A method and device for predicting pipe burst danger of hydraulic structures

Abstract

The present application discloses a method and device for predicting pipe burst danger for hydraulic structures, which belongs to the field of danger prediction and solves the problem of low accuracy in pipe burst danger prediction. The method includes: obtaining historical multi-dimensional feature data of hydraulic structures; determining the sensitive values ​​of each dimensional feature to update the standard pipe burst danger prediction network model to obtain the target pipe burst danger prediction network model; establishing the probability distribution of each dimensional feature, performing random sampling according to the probability distribution, obtaining the random parameter value of each dimensional feature, so as to simulate and predict the target pipe burst danger prediction network model; when the simulation result is higher than the preset risk coefficient, determining the target random parameter value of each dimensional feature corresponding to the simulation result to obtain the potential risk range of the feature data of each dimensional feature; predicting the pipe burst danger of hydraulic structures according to the potential risk range of the feature data and the target pipe burst danger prediction network model. Improve the prediction accuracy.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and device for predicting pipe burst hazards in hydraulic structures. Background Art

[0002] Piping mainly occurs in the foundation or contact zone of hydraulic structures such as levees and dams, especially when the seepage water carries fine-grained soil and forms holes or channels in the soil layer under the action of high water pressure. The holes or channels formed by piping will weaken the structural strength of levees and dams, making these projects prone to partial collapse or sliding under the action of water pressure. In extreme cases, piping may cause the overall collapse of levees and dams, posing a serious threat to the safety of life and property of people in downstream areas.

[0003] At present, in the process of pipe burst monitoring, remote sensing image technology is usually used to collect images of hydraulic structures, and the images are analyzed for intuitive factors such as pipe burst outlets and water surface changes, or prediction models are used to regularly monitor the environmental characteristics of hydraulic structures. However, since pipe bursts occur underwater, it is almost difficult to easily detect the early stage of pipe bursts in water surface changes. In addition, the factors causing pipe bursts are diverse, and the actual environment of hydraulic structures is also diverse. Existing prediction models are usually unable to adapt to the diverse and changing actual environment of hydraulic structures. In addition, since different hydraulic structures have different actual environments and the complexity of pipe bursts, the above two methods cannot determine the warning level between a single hydraulic structure and the risk of pipe bursts for a single dimensional feature, and thus cannot perform real-time warning analysis on the single dimensional feature data of hydraulic structures.

[0004] In summary, in the current scheme, the accuracy of prediction of pipe burst hazards in hydraulic structures is low. Summary of the invention

[0005] The embodiments of the present application provide a method and device for predicting pipe burst hazards of hydraulic structures, which are used to solve the problem of low accuracy in predicting pipe burst hazards of hydraulic structures.

[0006] The present application embodiment adopts the following technical solutions:

[0007] On the one hand, an embodiment of the present application provides a method for predicting piping danger of hydraulic structures, the method comprising: obtaining historical multi-dimensional feature data of the hydraulic structures for piping risks in a historical prediction database of hydraulic structures; the multi-dimensional features are related to geological characteristics, soil characteristics and hydrological characteristics; performing sensitivity analysis on the historical multi-dimensional feature data to determine the sensitive values ​​of each dimensional feature for changes in piping risk; updating a standard piping danger prediction network model according to the sensitive values ​​to obtain a target piping danger prediction network model for the hydraulic structures; establishing a probability distribution of each dimensional feature according to the data characteristics of each dimensional feature, According to the probability distribution, random sampling is performed to obtain random parameter values ​​of each dimensional feature; according to the random parameter values ​​of each dimensional feature, the target pipe burst hazard prediction network model is simulated and predicted; when the simulation result is higher than the preset risk coefficient, the target random parameter value of each dimensional feature corresponding to the simulation result is determined to obtain the characteristic data risk group of each dimensional feature; according to the probability distribution of each dimensional feature, the respective characteristic data risk groups are analyzed to obtain the potential risk range of the characteristic data of each dimensional feature; according to the potential risk range of the characteristic data and the target pipe burst hazard prediction network model, the pipe burst hazard prediction of the hydraulic structure is performed.

[0008] In one example, the sensitivity analysis of the historical multi-dimensional feature data is performed to determine the sensitivity level of each dimensional feature to the change in piping risk, specifically including: constructing a multivariate linear regression model based on each dimensional feature; the multivariate linear regression model is used to fit the relationship between the multi-dimensional feature and the piping risk; fitting the multivariate linear regression model based on the historical multi-dimensional feature data to obtain the regression coefficient of each dimensional feature; in a pre-constructed feature relationship map, determining multiple target feature node combinations with a correlation higher than a preset correlation threshold; wherein, in a single target feature node combination, the correlation between the first target feature node and the second target feature node is higher than the preset correlation threshold; according to the node dependency in each target feature node combination, compensating the regression coefficient of the corresponding target dimensional feature; determining the sensitive value of each dimensional feature based on the absolute values ​​corresponding to the compensated regression coefficient of the target dimensional feature and the regression coefficient of the non-target dimensional feature; the higher the absolute value, the higher the sensitive value.

[0009] In one example, the regression coefficient of the corresponding target dimension feature is compensated according to the node dependency in each target feature node combination, specifically including: retrieving the compensation coefficient corresponding to the correlation degree of each target feature node combination in the compensation coefficient mapping relationship table; the compensation coefficient is greater than 1; in each target feature node combination, if the first target feature node depends on the second target feature node, multiplying the compensation coefficient with the regression coefficient to obtain the compensated regression coefficient.

[0010] In one example, the standard piping hazard prediction network model is updated according to the sensitive value to obtain the target piping hazard prediction network model of the hydraulic structure, specifically including: respectively obtaining the first offset between the sensitive value of each dimensional feature and the set weight maximum value; obtaining the second offset between the maximum sensitive score and the set weight maximum value; respectively obtaining the ratio between the first offset and the second offset of each dimensional feature to obtain the weight factor of each dimensional feature; multiplying the weight factor of each dimensional feature with the weight of each dimensional feature in the standard piping hazard prediction network model to obtain the target piping hazard prediction network model of the hydraulic structure.

[0011] In one example, the feature data risk groups are analyzed according to the probability distribution of each dimensional feature to obtain the potential risk range of the feature data of each dimensional feature, which specifically includes: analyzing each feature data risk group to extract the data value range of each dimensional feature; in the probability distribution of each dimensional feature, according to the data value range of each dimensional feature, determining the minimum probability corresponding to each dimensional feature; in the probability distribution of each dimensional feature, adding the data values ​​below the minimum probability to the respective data value range to obtain the potential risk range of the feature data.

[0012] In one example, the piping hazard prediction of the hydraulic structure is performed based on the potential risk range of the characteristic data and the target piping hazard prediction network model, specifically including: obtaining the current dimensional characteristic data of the hydraulic structure; when the piping hazard prediction request is not automatically triggered, determining whether the current dimensional characteristic data is within the potential risk range of its own characteristic data; if so, sending a request to a multi-dimensional characteristic data acquisition node, and when obtaining the latest multi-dimensional characteristic data, inputting the multi-dimensional characteristic data into the target piping hazard prediction network model to obtain the piping risk coefficient of the hydraulic structure; and predicting the piping hazard of the hydraulic structure based on the piping risk coefficient.

[0013] In one example, the piping hazard prediction of the hydraulic structure based on the piping risk coefficient specifically includes: when the piping risk coefficient is lower than a preset risk coefficient threshold, obtaining meteorological information of the area where the hydraulic structure is located within a preset time interval; extracting the rainfall of the hydraulic structure within the preset time interval from the meteorological information; when the rainfall is higher than a preset rainfall threshold, generating piping hazard warning information of the hydraulic structure, and sending the piping hazard warning information to a management user terminal; when the rainfall is lower than a preset rainfall threshold, obtaining a historical feature data change curve for the current dimensional feature of the hydraulic structure within the same time interval last year; generating a feature data change curve for the current dimensional feature within the time interval; and predicting the piping hazard of the hydraulic structure by comparing the historical feature data change curve with the feature data change curve.

[0014] In one example, the prediction of pipe burst hazard of the hydraulic structure is performed based on the comparison between the historical characteristic data change curve and the characteristic data change curve, specifically including: judging whether the historical characteristic data change curve and the characteristic data change curve meet the trend consistency condition; if so, comparing the historical characteristic data change curve with the characteristic data change curve to generate a fluctuation data sequence; inputting the fluctuation data sequence into a pre-built fluctuation risk prediction network model to obtain the fluctuation risk coefficient of the hydraulic structure; when the fluctuation risk coefficient is higher than a preset fluctuation coefficient threshold, generating pipe burst hazard warning information of the hydraulic structure.

[0015] In one example, the historical feature data change curve is compared with the feature data change curve to generate a fluctuation data sequence, which specifically includes: calculating the difference between each time point of the feature data change curve and multiple historical time points of the historical feature data change curve; determining the historical time point with the smallest difference as the alignment time point of each time point; calculating the difference between the data value of each time point in the feature data change curve and the data value of each alignment time point in the historical feature data change curve, obtaining the fluctuation value of each time point, and generating a fluctuation data sequence.

[0016] On the other hand, an embodiment of the present application provides a device for predicting pipe burst hazards for hydraulic structures, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any of the above-described methods for predicting pipe burst hazards for hydraulic structures.

[0017] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0018] Taking into account the actual environment of hydraulic structures, using historical multi-dimensional feature data for sensitivity analysis can better capture the impact of different factors on the risk of piping for the hydraulic structure, thereby mining the more critical characteristic factors for the risk of piping, and then updating the standard piping hazard prediction network model, thereby obtaining a personalized model adapted to the hydraulic structure environment, and improving the accuracy of the piping risk of hydraulic structures. Furthermore, real-time analysis is performed on the data of a certain dimensional feature currently acquired, but instead of setting a fixed data threshold, the potential risk data range is predicted to be adapted to the actual environment of the hydraulic structure, thereby establishing a probability distribution and sampling, and using random parameter values ​​to simulate and predict the target piping hazard prediction network model, obtaining possible piping risk results, and pre-mining and inferring risk data under different circumstances, thereby providing more detailed risk information and further improving the accuracy of piping risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present application, some embodiments of the present application will be described in detail below in conjunction with the accompanying drawings, in which:

[0020] Figure 1 A flow chart of a method for predicting pipe burst danger of hydraulic structures provided in an embodiment of the present application;

[0021] Figure 2 A schematic diagram of the structure of a pipe burst hazard prediction device for hydraulic structures provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0023] Some embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0024] Figure 1 A schematic diagram of a flow chart of a method for predicting pipe burst danger of hydraulic structures provided in an embodiment of the present application. Certain input parameters or intermediate results in the flow chart allow manual intervention and adjustment to help improve accuracy.

[0025] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail by taking a server as an example.

[0026] It should be noted that the server may be a single device or a system consisting of multiple devices, that is, a distributed server, and this application does not make any specific limitation on this.

[0027] Figure 1 The process in includes the following steps:

[0028] S101: From a historical prediction database of hydraulic structures, historical multi-dimensional characteristic data of the hydraulic structures for piping risks are obtained; the multi-dimensional characteristics are related to geological characteristics, soil characteristics, and hydrological characteristics.

[0029] Among them, geological characteristics include geological structure, stratum distribution, geological faults, etc. Soil characteristics include soil type, soil density, soil permeability, soil moisture, etc. Hydrological characteristics include water level, water flow direction, water quality, water flow, etc.

[0030] It should be noted that the historical prediction database includes the multi-dimensional feature data corresponding to each prediction time point when the standard piping hazard prediction network model is used for prediction, as well as the piping risk coefficient at each prediction time point.

[0031] It should be noted that the training process of the standard piping hazard prediction network model is as follows:

[0032] Firstly, the sample multi-dimensional feature data and the corresponding piping risk coefficient are obtained. The sample multi-dimensional feature data are taken as input, and the corresponding piping risk coefficient is taken as the sample label. The neural network architecture of the standard piping hazard prediction network model is trained to obtain the standard piping hazard prediction network model.

[0033] Among them, the neural network architecture may include an input layer, a hidden layer and an output layer, and the input layer receives multi-dimensional feature data of the sample. These feature data can be numerical or categorical, depending on the characteristics of the data. The hidden layer is the middle layer in the neural network, which extracts and transforms the input data. The hidden layer may contain multiple neurons, each of which performs a weighted summation on the input data and performs a nonlinear transformation through an activation function. The role of the activation function is to introduce nonlinearity so that the neural network can learn complex patterns. The output layer predicts the pipe burst risk coefficient based on the output results of the hidden layer. The output layer usually has only one neuron, which uses a linear activation function to output the predicted pipe burst risk coefficient.

[0034] For example, when the hydraulic structure is a dam, the geological characteristics of the dam area can be analyzed by geological radar, and the soil characteristics can be analyzed by sampling the soil at multiple preset locations of the hydraulic structure. The water level of the dam is obtained by a water level gauge. The water quality is analyzed by a water quality analyzer, and the water flow direction of the dam is detected by the tracer method. For example, tracers such as radioactive isotopes or dyes are injected upstream of the dam, and then monitoring points are set downstream to detect the appearance time and concentration of the tracer, and the groundwater flow direction is inferred based on the migration speed and direction of the tracer.

[0035] It should be noted that when piping occurs in the dam body, the following signs may occur:

[0036] 1. Rising water level in the dam body: For example, if the water level on one side rises suddenly or continues to rise, it means that water flows from one side of the dam body into the other side through the pipe burst channel.

[0037] 2. Wet or saturated soil: Piping may cause the soil around the dam to become wet or saturated.

[0038] 3. Changes in water quality: The water brought out by pipe bursting may carry mud or other substances, causing the water to become turbid or contain sediment.

[0039] 5. Increased seepage: At the location where piping occurs, the seepage may increase significantly.

[0040] 6. Changes in soil pore water pressure: Use pressure sensors to monitor changes in soil pore water pressure. Piping may lead to a decrease in pore water pressure.

[0041] 7. Damage to geological structures: Piping may cause damage to geological structures, such as loosening, cracks or collapse.

[0042] S102: Perform sensitivity analysis on the historical multi-dimensional feature data to determine the sensitive value of each dimensional feature to the change of piping risk.

[0043] In some embodiments of the present application, since different hydraulic structures have different regional characteristics and construction structure characteristics, the importance factors that lead to pipe bursts are also different. Therefore, for different hydraulic structures, the correlation between different dimensional feature data and the pipe burst risk is mined, so as to focus on the dimensional feature data with a high correlation with the pipe burst risk of a single hydraulic structure. This can take into account the different relationships between the characteristics of different hydraulic structures and the pipe burst risk, thereby improving the accuracy of pipe burst prediction for different hydraulic structures.

[0044] Furthermore, there may be interactions between multi-dimensional features, which means that changes in one feature may be affected by other features. Therefore, in sensitivity analysis, these interactions are considered to obtain a more comprehensive understanding. Considering multivariate interactions can more comprehensively understand the impact of each dimensional feature on the change of piping risk. For example, the height of the water level will affect the soil permeability, the change of geological faults will affect the soil density, and the change of soil permeability will affect the water quality of the dam body.

[0045] Therefore, a feature relationship graph consisting of multiple dimensional feature nodes and associated edges between the multiple dimensional feature nodes is constructed.

[0046] Based on this, the process of sensitivity analysis of historical multi-dimensional feature data is as follows:

[0047] According to the characteristics of each dimension, a multivariate linear regression model is constructed. The multivariate linear regression model is used to fit the relationship between multidimensional characteristics and piping risk.

[0048] For example, the formula for the multiple linear regression model can be as follows:

[0049] Y=β0+β1*x1+β2*x2+β3*x3+ε

[0050] Among them, Y is the piping risk coefficient, β0 is the intercept, β1, β2, β3 are the regression coefficients of x1, x2, x3, respectively, x1, x2, x3 are the water level, soil moisture and geological fault value at the same prediction time point, respectively, and ε is the error term.

[0051] It should be noted that the number of the above independent variables is related to the dimensional feature data type. Here, water level, soil moisture and geological fault are used as examples for explanation.

[0052] Then, based on the historical multi-dimensional feature data, the multivariate linear regression model is fitted to obtain the regression coefficient of each dimensional feature.

[0053] Among them, the multi-dimensional characteristic data and piping risk coefficient of each prediction time point are respectively input into the multivariate linear regression model to solve the regression coefficient of each dimensional feature.

[0054] It should be noted that there is no fixed limit on the numerical range of the regression coefficient of the multivariate linear regression model. They can be positive, negative or zero, and the specific value depends on the characteristics of the data and the fit of the model. The size and sign of the regression coefficient indicate the strength and direction of the linear relationship between the independent variable and the dependent variable. Generally speaking, the larger the absolute value of the regression coefficient, the greater the influence of the independent variable on the dependent variable; a positive regression coefficient indicates that there is a positive correlation between the independent variable and the dependent variable, that is, when the independent variable increases, the dependent variable also tends to increase; a negative regression coefficient indicates that there is a negative correlation between the independent variable and the dependent variable, that is, when the independent variable increases, the dependent variable tends to decrease; a zero regression coefficient indicates that there is no linear relationship between the independent variable and the dependent variable.

[0055] Then, in the pre-constructed feature relationship graph, multiple target feature node combinations with correlations higher than a preset correlation threshold are determined, wherein in a single target feature node combination, the correlation between the first target feature node and the second target feature node is higher than the preset correlation threshold.

[0056] It should be noted that the correlation is a numerical value that indicates the strength of the correlation between nodes. The correlation calculation in the feature relationship graph can be calculated by comprehensively considering multiple factors, such as direct correlation, feature value similarity, and path length.

[0057] Then, according to the node dependencies in each target feature node combination, the regression coefficients of the corresponding target dimension features are compensated.

[0058] It should be noted that the compensation process is as follows:

[0059] First, in the compensation coefficient mapping relationship table, the compensation coefficient corresponding to the correlation degree of the target feature node combination is retrieved. The compensation coefficient is greater than 1.

[0060] In each target feature node combination, if the first target feature node depends on the second target feature node, the compensation coefficient is multiplied by the regression coefficient of the target dimension feature corresponding to the second target feature node to obtain the compensation regression coefficient.

[0061] It should be noted that when the first target feature node depends on the second target feature node, it means that when the dimensional feature of the second target feature node changes in data, it will cause a data change in the dimensional feature of the first target feature node. Therefore, after compensating the regression coefficient of the dimensional feature of the second target feature node, the coefficient value of the dimensional feature is increased, thereby increasing the importance of the risk of pipe burst caused by the dimensional feature.

[0062] Finally, the sensitivity value of each dimensional feature is determined according to the absolute values ​​corresponding to the compensation regression coefficient of the target dimensional feature and the regression coefficient of the non-target dimensional feature. The higher the absolute value, the higher the sensitivity value. For example, the sensitivity value of soil permeability is 0.3, and the sensitivity value of water level is 0.5.

[0063] S103: updating the standard piping hazard prediction network model according to the sensitive value to obtain the target piping hazard prediction network model of the hydraulic structure.

[0064] Among them, first, the first offset between the sensitive value of each dimensional feature and the set weight maximum value is calculated. Then, the second offset between the maximum sensitive score and the set weight maximum value is calculated. Then, the ratio between the first offset and the second offset of each dimensional feature is calculated to obtain the weight factor of each dimensional feature. Finally, the weight factor of each dimensional feature is multiplied by the weight of each dimensional feature in the standard piping hazard prediction network model to obtain the target piping hazard prediction network model of hydraulic structures.

[0065] It should be noted that the maximum weight setting can be set according to the actual weight range. For example, if the weight range is 0-1, the maximum weight setting is 1. When obtaining the first offset between the sensitive value of each dimensional feature and the set weight maximum value, the sensitive value of each dimensional feature is subtracted from the set weight maximum value, and the same applies to the second offset.

[0066] Based on this, the range of sensitive values ​​can be converted into values ​​between the weight range to match the weight range. In this way, the weight value after multiplication can still be within the weight range, accurately representing the relative importance of each dimensional feature to the prediction network model for the hydraulic structure.

[0067] S104: According to the data characteristics of each dimensional feature, a probability distribution of each dimensional feature is established, and random sampling is performed according to the probability distribution to obtain a random parameter value of each dimensional feature.

[0068] It should be noted that if the data of the dimension feature presents A distribution, you can choose A distribution as its probability distribution. The horizontal axis of the probability distribution represents the value axis of all possible results of the event, and the vertical axis represents the probability of occurrence or probability density corresponding to different results.

[0069] For example, when the dimension feature is soil density, the user observes the soil density data and finds that it presents a relatively concentrated normal distribution. Therefore, the user can mark the soil density as a normal distribution, and when the server determines that the soil density data is close to a normal distribution, it selects the normal distribution as its probability distribution.

[0070] Then, based on the statistical information of the historical soil density characteristic data, the mean and standard deviation of the normal distribution are estimated, for example, 1.8 and 0.2 respectively, thus determining the parameters.

[0071] Then, using a random sampling function from a normal distribution, generate multiple random parameter values ​​for soil density. This can be achieved by using a random number generator.

[0072] Finally, random sampling can be performed multiple times, for example, 100 times.

[0073] S105: According to the random parameter values ​​of the characteristics of each dimension, simulate and predict the target pipe burst hazard prediction network model.

[0074] Among them, the random parameter values ​​of the characteristics of each dimension are randomly combined to obtain multiple input combinations, and each input combination is input into the target pipe burst hazard prediction network model to obtain the simulation results.

[0075] S106: When the simulation result is higher than the preset risk coefficient, determine the target random parameter value of each dimensional feature corresponding to the simulation result to obtain a feature data risk group of each dimensional feature.

[0076] It should be noted that, for example, if the i-th simulation result of feature A exceeds the preset risk coefficient, the target random parameter value of feature A at the i-th time is determined when the risk coefficient is exceeded, and the target random parameter value is input into the feature data risk group of feature A. In other words, the feature data risk group of feature A may include multiple target random parameter values ​​of multiple simulations.

[0077] It should be noted that the preset risk coefficient can be set according to actual needs. For example, the preset risk coefficient is 0.6.

[0078] S107: Analyze the risk groups of the feature data according to the probability distribution of the features of each dimension to obtain the potential risk range of the feature data of each dimension.

[0079] Among them, when the probability is relatively small, it means that the possibility of the feature data being an abnormal value is higher. When determining an abnormal value, a value with a smaller probability than the abnormal value means that the possibility of abnormality is also relatively high. Therefore, the range of abnormal data values ​​is mined in advance.

[0080] Based on this, firstly, each characteristic data risk group is analyzed to extract the data value range of each dimensional feature.

[0081] Then, in the probability distribution of each dimensional feature, the minimum probability corresponding to each dimensional feature is determined according to the data value range of each dimensional feature.

[0082] Then, in the probability distribution of each dimensional feature, the data values ​​below the minimum probability are added to the respective data value range to obtain the potential risk range of the feature data.

[0083] S108: Predicting the piping hazard of the hydraulic structure according to the potential risk range of the characteristic data and the target piping hazard prediction network model.

[0084] In some embodiments of the present application, the collection method of each dimensional feature may be different. For example, some dimensional feature data can be collected in real time by sensors, while some dimensional feature data requires human intervention for collection. Alternatively, the collection time intervals of each dimensional feature are different. Or a fixed time interval for model analysis is set. Therefore, it is not possible to perform real-time model analysis on the feature data of each dimension. Therefore, some real-time collected data is monitored in advance according to the potential risk range of the feature data. It should be noted that when using the target pipe burst hazard prediction network model for prediction, the latest feature data of each dimensional feature is usually used.

[0085] Based on this, the current dimensional characteristic data of the hydraulic structure is obtained.

[0086] Then, when the pipe burst hazard prediction request is not automatically triggered, it is determined whether the current dimension feature data is within the potential risk range of its own feature data.

[0087] It should be noted that automatically triggering a pipe burst hazard prediction request refers to triggering a target pipe burst hazard prediction network model to perform a prediction.

[0088] If so, a request is sent to the multi-dimensional feature data acquisition node. When the latest multi-dimensional feature data is obtained, the multi-dimensional feature data is input into the target pipe burst hazard prediction network model to obtain the pipe burst risk coefficient of the hydraulic structure.

[0089] It should be noted that the multi-dimensional feature data acquisition node can be a dimensional feature data acquisition device and a management user terminal, or when the multi-dimensional feature data can be automatically collected by the acquisition device, it can also only include the dimensional feature data acquisition device.

[0090] Finally, based on the piping risk coefficient, piping hazard prediction for hydraulic structures is carried out.

[0091] In some embodiments of the present application, taking into account the concealment of pipe bursts, due to the appearance of abnormal data values ​​of dimensional characteristics, when the model analysis result shows that the risk factor is low, the impact of rainfall on pipe bursts is still considered, because rainfall will greatly increase the risk of pipe bursts.

[0092] Based on this, when the piping risk factor is lower than the preset risk factor threshold, the meteorological information of the area where the hydraulic structure is located within a preset time interval is obtained. For example, the time interval is one week before the current time.

[0093] Then, the rainfall of hydraulic structures within a preset time interval is extracted from the meteorological information.

[0094] Then, when the rainfall is higher than the preset rainfall threshold, pipe burst hazard warning information of the hydraulic structure is generated and sent to the management user terminal.

[0095] Furthermore, since the changing trend between characteristic data is also related to piping, for example, during piping, the water quality becomes more turbid, the soil density becomes lower, and the soil porosity becomes higher. If the analysis results are significantly different from the known normal fluctuation pattern, it means that there is a high probability of an abnormality.

[0096] Based on this, when the rainfall is lower than the preset rainfall threshold, the historical characteristic data change curve of the hydraulic structure in the same time interval last year for the current dimension feature is obtained. Among them, the X-axis of the historical characteristic data change curve is time, and the Y-axis is the dimension feature value.

[0097] Then, a feature data change curve of the current dimension feature within the time interval is generated, wherein the X-axis of the feature data change curve is time and the Y-axis is the dimension feature value.

[0098] It should be noted that the time intervals between adjacent time points of the characteristic data change curve and the historical characteristic data change curve are equal.

[0099] It should be noted that if the preset time interval exceeds the current time, the time in the characteristic data change curve ends at the current time.

[0100] Finally, by comparing the historical characteristic data change curve with the characteristic data change curve, the pipe burst hazard of hydraulic structures is predicted.

[0101] The process of curve comparison is as follows:

[0102] First, determine whether the historical characteristic data change curve and the characteristic data change curve meet the trend consistency condition.

[0103] It should be noted that the trend consistency condition refers to whether the historical feature data change curve and the feature data change curve are consistent in overall trend, for example, whether the shapes are consistent, whether they have the same upward or downward trend, etc.

[0104] If yes, the historical characteristic data change curve is compared with the characteristic data change curve to generate a fluctuation data sequence, wherein the fluctuation data sequence includes characteristic difference values ​​(fluctuation values) sorted by time points.

[0105] It should be noted that if the current dimension feature is automatically collected by a collection device, the collection time interval is usually the same, but the collection time points may be different. In addition, if the current dimension feature requires human intervention for collection, the collection time points are usually different.

[0106] Therefore, the comparison process is as follows:

[0107] First, the difference between each time point of the characteristic data change curve and a plurality of historical time points of the historical characteristic data change curve is calculated.

[0108] Then, the historical time point with the smallest difference is determined as the alignment time point of each time point.

[0109] It should be noted that the dimensional feature data usually does not fluctuate greatly in a short period of time, so the historical time point with the smallest difference is selected. For example, if the first time point in the feature data change curve is 7:00 a.m. on the day of A month B, and the first historical time point is 8:00 a.m. on the day of A month B, then the alignment time point of the first time point is 8:00 a.m.

[0110] Then, the difference between the data value at each time point in the characteristic data change curve and the data value at each aligned time point in the historical characteristic data change curve is calculated to obtain the fluctuation value at each time point and generate a fluctuation data sequence.

[0111] Then, the fluctuation data sequence is input into the pre-built fluctuation risk prediction network model to obtain the fluctuation risk coefficient of the hydraulic structure. For example, the fluctuation risk coefficient is 0.4.

[0112] It should be noted that the process of constructing the volatility risk prediction network model is as follows: obtain the sample volatility data sequence and volatility risk coefficient of the sample dimension feature data, use the sample volatility data sequence as the sample input, use the volatility risk coefficient as the sample label, train the architecture of the volatility risk prediction network model, and obtain the volatility risk prediction network model. Among them, the architecture of the volatility risk prediction network model please refer to the setting concept of the standard pipe burst hazard prediction network model, and no detailed description will be given here.

[0113] Finally, when the fluctuation risk coefficient is higher than the preset fluctuation coefficient threshold, the pipe burst hazard warning information of the hydraulic structure is generated.

[0114] It should be noted that when the historical characteristic data change curve and the characteristic data change curve do not meet the trend consistency condition, the pipe burst hazard warning information of the hydraulic structure is generated.

[0115] It should be noted that although the embodiments of the present application are based on Figure 1 Steps S101 to S108 are described in sequence, but this does not mean that steps S101 to S108 must be performed in a strict order. Figure 1 The order shown in the figure is to introduce and explain step S101 to step S108 in sequence, in order to facilitate those skilled in the art to understand the technical solution of the embodiment of the present application. In other words, in the embodiment of the present application, the order between step S101 to step S108 can be appropriately adjusted according to actual needs.

[0116] pass Figure 1 The method can take into account the actual environment of hydraulic structures, use historical multi-dimensional feature data for sensitivity analysis, and better capture the impact of different factors on the risk of pipe bursts for the hydraulic structure, thereby mining the more critical characteristic factors for the risk of pipe bursts, and then update the standard pipe burst hazard prediction network model, thereby obtaining a personalized model adapted to the hydraulic structure environment, and improving the accuracy of the pipe burst risk of hydraulic structures. Furthermore, real-time analysis is performed on the data of a certain dimensional feature currently acquired, but instead of setting a fixed data threshold, the potential risk data range is predicted according to the actual environment of the hydraulic structure, thereby establishing a probability distribution and sampling, and using random parameter values ​​to simulate and predict the target pipe burst hazard prediction network model, obtaining possible pipe burst risk results, and pre-mining and inferring risk data under different circumstances, thereby providing more detailed risk information and further improving the accuracy of pipe burst risk prediction.

[0117] It is understandable that as the actual environments of different hydraulic structures are different, the target piping hazard prediction network models are different, and the risk ranges of the dimensional feature data are also different.

[0118] Based on the same idea, some embodiments of the present application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0119] Figure 2 A schematic diagram of a device for predicting pipe burst danger of hydraulic structures provided in an embodiment of the present application includes:

[0120] at least one processor; and,

[0121] a memory communicatively connected to the at least one processor; wherein,

[0122] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned methods for predicting pipe burst hazards for hydraulic structures.

[0123] Some embodiments of the present application provide a non-volatile computer storage medium for predicting piping hazards of hydraulic structures, which stores computer executable instructions, and the computer executable instructions can execute any of the above-mentioned methods for predicting piping hazards of hydraulic structures.

[0124] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0125] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0126] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0130] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0131] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0132] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0133] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0134] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the technical principle of the present application should fall within the protection scope of the present application.

Claims

1. A method for predicting pipe burst danger of hydraulic structures, characterized in that: The method comprises: In a historical prediction database of hydraulic structures, historical multi-dimensional characteristic data of the hydraulic structures for piping risks are obtained; the multi-dimensional characteristics are related to geological characteristics, soil characteristics and hydrological characteristics; Performing sensitivity analysis on the historical multi-dimensional feature data to determine the sensitive value of each dimensional feature to the change of piping risk; The standard piping hazard prediction network model is updated according to the sensitive value to obtain the target piping hazard prediction network model of the hydraulic structure; According to the data characteristics of each dimensional feature, a probability distribution of each dimensional feature is established, and random sampling is performed according to the probability distribution to obtain a random parameter value of each dimensional feature; According to the random parameter values ​​of the characteristics of each dimension, a simulation prediction is performed on the target piping hazard prediction network model; When the simulation result is higher than the preset risk coefficient, the target random parameter value of each dimensional feature corresponding to the simulation result is determined to obtain a feature data risk group of each dimensional feature; According to the probability distribution of each dimension feature, analyze the risk group of each feature data to obtain the potential risk range of the feature data of each dimension feature; Predicting the piping hazard of the hydraulic structure according to the potential risk range of the characteristic data and the target piping hazard prediction network model; The sensitivity analysis of the historical multi-dimensional feature data is performed to determine the sensitive value of each dimensional feature to the change of piping risk, specifically including: According to the characteristics of each dimension, a multivariate linear regression model is constructed; the multivariate linear regression model is used to fit the relationship between the multidimensional characteristics and the piping risk; According to the historical multi-dimensional feature data, the multivariate linear regression model is fitted to obtain the regression coefficient of each dimensional feature; In a pre-constructed feature relationship graph, a plurality of target feature node combinations having a correlation degree higher than a preset correlation threshold are determined; wherein, in a single target feature node combination, a correlation degree between a first target feature node and a second target feature node is higher than the preset correlation threshold; According to the node dependency in each target feature node combination, the regression coefficient of the corresponding target dimension feature is compensated; The sensitive value of each dimensional feature is determined according to the absolute values ​​corresponding to the compensation regression coefficient of the target dimensional feature and the regression coefficient of the non-target dimensional feature; the higher the absolute value, the higher the sensitive value.

2. The method according to claim 1, characterized in that The step of compensating the regression coefficient of the corresponding target dimension feature according to the node dependency relationship in each target feature node combination specifically includes: In the compensation coefficient mapping relationship table, the compensation coefficient corresponding to the correlation degree of each target feature node combination is retrieved; the compensation coefficient is greater than 1; In each target feature node combination, if the first target feature node depends on the second target feature node, the compensation coefficient is multiplied by the regression coefficient to obtain a compensation regression coefficient.

3. The method according to claim 1, characterized in that The updating of the standard piping hazard prediction network model according to the sensitive value to obtain the target piping hazard prediction network model of the hydraulic structure specifically includes: Calculate the first offset between the sensitive value of each dimensional feature and the set weight maximum value respectively; Obtaining a second offset between the maximum sensitivity score and the set weight maximum value; The ratio between the first offset and the second offset of each dimensional feature is calculated respectively to obtain the weight factor of each dimensional feature; The weight factor of each dimensional feature is multiplied by the weight of each dimensional feature in the standard piping hazard prediction network model to obtain the target piping hazard prediction network model of the hydraulic structure.

4. The method according to claim 1, characterized in that According to the probability distribution of each dimensional feature, the respective feature data risk groups are analyzed to obtain the potential risk range of the feature data of each dimensional feature, which specifically includes: Analyze each feature data risk group and extract the data value range of each dimension feature; In the probability distribution of each dimensional feature, the minimum probability corresponding to each dimensional feature is determined according to the data value range of each dimensional feature; In the probability distribution of each dimensional feature, data values ​​below the minimum probability are added to their respective data value ranges to obtain the potential risk range of the feature data.

5. The method according to claim 1, characterized in that: The step of predicting the piping hazard of the hydraulic structure according to the potential risk range of the characteristic data and the target piping hazard prediction network model specifically includes: Acquiring current dimensional characteristic data of the hydraulic structure; When a piping hazard prediction request is not automatically triggered, determining whether the current dimension feature data is within the potential risk range of its own feature data; If yes, a request is sent to a multi-dimensional feature data acquisition node, and when the latest multi-dimensional feature data is acquired, the multi-dimensional feature data is input into a target piping hazard prediction network model to obtain a piping risk coefficient of the hydraulic structure; According to the piping risk coefficient, piping hazard prediction is performed on the hydraulic structure.

6. The method according to claim 5, characterized in that The predicting of piping danger of the hydraulic structure according to the piping risk coefficient specifically includes: When the piping risk coefficient is lower than a preset risk coefficient threshold, obtaining meteorological information of the area where the hydraulic structure is located within a preset time interval; Extracting the rainfall of the hydraulic structure within the preset time interval from the meteorological information; When the rainfall is higher than a preset rainfall threshold, generating pipe burst danger warning information of the hydraulic structure, and sending the pipe burst danger warning information to a management user terminal; When the rainfall is lower than a preset rainfall threshold, obtaining a historical characteristic data change curve of the hydraulic structure for the current dimensional characteristic in the same time interval last year; Generate a feature data change curve of the current dimensional feature within the time interval; According to the comparison between the historical characteristic data change curve and the characteristic data change curve, the pipe burst hazard of the hydraulic structure is predicted.

7. The method according to claim 6, characterized in that The prediction of pipe burst danger of the hydraulic structure according to the comparison between the historical characteristic data change curve and the characteristic data change curve specifically includes: Determine whether the historical characteristic data change curve and the characteristic data change curve meet the trend consistency condition; If yes, compare the historical characteristic data change curve with the characteristic data change curve, Generate a volatility data series; Inputting the fluctuation data sequence into a pre-built fluctuation risk prediction network model to obtain a fluctuation risk coefficient of the hydraulic structure; When the fluctuation risk coefficient is higher than a preset fluctuation coefficient threshold, pipe burst hazard warning information of the hydraulic structure is generated.

8. The method according to claim 7, characterized in that The comparing the historical characteristic data change curve with the characteristic data change curve to generate a fluctuation data sequence specifically includes: Calculate the difference between each time point of the characteristic data change curve and multiple historical time points of the historical characteristic data change curve; The historical time point with the smallest difference is determined as the alignment time point for each time point; The difference between the data value at each time point in the characteristic data change curve and the data value at each aligned time point in the historical characteristic data change curve is calculated to obtain the fluctuation value at each time point and generate a fluctuation data sequence.

9. A device for predicting pipe burst danger of hydraulic structures, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for predicting pipe burst hazards for hydraulic structures as described in any one of claims 1 to 8.

Images