate P k To monitoring unit P i The Euclidean distance of .

[0077] Check convergence. If ‖P k+1 -P k ‖<∈, where ∈ is a preset small positive number, the iteration stops; otherwise, the iteration continues.

[0078] Finally, verify whether the final position is reasonable and ensure that it is indeed the position that minimizes the sum of distances to all hydrological monitoring units. Cross-validation can be performed by comparing the results obtained under different initialization conditions.

[0079] S202: after the calibration unit is placed at the placement position, calibration data is collected at every preset time period;

[0080] The specific steps include:

[0081] S401 sets a calibration cycle;

[0082] According to the characteristics of the equipment and monitoring needs, set a reasonable calibration cycle (such as daily, weekly or monthly). This cycle should be able to capture changes in a timely manner without being too frequent and increasing the workload.

[0083] S402 allows the calibration unit to run in the actual environment for a preset period of time before collecting calibration data.

[0084] After the calibration unit enters the actual monitoring environment, give it enough time to adapt to the new environmental conditions to reduce errors caused by sudden changes in temperature, humidity and other factors.

[0085] Ensure that the timestamps of the calibration unit and all other hydrological monitoring units are synchronized to facilitate subsequent data comparison and analysis.

[0086] Start the data collection process for the calibration unit and all other hydrological monitoring units during the same time period to ensure that all equipment is measuring under the same conditions to obtain comparable data sets.

[0087] S203 obtains monitoring data of all hydrological monitoring units with the same timestamp as the calibration data;

[0088] At this stage, real-time monitoring parameters with the same timestamp as the calibration data are obtained from all hydrological monitoring units. This step is crucial because it ensures that we are comparing data from the same moment, thus improving the accuracy of the analysis results.

[0089] S204 uses polynomial regression analysis to obtain the relationship between the calibration data and the monitoring data, and obtains a monitoring numerical linkage model.

[0090] The specific steps include:

[0091] S301 obtains temperature, humidity, and atmospheric pressure data and uses them as input features to predict the difference between calibration data and monitoring data;

[0092] The environmental condition information (such as temperature, humidity, and atmospheric pressure) around each monitoring unit is collected and used as input features to predict the difference pattern between the calibration data and the monitoring data.

[0093] S302 divides the data into a training set and a test set;

[0094] The dataset is divided into a training set and a test set at a ratio of 70%-80%, the former is used to train the model, and the latter is used to evaluate the model performance. Ensure that the training set can fully represent the entire data distribution, and the test set is large enough to accurately reflect the generalization ability of the model.

[0095] S303 uses a linear regression algorithm to fit a polynomial regression model on the training set;

[0096] Select the quadratic polynomial order and fit the polynomial regression model on the training set by least squares or other methods. The goal of this step is to find an optimal set of parameters that minimizes the error between the predicted output and the actual observed value.

[0097] S304 uses the test set to test and adjust the polynomial regression model to obtain a monitoring numerical linkage model.

[0098] Use the test set to verify the fitted polynomial regression model and check its prediction accuracy and stability. If the model performance is poor, you need to return to the previous steps and readjust the feature selection, algorithm selection or hyperparameter settings until the most suitable configuration is found. In order to improve the generalization ability of the model, the K-fold cross-validation method can be used to repeatedly train and test the model on different subsets, and finally take the average score as the performance indicator of the model.

[0099] S102 obtains the calibration monitoring value measured by the calibration unit entering the monitoring environment every preset time period, and calculates the predicted detection value of each hydrological monitoring unit based on the calibration monitoring value and the monitoring value linkage model;

[0100] According to the characteristics of the equipment and monitoring needs, set a reasonable calibration cycle (such as daily, weekly or monthly). This cycle should be able to capture potential changes in a timely manner, but not too frequent and increase unnecessary workload.

[0101] Move the calibration unit to the actual monitoring environment according to the predetermined schedule. This step involves physically moving it or adjusting its position by remote control. Let the calibration unit run in the new environment for a period of time (such as a few minutes to half an hour) to adapt it to the new environmental conditions and reduce errors caused by sudden changes in factors such as temperature and humidity. Ensure that the timestamps of the calibration unit and all other hydrological monitoring units are synchronized to facilitate subsequent data comparison and analysis.

[0102] Start the data collection process for the calibration unit and all other hydrological monitoring units in the same time period, ensuring that all devices are measured under the same conditions to obtain comparable data sets. Record all data obtained during this calibration process, including timestamps, geographic locations, measurement results, and other information to provide a basis for subsequent analysis. Call the previously established monitoring numerical linkage model, which has been trained and can accurately predict the behavior of other hydrological monitoring units. Use the data obtained from the current calibration as input and pass it to the linkage model to directly calculate the predicted value.

[0103] S103 calculates the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, and determines whether the deviation is within a predetermined target range. When the deviation is not within the target range, obtains historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type;

[0104] The specific steps include:

[0105] S501 obtains the predicted detection value and the real-time monitoring parameters at the corresponding time point;

[0106] Ensure that the calibration unit and all other hydrological monitoring units collect data in the same time period to obtain data sets with the same timestamp. Extract the predicted detection value and corresponding real-time monitoring parameters of each hydrological monitoring unit at a specific time point from the database. This includes but is not limited to key indicators such as temperature, humidity, precipitation, flow rate, etc.

[0107] S502 removes abnormal values ​​and missing values ​​in the real-time monitoring parameters, and removes the predicted detection values ​​at the corresponding time points;

[0108] Use statistical methods (such as Z-score, IQR) to identify and mark outliers. Data points that are obviously wrong should be deleted. For missing data points, interpolation methods (such as linear interpolation, spline interpolation) or the average value of adjacent data points can be used to fill them. At the same time, remove the predicted detection values ​​of time points that cannot be reliably filled.

[0109] S503 calculates the deviation value between the predicted detection value and the real-time monitoring parameter at the corresponding time point;

[0110] Select percentage error to calculate the deviation value. For each hydrological monitoring unit, batch calculate the deviation value between its predicted detection value and the actual detection value. Save these deviation values ​​and their corresponding timestamps for subsequent analysis. You can use charts (such as line charts, scatter plots, heat maps, etc.) to intuitively display the relationship between the predicted value and the actual value, helping to more clearly understand the similarities and differences between the two.

[0111] S504: When the deviation value is not within the target range, historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type are obtained.

[0112] Based on historical data, domain knowledge or expert advice, set a reasonable "normal" range (for example, ±5%) for each monitoring parameter. This range can distinguish between normal fluctuations and abnormal situations. For each hydrological monitoring unit, compare its deviation value with the preset target range one by one. When it is found that the deviation value of a monitoring unit exceeds the preset range, immediately mark the unit as a "potential fault" and record the specific degree of deviation and time point. Using geographic information system (GIS) tools, calculate the spatial distance between the target hydrological monitoring unit and other monitoring units of the same type, and find the closest unit. Extract all monitoring records of these two units in the past period of time (for example, the past week or month) from the database. Ensure that the data covers measurements in different seasons, weather conditions and working environments.

[0113] S104 establishes a first data fluctuation feature of the target monitoring unit and a second data fluctuation feature of the closest hydrological monitoring unit of the same type based on the degree of change of the historical monitoring data;

[0114] Extract all monitoring records of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type in the past period of time (e.g., the past week or month) from the database. Ensure that the data covers measurements in different seasons, weather conditions, and working environments.

[0115] Calculate the amount of change in data between adjacent time periods (such as differences) to capture the trend of data changes over time. Calculate moving averages of different window sizes (such as 7 days, 30 days) to smooth short-term fluctuations and highlight long-term trends. Use linear regression or other trend analysis methods to evaluate the trend of data changes over time and identify rising, falling or stable patterns.

[0116] Based on the above analysis results, a series of key features that can reflect the fluctuation characteristics of the target monitoring unit data are defined, including but not limited to the average rate of change, fluctuation amplitude, cycle length, trend slope, etc.

[0117] Each feature is quantified into a specific value to form a feature vector, which will be used for subsequent comparative analysis.

[0118] Repeat the above process: perform the same process on the closest hydrological monitoring unit of the same type to extract its second data fluctuation characteristics.

[0119] S105 compares the first data fluctuation characteristic of the current monitoring unit with the second data fluctuation characteristic of the nearest hydrological monitoring unit of the same type, and if the difference value exceeds a preset threshold, determines that the target monitoring unit is faulty;

[0120] In this step, we will conduct a detailed comparative analysis of the first data fluctuation characteristics of the current monitoring unit (target monitoring unit) and the second data fluctuation characteristics of the closest hydrological monitoring unit of the same type. By quantifying the difference between the two and comparing it with the preset threshold, it is possible to effectively identify whether the target monitoring unit has a fault. The following are the detailed implementation steps:

[0121] According to the actual situation, the cosine similarity index is selected to quantify the difference between the two feature vectors. For each feature dimension, the difference values ​​between the target monitoring unit and the adjacent units of the same type are calculated in batches. These difference values ​​and their corresponding timestamps are saved for subsequent analysis.

[0122] Based on historical data, determine a reasonable preset threshold that can distinguish normal fluctuations from abnormal situations, but not be too sensitive to cause frequent false alarms.

[0123] The difference value at each time point is compared with the preset threshold one by one. If the difference value at a certain time point is found to exceed the preset threshold, the time point is immediately marked as the moment of "potential failure" of the target monitoring unit, and the specific deviation degree and time point are recorded.

[0124] For the time points marked as "potential failures", the data performance in the period before and after is further comprehensively evaluated to eliminate the impact of randomness and short-term fluctuations. If multiple consecutive time points exceed the threshold, it is more likely to be considered that there is indeed a failure.

[0125] S106 obtains characteristic information of the faulty hydrological monitoring unit and notifies relevant personnel.

[0126] The specific steps include:

[0127] S601 extracts basic information of the faulty hydrological monitoring unit from a database, wherein the basic information includes equipment model, installation location, operation time, and maintenance history;

[0128] Equipment Model Records the specific model and version number of the hydrological monitoring unit, which is very important for evaluating the performance of the equipment and selecting appropriate maintenance methods. Installation Location Obtains the exact geographic location of the unit (longitude, latitude) and a description of the physical environment in which it is located (such as river name, monitoring station number). This helps field engineers quickly locate the device.

[0129] Run time counts the length of time the unit has been in operation since installation, including total run time and time since the most recent restart. Long periods of continuous operation can cause equipment aging or wear. Organize the unit's maintenance history, including the last calibration date, replacement parts list, fault repair log, etc. This information can help analyze the cause of the current fault and provide a reference for future preventive maintenance.

[0130] S602 generates notification information based on basic information;

[0131] Design a standardized notification template, including title, body, attachments, etc. Make sure the notification content is concise and clear while covering all key information. Automatically populate variable fields in the notification template based on information extracted from the database, such as equipment model, installation location, operating hours, maintenance history, etc. If necessary, add additional instructions or suggestions to the notification, such as preliminary diagnostic results, tools and spare parts list to bring, etc.

[0132] S603 uses multiple communication channels to deliver the notification information to relevant processing personnel.

[0133] In case of emergency, a short notification message is sent to the designated contact via the SMS platform to ensure immediate attention.

[0134] Use the company's internal mail system or external service to send a formal notification email with a detailed fault report and supporting documents. This method is suitable for delivering more complex information and documents.

[0135] Use enterprise-level instant messaging software (such as WeChat, DingTalk, Slack, etc.) to create a dedicated fault response group to share the latest progress and discuss solutions in real time.

[0136] For particularly serious situations, call the person in charge directly to make a verbal notification and confirm that the other party has received and understood the relevant information.

[0137] If your organization has a dedicated mobile app or management platform, you can use the in-app push notification feature to send alerts to relevant personnel, reminding them to check the latest fault information.

[0138] S107 classifies the fault type of the hydrological monitoring unit based on the characteristic information.

[0139] The specific steps include:

[0140] S701 collects historical operating status data from hydrological monitoring units, including data under normal operating conditions and data under known fault conditions;

[0141] Establish a comprehensive historical working status data set, which includes data records of hydrological monitoring units under normal operating conditions and known different fault conditions. These data come from multiple dimensions to ensure the diversity and representativeness of the samples. Specifically, we will collect the following types of data:

[0142] S702 extracts fault classification features based on historical working status data to obtain a training data set, where the fault classification features include temperature, humidity, flow rate, and pressure;

[0143] Next, extract the feature values ​​that can effectively characterize the fault type from the collected historical data. The "fault classification features" mentioned here refer to the key indicators that best reflect the fault characteristics, such as temperature, humidity, flow rate and pressure. Determine which features are the most important through statistical analysis, signal processing technology or the experience of domain experts, and use them as input variables for subsequent model training.

[0144] We combine the extracted features into feature vectors, each of which represents a snapshot of the state of the hydrological monitoring unit at a specific moment. We then label these feature vectors according to whether they correspond to a fault state (for example, 0 for normal and 1 for a certain type of fault), thus forming our training data set.

[0145] S703 uses the training data set to train the support vector machine model to obtain a discrimination model;

[0146] With the prepared training data set, you can start training the support vector machine model. SVM is a supervised learning method that excels at handling classification problems in high-dimensional space. It finds an optimal hyperplane that maximizes the interval between two types of data to distinguish different fault categories. During the training process, the model parameters are adjusted to optimize the classification effect until a satisfactory accuracy is achieved.

[0147] S704 obtains the fault type based on the resolution model and the fault classification feature data of the currently detected hydrological monitoring unit.

[0148] After completing the model training, we obtained a fully verified discrimination model. When a new hydrological monitoring unit has a suspected fault, we can collect its current feature data in real time and input it into this discrimination model for prediction. The model will output the most likely fault type to help operation and maintenance personnel quickly locate the problem and take appropriate maintenance measures.

[0149] S108 matches the replacement time period of the hydrological monitoring unit according to the fault type.

[0150] The specific steps include:

[0151] S801 obtains the life prediction data provided by the equipment manufacturer to evaluate the expected life of each hydrological monitoring unit;

[0152] First, obtain official life prediction data for each hydrological monitoring unit from the equipment manufacturer. These data are usually based on laboratory testing and long-term field experience, including but not limited to mean time between failures (MTBF), maximum service life, and performance degradation rate under different environmental conditions.

[0153] S802 sets a replacement schedule based on expected life and failure type;

[0154] Divide all known failures into predictable failures (such as natural wear and tear, aging) and sudden failures (such as natural disasters, accidental damage). For each failure mode, analyze its frequency of occurrence, scope of impact and possible consequences. For predictable failures, determine a reasonable preventive replacement cycle based on the expected life assessment results. This cycle should take into account a safety margin to ensure that the replacement is completed before the expected failure occurs, thereby avoiding service interruptions. Generate a schedule that can be updated in real time based on the latest assessment results. For example, if it is found that the actual life of a certain type of equipment is longer or shorter than expected, the corresponding schedule should be adjusted in a timely manner.

[0155] The maintenance interval of the sensor can be calculated using the following formula:

[0156] R sensor = k * T

[0157] R sensor is the sensor maintenance or replacement time interval (hours); k is the proportionality coefficient, which can be determined according to the sensor type and environmental conditions. For example, for a water quality sensor, k=0.05.

[0158] Assume that a water quality sensor has been used for 3000 hours and the proportionality coefficient k = 0.05:

[0159] R sensor =0.05.3000=150 hours, so it is recommended to clean or replace it within 150 hours.

[0160] If it is a circuit fault, the following formula can be used for calculation:

[0161]

[0162] Where MTBF is the mean time between failures (hours); S is the safety time window; τ is the time constant. WF is the rate of entering the metastable state, F D is the fault detection probability.

[0163] Assume that the circuit parameters are S = 10, T = 5, T WF =0.01, F D =0.1, the calculated result is that it is recommended to conduct a circuit inspection every 7389 hours.

[0164] If it is a communication failure, the maintenance interval can be calculated using the following formula:

[0165]

[0166] Where R comm is the maintenance interval of the communication equipment (hours); λ is the failure rate (unit: times / hour). Assuming that the failure rate of the communication module is λ = 0.001 times / hour, it is recommended to check the communication module every 1000 hours.

[0167] S803 obtains the current fault type and usage time, and matches the corresponding replacement time period.

[0168] When a fault alarm occurs, a preliminary diagnosis is quickly performed to determine the specific type of fault. If it is a sudden fault, the emergency response process is immediately initiated; if it is a foreseeable fault, further check whether it is close to the preset replacement time point. Then, based on the current fault type and the accumulated usage time of the equipment, find the pre-set replacement schedule and the most appropriate replacement time period. For equipment that is about to expire, plan and arrange replacement tasks in advance; for equipment that has exceeded the safety margin, replacement should be arranged as soon as possible.

[0169] Second embodiment

[0170] See also Fig. 9The present invention also provides a hydrological situation monitoring fault warning system, comprising a calibration unit 101, a plurality of hydrological monitoring units 102, a linkage model generation module 103, a detection value prediction module 104, a first deviation calculation module 105, a data feature calculation module 106, a second deviation calculation module 107 and a fault alarm module 108; the linkage model generation module 103 is used to establish a monitoring value linkage model between the calibration unit 101 and each hydrological monitoring unit 102 under the condition of preset environmental parameters, the environmental parameters including temperature, humidity, and altitude data, and the calibration unit 101 is arranged between the plurality of hydrological monitoring units 102; the detection value prediction module 104 is used to obtain the calibration monitoring value measured by the calibration unit 101 entering the monitoring environment every preset time period, and calculate the predicted detection value of each hydrological monitoring unit 102 based on the calibration monitoring value and the monitoring value linkage model; the first deviation calculation module 105 05, used to calculate the deviation value between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit 102, and determine whether the deviation value is within the predetermined target range. When the deviation value is not within the target range, the historical monitoring data of the target hydrological monitoring unit 102 and the nearest hydrological monitoring unit 102 of the same type are obtained; the data feature calculation module 106, used to establish the first data fluctuation feature of the target monitoring unit and the second data fluctuation feature of the nearest hydrological monitoring unit 102 of the same type based on the degree of change of the historical monitoring data; the second deviation calculation module 107, used to compare the first data fluctuation feature of the current monitoring unit with the second data fluctuation feature of the nearest hydrological monitoring unit 102 of the same type. If the difference value exceeds the preset threshold, it is determined that the target monitoring unit has a fault; the fault alarm module 108, used to obtain the characteristic information of the faulty hydrological monitoring unit 102 and notify relevant personnel.

[0171] In this embodiment, the calibration unit 101 is placed between multiple hydrological monitoring units 102. Its main task is to provide a reference standard to ensure the data accuracy of all monitoring equipment. It will regularly enter the monitoring environment to measure and adjust other hydrological monitoring units 102 based on the results.

[0172] The hydrological monitoring units 102 are distributed in different geographical locations to collect important hydrological parameters such as water flow velocity, water level, water quality components, etc. in real time.

[0173] Considering that factors such as temperature, humidity and altitude in the natural environment will affect the hydrological data, the linkage model generation module 103 constructs a set of mathematical models that can reflect the relationship between these factors and the readings of each monitoring point. In this way, even under changing external conditions, the consistency between the predicted value and the actual observed value can be maintained.

[0174] The detection value prediction module 104 is based on the previously established linkage model and combines the fresh correction information from the calibration unit 101. The module can make an estimate of the value that should appear at each hydrological monitoring point at a certain time in the future. This step is crucial for discovering potential problems in advance.

[0175] The first deviation calculation module 105 is used to compare the predicted value with the actual reading currently obtained from each hydrological monitoring unit 102 to calculate the difference between the two. If the error at a certain point exceeds the set safety limit, it means that there is a risk of failure and further investigation is required.

[0176] Data feature calculation module 106: After initially determining that an abnormality exists, the data feature calculation module 106 obtains the behavior patterns of the target monitoring point and similar types of nearby sites over a period of time to find out unique trends or patterns that indicate a fault.

[0177] On this basis, the second deviation calculation module 107 compares and analyzes the behavior characteristics of the above two locations. If it is found that the difference between them significantly deviates from the normal range, it is very likely that a hardware failure or software error has occurred at the target monitoring point, and measures must be taken immediately.

[0178] Finally, once it is confirmed that a problem has indeed occurred in a hydrological monitoring unit 102, the fault alarm module 108 will automatically trigger the alarm mechanism, quickly notifying relevant personnel of the specific location of the fault and other relevant information so that they can intervene as soon as possible to prevent the situation from deteriorating.

[0179] This system not only improves the efficiency and accuracy of hydrological monitoring, but also reduces the need for human intervention through intelligent means, making water resource management and disaster prevention more scientific and reasonable.

[0180] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A hydrological condition monitoring fault early warning method based on artificial intelligence, It is characterized in that The method comprises: establishing a monitoring numerical linkage model between a calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters, wherein the environmental parameters include temperature, humidity, and altitude data, and the calibration unit is arranged between the plurality of hydrological monitoring units; Obtaining calibration monitoring values ​​measured by the calibration unit entering the monitoring environment at preset time intervals, and calculating predicted detection values ​​of each hydrological monitoring unit based on the calibration monitoring values ​​and the monitoring numerical linkage model; Calculate the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, and determine whether the deviation is within the predetermined target range. When the deviation is not within the target range, obtain the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type; Based on the degree of change of historical monitoring data, a first data fluctuation characteristic of the target monitoring unit and a second data fluctuation characteristic of the closest hydrological monitoring unit of the same type are established; Compare the first data fluctuation feature of the current monitoring unit with the second data fluctuation feature of the nearest hydrological monitoring unit of the same type, and if the difference value exceeds a preset threshold, determine that the target monitoring unit is faulty; Obtain characteristic information of faulty hydrological monitoring units and notify relevant personnel; Classify the fault types of the hydrological monitoring units based on the characteristic information; Match the replacement time period of the hydrological monitoring unit according to the fault type.

2. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 1, characterized in that: The types of hydrological monitoring units include rain gauges, flow meters and water quality analyzers.

3. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 2, characterized in that: The specific steps of establishing a monitoring numerical linkage model between the calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters include: Calculate the placement position of the calibration unit based on the position information of all hydrological monitoring units; After placing the calibration unit at the placement position, collecting calibration data at preset time intervals; Obtain monitoring data of all hydrological monitoring units with the same timestamp as the calibration data; The relationship between calibration data and monitoring data is obtained by using polynomial regression analysis, and a monitoring numerical linkage model is obtained.

4. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 3, characterized in that: The specific steps of using polynomial regression analysis to obtain the relationship between calibration data and monitoring data and to obtain the monitoring numerical linkage model include: Obtain temperature, humidity, and atmospheric pressure data and use them as input features to predict the difference between calibration data and monitoring data; Split the data into training and testing sets; Use the linear regression algorithm to fit a polynomial regression model on the training set; The test set is used to test and adjust the polynomial regression model to obtain the monitoring numerical linkage model.

5. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 4, characterized in that: The specific steps of collecting calibration data at preset time intervals after the calibration unit is placed at the placement position include: Set a calibration cycle; Allow the calibration unit to run in the actual environment for a preset period of time before collecting calibration data.

6. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 5, characterized in that: The specific steps of calculating the deviation between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, judging whether the deviation is within a predetermined target range, and obtaining the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type when the deviation is not within the target range include: Obtain predicted detection values ​​and real-time monitoring parameters at corresponding time points; Remove outliers and missing values ​​from real-time monitoring parameters, and remove predicted detection values ​​at corresponding time points; Calculate the deviation between the predicted detection value and the real-time monitoring parameter at the corresponding time point; When the deviation value is not within the target range, the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type are obtained.

7. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 6, characterized in that: The specific steps of obtaining characteristic information of the faulty hydrological monitoring unit and notifying relevant personnel include: Extracting basic information of the faulty hydrological monitoring unit from a database, the basic information including equipment model, installation location, operation time, and maintenance history; Generate notification information based on basic information; Use multiple communication channels to deliver notification information to relevant processing personnel.

8. The artificial intelligence-based hydrological condition monitoring fault early warning method according to claim 7, characterized in that: The specific steps of classifying the fault types of the hydrological monitoring unit based on the characteristic information include: Collect historical operating status data from hydrological monitoring units, including data under normal operating conditions as well as data under known fault conditions; Extract fault classification features based on historical working status data to obtain a training data set, wherein the fault classification features include temperature, humidity, flow rate, and pressure; The support vector machine model is trained using the training data set to obtain a discrimination model; The fault type is obtained based on the discrimination model and the fault classification feature data of the currently detected hydrological monitoring unit.

9. A hydrological condition monitoring fault early warning system, applied to the hydrological condition monitoring fault early warning method based on artificial intelligence as claimed in claims 1 to 8, characterized in that: include: It includes a calibration unit, a plurality of hydrological monitoring units, a linkage model generation module, a detection value prediction module, a first deviation calculation module, a data feature calculation module, a second deviation calculation module and a fault alarm module; The linkage model generation module is used to establish a monitoring numerical linkage model between the calibration unit and each hydrological monitoring unit under the condition of preset environmental parameters, wherein the environmental parameters include temperature, humidity, and altitude data, and the calibration unit is arranged between multiple hydrological monitoring units; The detection value prediction module is used to obtain the calibration monitoring value measured by the calibration unit entering the monitoring environment every preset time period, and calculate the predicted detection value of each hydrological monitoring unit based on the calibration monitoring value and the monitoring value linkage model; The first deviation calculation module is used to calculate the deviation value between the predicted detection value and the real-time monitoring parameter actually detected by each hydrological monitoring unit, and determine whether the deviation value is within a predetermined target range. When the deviation value is not within the target range, the historical monitoring data of the target hydrological monitoring unit and the nearest hydrological monitoring unit of the same type are obtained; The data feature calculation module is used to establish a first data fluctuation feature of a target monitoring unit and a second data fluctuation feature of a hydrological monitoring unit of the same type that is closest to the target monitoring unit based on the degree of change of historical monitoring data; The second deviation calculation module is used to compare the first data fluctuation characteristic of the current monitoring unit with the second data fluctuation characteristic of the nearest hydrological monitoring unit of the same type, and if the difference value exceeds a preset threshold, it is determined that the target monitoring unit is faulty; The fault alarm module is used to obtain characteristic information of the faulty hydrological monitoring unit and notify relevant personnel.

Images