Drainage pipe network liquid level and risk prediction method

Abstract

The invention discloses a drainage pipe network liquid level and risk prediction method, relates to the technical field of drainage pipe networks, and solves the problems that in the prior art, a hydrology and hydrodynamics mechanism model needs a large amount of basic data and is complex in source, and if the data precision and accuracy are not enough, the deviation between a calculation result of the model and an actual result is large. The invention relates to a method for judging rainy days and screening rainfall periods according to rainfall data. A least square method is adopted to establish a multiple linear regression relation among the independent variable accumulative rainfall capacity, the rainfall duration and the dependent variable pipe network net rising liquid level, algorithm training is completed, and therefore the drainage pipe network liquid level is predicted; the critical rainfall and the rainfall are calculated through the set drainage pipe network liquid level threshold value, the rainfall is compared with the critical rainfall, and whether the drainage pipe network liquid level has risks or not can be judged; according to the method, drainage pipe network facilities and related topographic data do not need to be considered, and the over-fitting phenomenon possibly caused by complicated parameter training is avoided.

Description

Technical Field

[0001] The present invention belongs to the field of drainage pipe networks and relates to intelligent algorithm technology, and specifically is a method for predicting liquid level and risk of drainage pipe networks. Background Art

[0002] Predicting the liquid level in the drainage network is an effective way to detect and prevent drainage accidents in advance. By predicting the liquid level in the drainage network, we can understand the changes in the water level in the pipe in advance, so as to arrange the drainage operations reasonably. Under extreme weather conditions such as heavy rain, the prediction method can respond quickly and guide the drainage system to carry out drainage operations in advance, effectively prevent the occurrence of floods and waterlogging disasters, and improve the city’s flood control and drainage capabilities. By real-time monitoring and analysis of water level data in the drainage network and accurately predicting water level changes, water resources can be reasonably allocated to avoid waste and leakage of water resources and improve the efficiency of water resource utilization. At the same time, the prediction method also helps to promptly detect and deal with sewage leakage problems and protect water resources from pollution.

[0003] The existing technology predicts the liquid level of the drainage network based on the traditional urban hydrological and hydrodynamic model. The hydrological and hydrodynamic mechanism model requires a large amount of basic data with complex sources. Among them, structural data such as drainage pipes and ground elevations mostly come from surveying or drawings, and professional parameters mostly come from the experience of modelers. If the data precision and accuracy are not enough, the calculation results of the model will deviate greatly from the actual results. At the same time, if the scope of modeling is large, the number of nodes involved in the calculation of the model will be large, and the calculation time is usually long, resulting in reduced timeliness of the prediction.

[0004] The present invention provides a drainage network liquid level and risk prediction method to solve the above technical problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a drainage network liquid level and risk prediction method, which is used to solve the technical problem that when the prior art predicts the drainage network liquid level, the hydrological and hydrodynamic mechanism model requires a large amount of basic data and the source is complex, resulting in a large deviation between the model's calculation results and the actual results.

[0006] To achieve the above-mentioned object, a first aspect of the present invention provides a drainage network liquid level and risk prediction method, comprising:

[0007] S100: obtaining a number of drainage network inspection well liquid level monitoring points and rain gauge points from the database, monitoring the nearest liquid level monitoring point through the rain gauge, obtaining monitoring data and performing preprocessing;

[0008] S200: Determine the time interval and extract the rainfall period; construct and train a drainage network liquid level prediction model; wherein the drainage network liquid level prediction model is constructed based on a multivariate linear regression relationship;

[0009] S300: Set the liquid level threshold level, calculate the critical rainfall and rainfall; input the rainfall into the drainage network liquid level prediction model, and obtain the maximum predicted liquid level of the drainage network through the drainage network liquid level prediction model; compare the rainfall with the critical rainfall to determine whether there is a risk in the drainage network liquid level; if yes, output the risk level and the time when the risk occurs; if no, continue to determine.

[0010] Preferably, the acquiring of monitoring data and preprocessing includes:

[0011] Obtain the monitoring data of each liquid level monitoring point and the corresponding rain gauge for all historical periods, and remove invalid monitoring data with null values;

[0012] By formula Calculate the average value of effective monitoring data;

[0013] By formula The standard deviation is calculated; where x n is the monitoring data; n is the number of rainfall events, and the value range of n is: [2, +∞);

[0014] The values ​​within the interval [u-3δ, u+3δ] are regarded as normal values ​​of valid monitoring data.

[0015] The normal value range of valid monitoring data determined by the present invention through standard deviation helps to identify abnormal values; data points outside this range may indicate abnormal conditions, measurement errors or extreme events, which require further investigation or processing.

[0016] Preferably, the extraction of rainfall period includes:

[0017] Retrieve the normal value of effective monitoring data and set the start time of a rainfall period T s Time until the end of rainfall T E As the rainy day period, the historical rainfall monitoring data of the rain gauge is traversed from front to back and all the rainy day periods of each rain gauge are counted, and the monitoring data in the rainy period is accumulated to obtain the cumulative rainfall P;

[0018] The rainfall duration W of each rainfall is obtained by subtracting the rainfall start time from the rainfall end time;

[0019] The average hourly rainfall is obtained by dividing the accumulated rainfall by the rainfall duration;

[0020] Obtain the rainfall impact time and mark it as Δt; take the end time of the previous rainfall minus the start time of the next rainfall as the time interval between two adjacent rainfalls; determine whether the time interval is less than Δt; if so, merge the two rainfall periods and the subsequent Δt as a new rainy day period; if not, no merging is required.

[0021] It should be noted that the rainfall impact time refers to the time after the rainfall stops, during which the remaining water on the ground and in the drainage network continues to be discharged, thus affecting drainage. The actual value of Δt is set by the staff based on the operating conditions.

[0022] The present invention helps to accurately understand the intensity and duration of each rainfall by calculating the cumulative rainfall and rainfall duration, and provides a reliable data basis for rainfall analysis; by merging two adjacent rainfall periods with a time interval shorter than the rainfall impact time, a complete and effective rainfall period is reasonably obtained.

[0023] Preferably, the construction and training of the drainage network liquid level prediction model includes:

[0024] The accumulated rainfall and rainfall duration of each rainfall event are retrieved from the pre-processed historical monitoring data as independent variable data, as well as the maximum elevated liquid level dependent variable data, to form the initial sample data for algorithm training; the initial sample data is input into the drainage network liquid level prediction model;

[0025] Obtain the accumulated rainfall and rainfall duration of each corresponding liquid level monitoring point in the new rainy period; obtain the net increased liquid level of the pipe network through the absolute value of the difference between the maximum increased liquid level dependent variable data and the initial pipe network liquid level at the start of rainfall; merge the accumulated rainfall, rainfall duration, net increased liquid level of the pipe network and the initial sample data in the new rainy period as secondary sample data, and re-input them into the drainage pipe network liquid level prediction model.

[0026] It should be noted that in order to obtain the accumulated rainfall and rainfall duration of each corresponding liquid level monitoring point in the new rainy period, it is necessary to set the accuracy of the monitoring equipment and determine whether the calculated net increase in liquid level is less than the equipment accuracy value; if so, the data obtained by the corresponding monitoring equipment can be discarded; if not, the data obtained by the corresponding monitoring equipment will be used as sample data.

[0027] The present invention can ensure the data quality and accuracy of the input model by using preprocessed historical monitoring data as initial sample data; by introducing new monitoring data of rainy days and calculating the net rise in the pipe network liquid level, it can enrich the diversity of sample data and improve the model's adaptability to actual rainfall and liquid level changes; through the combined use of initial sample data and secondary sample data, the model can better learn the complex relationship between rainfall amount and rainfall duration and liquid level changes.

[0028] Preferably, the drainage network liquid level prediction model is constructed based on a multivariate linear regression relationship, including:

[0029] Retrieve sample data, calculate the average cumulative rainfall of multiple rainfall events and mark it as The average rainfall duration is marked as The average net rise in liquid level is marked as The sample data includes: initial sample data and secondary sample data;

[0030] By formula The variable coefficient b1 is calculated;

[0031] By formula The variable coefficient b2 is calculated; among them, Spp, Sww, Spw, Sph, Swh, Shh are all expressed as residual values;

[0032] By formula The constant term is calculated;

[0033] in:

[0034]

[0035] By formula Calculate the ratio of the regression sum of squares to the total deviation sum of squares;

[0036] Set the ratio threshold to determine R 2 Is it greater than or equal to the ratio threshold? If yes, the predicted net rise liquid level H can be obtained through the function H=a+b1×P+b2×W, and H is used as the output of the drainage network liquid level prediction model; where a is a constant term, and b1 and b2 are variable coefficients; if no, the calculation process is terminated.

[0037] It should be noted that the actual net increase in liquid level is marked as Hi; The relative error between the actual net liquid level rise and the predicted net liquid level rise is calculated, and rainfall with an absolute value of the relative error exceeding 100% is removed from the rainfall history database and does not participate in the calculation.

[0038] The present invention reflects the influence of cumulative rainfall and rainfall duration on the net liquid level rise by calculating the variable coefficient, which helps to understand the complex relationship between rainfall and liquid level change; by constructing the absolute value of the difference between multiple groups of actual values ​​and average values, it helps to eliminate abnormal values ​​with large errors.

[0039] Preferably, the setting of the liquid level threshold level includes:

[0040] The maximum predicted liquid level Hmax of the drainage network is retrieved, and the maximum liquid level of the drainage inspection well is set to three thresholds: Level Ⅰ, Ⅱ, and Ⅲ; among which Level Ⅲ is the lowest and Level Ⅰ is the highest;

[0041] The distance from the top of the downstream pipeline to the bottom of the well is marked as h1, and the distance from the wellhead to the bottom of the well is marked as h;

[0042] Set the level III threshold to the full pipe level of the downstream pipeline, then hIII = h1;

[0043] Retrieve the safe height and mark it as t, mark the level I threshold as hI, and set hI = ht;

[0044] The level II threshold is the middle value between the level III threshold and the level I threshold.

[0045] It should be noted that the safety height is the distance below the wellhead set to prevent overflow of the drainage network.

[0046] By setting three-level thresholds, namely, Level I, Level II, and Level III, the present invention can clearly define different risk levels, thereby achieving early warning of the liquid level risk of the drainage network; when the current liquid level is lower than the Level III threshold, it indicates that the drainage network is in a relatively safe state; and when the liquid level exceeds these thresholds, the early warning mechanism can be immediately triggered, and necessary preventive measures can be taken to avoid escalation of risks during subsequent rainfall.

[0047] Preferably, the critical rainfall is calculated by:

[0048] Set two critical rainfall index times a1 and a2, where a2>a1;

[0049] According to the trained multiple linear regression relationship H=a+b1×P+b2×W, according to the set liquid level thresholds of different levels and the critical rainfall index time a1, a2, H and W are respectively substituted into the multiple linear regression relationship to obtain the critical rainfall of the corresponding liquid level threshold level, and the critical rainfall corresponding to different liquid level threshold levels are marked as PI, PII, and PIII respectively.

[0050] Preferably, the calculation method of the rainfall includes:

[0051] Obtain forecast rainfall data from a third party, use the forecast rainfall data as a starting point to search for historical rainfall data at rainfall stations before the current time, and determine whether the time interval between the end time of the historical rainfall data and the start time of the forecast rainfall is less than the set rainfall time interval; if yes, combine the two rainfalls to obtain a complete rainfall period; if no, use the forecast rainfall period as the complete rainfall period;

[0052] The start time of the complete rainfall period is marked as ts, the end time of the complete rainfall period is marked as te, the current rainfall time is marked as t0, and each time after t0 is marked as tm i ; Wherein, i is the number of rainfall periods, i = 1, 2, 3…, n; n is a positive integer;

[0053] Judge tm i Whether the absolute value of the difference with t0 is greater than a1;

[0054] Yes, then statistics (tm i -a1,tm i ) and marked as Ph1, and judge tm i Is the absolute value of the difference from t0 greater than a2? If yes, then count (tm i -a2,tm i ) and marked as Ph2; otherwise, count (tm i -a1,tm i ) and marked as Ph3;

[0055] If no, rainfall is not calculated.

[0056] It should be noted that the rainfall interval is set by the staff based on the actual rainfall conditions.

[0057] The present invention combines the historical rainfall before the current moment and the future forecast rainfall to form a complete rainfall period, and sets the rainfall time interval to determine whether the historical rainfall has an impact on the forecast rainfall, which is beneficial to continuous monitoring of the changing rainfall and makes the predicted maximum predicted liquid level of the drainage network more accurate.

[0058] Preferably, obtaining the maximum predicted liquid level of the drainage network through the drainage network liquid level prediction model includes:

[0059] The rainfall and duration of the entire rainfall period are input into the drainage network liquid level prediction model to obtain the net rise in the network liquid level value under this rainfall.

[0060] The liquid level value of the pipe network corresponding to the time when rainfall starts is marked as H0, and the net increase in liquid level is calculated by The sum of the current network liquid level value H0 is the maximum predicted liquid level Hmax of the drainage network.

[0061] Preferably, the determining whether there is a risk in the drainage network liquid level includes:

[0062] Retrieve the Ph1, Ph2, and Ph3 values ​​obtained during the forecast rainfall period, and determine whether the Ph1, Ph2, and Ph3 values ​​are all lower than the Level III threshold; if so, there is no risk in the drainage network liquid level corresponding to the forecast rainfall period; if not, there is a risk in the drainage network liquid level; use the highest risk level obtained as the predicted liquid level risk level, record the predicted liquid level risk levels to form a drainage network liquid level risk prediction table, and obtain all liquid level risk periods during the forecast rainfall period based on the drainage network liquid level risk prediction table.

[0063] The present invention helps to quickly identify whether there is a risk in the liquid level of the pipe network under the current rainfall conditions by judging whether the rainfall index at the predicted rainfall time is lower than the critical rainfall of level III. At the same time, it helps to take preventive measures in time to reduce the problems of poor drainage and waterlogging caused by excessively high liquid levels. By retrieving Ph1, Ph2, and Ph3 values ​​and judging whether the Ph1, Ph2, and Ph3 values ​​are all lower than the critical rainfall of level III, it can be predicted whether there is a liquid level risk in the drainage pipe network, which helps to plan maintenance resources in advance, optimize scheduling, and ensure that effective response measures are taken before the risk occurs. According to the drainage pipe network liquid level risk prediction table, time periods and areas that need to be focused on can be identified, thereby optimizing resource allocation.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The drainage network liquid level prediction method based on historical data statistics proposed in the present invention obtains the liquid level monitoring points and all rain gauge points of the drainage network, matches each liquid level monitoring point with the rain gauge closest to it, and obtains relatively realistic rainfall, rainfall duration and other data for each liquid level monitoring point, thereby ensuring the reliability of the sample data of the liquid level prediction model; when the model is calculated for the first time, the monitoring data of all historical periods of each liquid level monitoring point and the corresponding rain gauge are obtained, and then the algorithm training is started regularly, and the historical monitoring data is merged with the sample data used in the last training to form new sample data, thereby ensuring the real-time nature of the sample data of the liquid level prediction model, and at the same time continuously accumulating historical rainfall and network liquid level data to improve the accuracy of the model.

[0066] 2. The present invention realizes the prediction of the network liquid level by establishing a multivariate linear relationship between rainfall index and network liquid level, and takes the accumulated rainfall and rainfall time as the main factors affecting the change of drainage network liquid level. It can not only characterize the situation where the total rainfall leads to the rise of liquid level, but also highlight the wide range of scenarios where the instantaneous intensity of rainfall affects the change of liquid level. The selected rainfall index is reasonable and covers a wide range of rainfall scenarios. At the same time, it does not rely on a large amount of basic data such as drainage network and related facilities, and the data acquisition is convenient and the model calculation accuracy is high.

[0067] 3. The method proposed in the present invention can predict whether there is a risk in the drainage network liquid level, obtain the risk level and the time when the risk occurs, based on the actual rainfall data and dynamic forecast data, by real-time and rolling calculation of the critical rainfall index of rainfall, and compare and judge with the defined index threshold, with fast prediction time and high efficiency. This method does not directly predict the drainage network liquid level risk by predicting the network liquid level, which indirectly reduces the error in liquid level prediction, has high prediction accuracy, and can timely and effectively discover and prevent the occurrence of drainage accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] 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.

[0069] Figure 1 It is a schematic diagram of the process of the pipeline network liquid level prediction method of the present invention;

[0070] Figure 2 It is a schematic diagram of a specific process of monitoring data processing of the present invention;

[0071] Figure 3 A rainy day period diagram for traversing rainfall monitoring data of the present invention;

[0072] Figure 4 This is a schematic diagram of the rainy day judgment logic of the present invention;

[0073] Figure 5 A schematic diagram of the specific process of model training of the present invention;

[0074] Figure 6 A schematic diagram of setting the liquid level threshold level of the present invention;

[0075] Figure 7 A schematic diagram of the steps of obtaining rainfall of different critical rainfall indicators according to the present invention;

[0076] Figure 8 Schematic diagram of the steps for judging the liquid level risk of the drainage network according to the present invention. DETAILED DESCRIPTION

[0077] The technical solution of the present invention will be described clearly and completely in conjunction with the embodiments below. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Please refer to Figure 1The first embodiment of the present invention provides a method for predicting the liquid level and risk of a drainage network, including:

[0078] S100: Obtain several drainage network inspection well liquid level monitoring points and rain gauge points from the database, monitor the nearest liquid level monitoring point through the rain gauge, obtain monitoring data and pre-process; S200: Determine the time interval and extract the rainfall period; Construct and train the drainage network liquid level prediction model; wherein, the drainage network liquid level prediction model is constructed based on the multivariate linear regression relationship; S300: Set the liquid level threshold level, calculate the critical rainfall and rainfall; input the rainfall into the drainage network liquid level prediction model, and obtain the maximum predicted liquid level of the drainage network through the drainage network liquid level prediction model; compare the rainfall with the critical rainfall to determine whether there is a risk in the drainage network liquid level; if yes, output the risk level and the time when the risk occurs; if not, continue to judge.

[0079] See also Figure 2 , the specific process of monitoring data processing, obtains the monitoring data of each liquid level monitoring point and the corresponding rain gauge in all historical periods, and removes invalid monitoring data with null values; through the formula Calculate the average value of effective monitoring data;

[0080] By formula The standard deviation is calculated; where x n is the monitoring data; n is the number of rainfall events, and the value range of n is: [2, +∞);

[0081] The value within the interval [u-3δ, u+3δ] is taken as the normal value of the effective monitoring data. The normal value of the effective monitoring data is retrieved and the start time of a rainfall period T is taken as E Time until the end of rainfall T s As the rainy day period, the historical rainfall monitoring data of the rain gauge is traversed from the beginning to the end and all the rainy day periods of each rain gauge are counted. The monitoring data in the rainy period is accumulated to obtain the cumulative rainfall P (mm); the rainfall duration W (min) of each rainfall is obtained by subtracting the rainfall start time from the rainfall end time; the average hourly rainfall is obtained by dividing the cumulative rainfall by the rainfall duration;

[0082] Get the rainfall impact time and mark it as Δt; take the end time of the previous rainfall minus the start time of the next rainfall as the time interval between two adjacent rainfalls; determine whether the time interval is less than Δt; if yes, merge the two rainfall periods and the subsequent Δt as a new rainy day period; if no, no merging is required. For example, now obtain the monitoring data of each liquid level monitoring point and the corresponding rain gauge for all historical periods, and remove the invalid monitoring data with null values;

[0083] As shown in the figure below, Figure 3is each rainfall period Ti (ti0~ti1) obtained after traversing the rainfall monitoring data; where Ti, ti0, ti1 represent the rainfall time, start time and end time of the i-th rainfall period respectively;

[0084] Figure 4 For Figure 3 The logic diagram of rainy day judgment after merging based on the above. The time interval (t11-t20) between the rainfall in T1 (t10-t11) and the rainfall in T2 (t20-t21) is less than Δt. The two rainfalls are merged, and T12 (t10-t21), T3 (t30-t31), T4 (t40-t41), and T5 (t50-t51) are extended by Δt to form rainy day periods Y1, Y2, Y3, and Y4. Please refer to Figure 5 , the specific process of model training, retrieve the pre-processed historical monitoring data to obtain the cumulative rainfall and rainfall duration independent variable data of each rainfall, as well as the maximum liquid level rise dependent variable data, to form the initial sample data for algorithm training; input the initial sample data into the drainage network liquid level prediction model; obtain the cumulative rainfall and rainfall duration of each corresponding liquid level monitoring point in the new rainy day period; obtain the net liquid level rise of the network through the absolute value of the difference between the maximum liquid level rise dependent variable data and the initial network liquid level at the start of the rainfall; merge the cumulative rainfall, rainfall duration, net liquid level rise of the network and the initial sample data in the new rainy day period as secondary sample data, and re-input them into the drainage network liquid level prediction model;

[0085] The average cumulative rainfall of multiple rainfall events is calculated based on the sample data and marked as The average rainfall duration is marked as The average net rise in liquid level is marked as The sample data includes: initial sample data and secondary sample data;

[0086] By formula The variable coefficient b1 is calculated;

[0087] By formula The variable coefficient b2 is calculated; among them, Spp, Sww, Spw, Sph, Swh, Shh are all expressed as residual values;

[0088] By formula The constant term is calculated;

[0089] Through the function Get the predicted net rise level Will As the output of the drainage network liquid level prediction model; where a is a constant term, b1 and b2 are variable coefficients;

[0090] By formula Calculate the ratio of the regression sum of squares to the total deviation sum of squares;

[0091] Judge R 2 Is it greater than or equal to 0.6? If yes, the fitting function is used for the next step of drainage network liquid level prediction. If not, the calculation process ends. The network liquid level value corresponding to the start of rainfall is marked as H0. By calculating the net rise liquid level value The sum of the current network liquid level value H0 is the maximum predicted liquid level Hmax of the drainage network.

[0092] For example, 13 rainfall events in a certain place are selected to form a rainfall process data statistical table (as shown in Table 1). The cumulative rainfall range is 13.5mm-53mm, and the rainfall duration range is 50min-1610min. The parameters of rainfall events No. 1 and 2, 7 and 8 are similar. So rainfall events No. 2 and 7 are selected as the network liquid level prediction verification respectively. The remaining rainfall events are used for the training of the multivariate linear regression algorithm. The prediction results are compared and analyzed, and the data in Table 1 are input into the drainage network liquid level prediction model.

[0093] Table 1 Rainfall process data record table

[0094]

[0095] in:

[0096]

[0097] The actual net raised liquid level is marked as Hi;

[0098] By formula The relative error between the actual net liquid level rise and the predicted net liquid level rise is calculated;

[0099]

[0100] The rainfall event No. 2 was selected as the prediction, and the remaining 12 rainfall events were used for multivariate linear regression analysis.

[0101] The least squares method was used to conduct a multivariate linear regression analysis on the net rise in the pipe network liquid level (m), rainfall duration (min) and cumulative rainfall (mm) in 12 rainfall events. The analysis results showed that there was a quantitative relationship between the variables, as follows:

[0102] The net rise in the liquid level of the pipe network = a + b1 × rainfall duration + b2 × accumulated rainfall;

[0103] The training parameters obtained are: a=0.613911, b1=-0.000530, b2=0.022133, R2=0.738.

[0104] Based on the quantitative relationship of the multiple linear regression analysis results, the net rise in the liquid level of the pipe network in the No. 2 rainfall event is predicted. The relative error between the multiple linear regression and the prediction results is shown in the following table (2). The actual net rise in the liquid level is 0.82m, and the predicted net rise in the liquid level is 0.89m. The relative error is 8.5%, and the prediction result has a high accuracy rate.

[0105] Table 2 Comparison of actual and predicted net rise in water level for rainfall event No. 2

[0106]

[0107] The rainfall event No. 7 was selected as the prediction, and the remaining 12 rainfall events were used for multivariate linear regression analysis.

[0108] The least squares method was used to conduct a multivariate linear regression analysis on the net rise in the pipe network liquid level (m), rainfall duration (min) and cumulative rainfall (mm) in 12 rainfall events. The analysis results showed that there was a quantitative relationship between the variables, as follows:

[0109] Net rise in pipe network liquid level = a + b1 × rainfall duration + b2 × cumulative rainfall

[0110] The training parameters obtained are: a=0.569172, b1=-0.000499, b2=0.022641, R2=0.747.

[0111] Based on the quantitative relationship of the multiple linear regression analysis results, the net rise in the liquid level of the pipeline network in the rainfall event No. 2 is predicted. The relative error between the multiple linear regression and the prediction results is shown in the following table (3); the actual net rise in the liquid level is 1.27m, and the predicted net rise in the liquid level is 1.10m. The relative error is 9.2%, and the prediction result has a high accuracy rate.

[0112] Table 3 Comparison of actual and predicted net rise in water level for rainfall event No. 7

[0113]

[0114] See also Figure 6 , set the liquid level threshold level, retrieve the maximum predicted liquid level Hmax of the drainage network, and set the drainage inspection well liquid level to three levels of thresholds: Ⅰ, Ⅱ, and Ⅲ; among which, Ⅲ is the lowest and Ⅰ is the highest;

[0115] The distance from the top of the downstream pipeline to the bottom of the well is marked as h1, and the distance from the wellhead to the bottom of the well is marked as h;

[0116] Set the level III threshold to the full pipe level of the downstream pipeline, then hIII = h1;

[0117] Retrieve the safe height and mark it as t, mark the level I threshold as hI, and set hI = ht;

[0118] The level II threshold is the middle value between the level III threshold and the level I threshold.

[0119] See also Figure 7 , obtain the rainfall of different critical rainfall indicators, set two critical rainfall indicator times a1, a2; where a2>a1;

[0120] According to the trained multiple linear regression relationship H=a+b1×P+b2×W, according to the set liquid level thresholds of different levels and the critical rainfall index time a1, a2, H and W are respectively substituted into the multiple linear regression relationship to obtain the critical rainfall of the corresponding liquid level threshold level, and the critical rainfall corresponding to different liquid level threshold levels are marked as PI, PII, and PIII respectively.

[0121] Obtain forecast rainfall data from a third party, use the forecast rainfall data as a starting point to search for historical rainfall data at rainfall stations before the current time, and determine whether the time interval between the end time of the historical rainfall data and the start time of the forecast rainfall is less than the set rainfall time interval; if yes, combine the two rainfalls to obtain a complete rainfall period; if no, use the forecast rainfall period as the complete rainfall period;

[0122] The start time of the complete rainfall period is marked as ts, the end time of the complete rainfall period is marked as te, the current rainfall time is marked as t0, and each time after t0 is marked as tm i ; Wherein, i is the number of rainfall periods, i = 1, 2, 3…, n; n is a positive integer;

[0123] Judge tm i Is the absolute value of the difference from t0 greater than a1? If so, count the rainfall in the a1 period and mark it as Ph1, and determine tm i Is the absolute value of the difference from t0 greater than a2? If yes, then count (tm i -a1,tm i ) and marked as Ph2; otherwise, count (tm i -a2,tm i ) and marked as Ph3;

[0124] If no, rainfall is not calculated.

[0125] For example, set the two critical rainfall index times a1 to 1 hour and a2 to 3 hours; obtain the actual data of the rainfall station closest to the monitoring point and the forecast rainfall data of the monitoring point, combine the actual rainfall with the forecast rainfall, and calculate the 1-hour rainfall and 3-hour rainfall in a rolling manner; set the rainfall time interval to 15 minutes;

[0126] Bring the set level III threshold into the multivariate linear regression relationship.

[0127] The critical 1-hour rainfall of level III can be obtained by setting H and W to 60 and 180, representing 1 hour and 3 hours of rainfall duration respectively. Critical 3-hour rainfall

[0128] According to the above method, the critical rainfall value of each monitoring point is calculated and recorded in the three-level threshold table of critical rainfall of the drainage network, as shown in Table 4 below;

[0129] Table 4 Three-level thresholds of critical rainfall in drainage network

[0130]

[0131]

[0132] Taking the current starting time t0 as the starting time, obtain the historical rainfall data of the rainfall station, and according to the above-mentioned rainfall period extraction and rainy day judgment method, obtain the historical rainfall period of ts-t0, superimpose the forecast rainfall period of t0-te, and obtain the complete rainfall period of ts-te;

[0133] During the forecast rainfall period of t0-te, the rainfall index is calculated according to the rainfall data interval ΔT, and the tm of each interval ΔT after t0 is traversed. i time, if tm i -t0>1h, rolling calculation of each tm i The rainfall in the hour before time Ph1; if tm i - If t0<1h, the rainfall in the previous hour will not be calculated;

[0134] If tm i -t0>3h, rolling calculation of each tm i The rainfall in the first 3 hours of the moment is Ph3;

[0135] If tm i - If t0<3h, the rainfall in the first 3 hours will not be calculated;

[0136] Thus, the rainfall amount and end time of several 1-hour and 3-hour rainfall periods are obtained.

[0137] See also Figure 8, the steps of judging the risk of liquid level in the drainage network are to retrieve the rainfall value obtained during the forecast rainfall period to determine whether it is lower than the critical rainfall of Level III; if so, there is no risk in the liquid level of the drainage network corresponding to the forecast rainfall period; if not, there is a risk in the liquid level of the drainage network; the highest risk level obtained is used as the predicted liquid level risk level, and the predicted liquid level risk levels are recorded to form a drainage network liquid level risk prediction table, and all liquid level risk periods in the rainfall period are obtained according to the drainage network liquid level risk prediction table.

[0138] For example, t m,i The rainfall in the previous hour and the rainfall in the previous 3 hours are compared with the third-level critical rainfall respectively to determine the risk level of the rainfall in the previous hour and the rainfall in the previous 3 hours. If both are less than level III, it means that the liquid level of the drainage network is temporarily free of risk. Otherwise, the highest risk level determined by the rainfall in the previous hour and the rainfall in the previous 3 hours is taken as the predicted liquid level risk level. The predicted liquid level risk level at the time of the forecast rainfall, the rainfall in the 1-hour or 3-hour rainfall period, and the end time tm,i are recorded, as shown in Table 5 below;

[0139] Table 5 Drainage network liquid level risk prediction table

[0140] Ending time <![CDATA[1-hour rainfall amount P h > <![CDATA[3-hour rainfall amount P 3h > Liquid level risk level <![CDATA[t e,1 ]]> <![CDATA[P h,1 ]]> — <![CDATA[X1]]> <![CDATA[t e,2 ]]> <![CDATA[P h,2 ]]> — <![CDATA[X2]]> <![CDATA[t e,3 ]]> <![CDATA[P h,3 ]]> — <![CDATA[X3]]> <![CDATA[t e,4 ]]> <![CDATA[P h,4 ]]> — <![CDATA[X4]]> <![CDATA[t e,5 ]]> <![CDATA[P h,5 ]]> — <![CDATA[X5]]> <![CDATA[t e,6 ]]> <![CDATA[P h,6 ]]> — <![CDATA[X6]]> … … … … <![CDATA[t e,i ]]> <![CDATA[P h,i ]]> <![CDATA[P 3h,i ]]> <![CDATA[X i ]]> … … … … <![CDATA[t e,n ]]> <![CDATA[P h,n ]]> <![CDATA[P 3h,n ]]> <![CDATA[X n ]]>

[0141] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0142] The working principle of the present invention is as follows: the present invention obtains a number of drainage network inspection well liquid level monitoring points and rain gauge points from a database, monitors the liquid level monitoring point closest to it through a rain gauge, obtains monitoring data and performs preprocessing; determines the time interval, and extracts the rainfall period; constructs and trains a drainage network liquid level prediction model; wherein the drainage network liquid level prediction model is constructed based on a multivariate linear regression relationship; sets a liquid level threshold level, sets a critical rainfall index and obtains the critical rainfall; inputs the rainfall into the drainage network liquid level prediction model, and obtains the maximum predicted liquid level of the drainage network through the drainage network liquid level prediction model; determines whether there is a risk in the drainage network liquid level corresponding to the rainfall; if yes, issues a warning signal; if not, continues to determine.

[0143] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A drainage network liquid level and risk prediction method, characterized in that: include: S100: obtaining a number of drainage network inspection well liquid level monitoring points and rain gauge points from the database, monitoring the nearest liquid level monitoring point through the rain gauge, obtaining monitoring data and performing preprocessing; S200: Determine the time interval and extract the rainfall period; construct and train a drainage network liquid level prediction model; wherein the drainage network liquid level prediction model is constructed based on a multivariate linear regression relationship; S300: setting the liquid level threshold level, calculating the critical rainfall and rainfall; inputting the rainfall into the drainage network liquid level prediction model, and obtaining the maximum predicted liquid level of the drainage network through the drainage network liquid level prediction model; comparing the rainfall with the critical rainfall to determine whether there is a risk in the drainage network liquid level; If yes, output the risk level and the time when the risk occurs; if no, continue to judge.

2. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The obtaining of monitoring data and preprocessing includes: Obtain the monitoring data of each liquid level monitoring point and the corresponding rain gauge for all historical periods, and remove invalid monitoring data with null values; By formula Calculate the average value of effective monitoring data; By formula The standard deviation is calculated; where x n is the monitoring data; n is the number of rainfall events, and the value range of n is: [2, +∞); The values ​​within the interval [u-3δ, u+3δ] are regarded as normal values ​​of valid monitoring data.

3. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The step of extracting the rainfall period comprises: Retrieve the normal value of effective monitoring data and set the start time of a rainfall period T s Time until the end of rainfall T E As the rainy day period, the historical rainfall monitoring data of the rain gauge is traversed from front to back and all the rainy day periods of each rain gauge are counted, and the monitoring data in the rainy period is accumulated to obtain the cumulative rainfall P; The rainfall duration W of each rainfall is obtained by subtracting the rainfall start time from the rainfall end time; The average hourly rainfall is obtained by dividing the accumulated rainfall by the rainfall duration; Obtain the rainfall impact time and mark it as Δt; take the end time of the previous rainfall minus the start time of the next rainfall as the time interval between two adjacent rainfalls; determine whether the time interval is less than Δt; if so, merge the two rainfall periods and the subsequent Δt as a new rainy day period; if not, no merging is required.

4. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The construction and training of the drainage network liquid level prediction model includes: The accumulated rainfall and rainfall duration of each rainfall event are retrieved from the pre-processed historical monitoring data as independent variable data, as well as the maximum elevated liquid level dependent variable data, to form the initial sample data for algorithm training; the initial sample data is input into the drainage network liquid level prediction model; Obtain the accumulated rainfall and rainfall duration of each corresponding liquid level monitoring point in the new rainy period; obtain the net increased liquid level of the pipe network through the absolute value of the difference between the maximum increased liquid level dependent variable data and the initial pipe network liquid level at the start of rainfall; merge the accumulated rainfall, rainfall duration, net increased liquid level of the pipe network and the initial sample data in the new rainy period as secondary sample data, and re-input them into the drainage pipe network liquid level prediction model.

5. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The drainage network liquid level prediction model is constructed based on a multivariate linear regression relationship and includes: Retrieve sample data, calculate the average cumulative rainfall of multiple rainfall events and mark it as The average rainfall duration is marked as The average net rise in liquid level is marked as The sample data includes: initial sample data and secondary sample data; By formula The variable coefficient b1 is calculated; By formula The variable coefficient b2 is calculated; among them, Spp, Sww, Spw, Sph, Swh, Shh are all expressed as residual values; By formula The constant term is calculated; in: By formula Calculate the ratio of the regression sum of squares to the total deviation sum of squares; Set the ratio threshold to determine R 2 Is it greater than or equal to the ratio threshold? If yes, the predicted net rise liquid level H can be obtained through the function H=a+b1×P+b2×W, and H is used as the output of the drainage network liquid level prediction model; where a is a constant term, and b1 and b2 are variable coefficients; if no, the calculation process is terminated.

6. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The setting of the liquid level threshold level includes: The maximum predicted liquid level Hmax of the drainage network is retrieved, and the liquid level of the drainage inspection well is set to three thresholds: Ⅰ, Ⅱ, and Ⅲ; among which, Ⅲ is the lowest and Ⅰ is the highest; The distance from the top of the downstream pipeline to the bottom of the well is marked as h1, and the distance from the wellhead to the bottom of the well is marked as h; Set the level III threshold to the full pipe level of the downstream pipeline, then hIII = h1; Retrieve the safe height and mark it as t, mark the level I threshold as hI, and set hI = ht; The level II threshold is the middle value between the level III threshold and the level I threshold.

7. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The calculation method of the critical rainfall includes: Set two critical rainfall index times a1 and a2, where a2>a1; According to the trained multiple linear regression relationship H=a+b1×P+b2×W, according to the set liquid level thresholds of different levels and the critical rainfall index time a1, a2, H and W are respectively substituted into the multiple linear regression relationship to obtain the critical rainfall of the corresponding liquid level threshold level, and the critical rainfall corresponding to different liquid level threshold levels are marked as PI, PII, and PIII respectively.

8. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The calculation method of the rainfall includes: Obtain forecast rainfall data from a third party, use the forecast rainfall data as a starting point to search for historical rainfall data at rainfall stations before the current time, and determine whether the time interval between the end time of the historical rainfall data and the start time of the forecast rainfall is less than the set rainfall time interval; if yes, combine the two rainfalls to obtain a complete rainfall period; if no, use the forecast rainfall period as the complete rainfall period; The start time of the complete rainfall period is marked as ts, the end time of the complete rainfall period is marked as te, the current rainfall time is marked as t0, and each time after t0 is marked as tm i ; Wherein, i is the number of rainfall periods, i = 1, 2, 3…, n; n is a positive integer; Judge tm i Whether the absolute value of the difference with t0 is greater than a1; Yes, then statistics (tm i -a1,tm i ) and marked as Ph1, and judge tm i Is the absolute value of the difference from t0 greater than a2? If yes, then count (tm i -a2,tm i ) and is marked as Ph2; If not, then statistics (tm i -a1,tm i ) and marked as Ph3; If no, rainfall is not calculated.

9. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The method of obtaining the maximum predicted liquid level of the drainage network by using the drainage network liquid level prediction model includes: The accumulated rainfall and rainfall duration of the complete rainfall period are input into the drainage network liquid level prediction model to obtain the net rise liquid level value of the network under the corresponding rainfall. The network liquid level value corresponding to the start of rainfall is marked as H0, and the net rise liquid level value is calculated by The sum of the current network liquid level value H0 is the maximum predicted liquid level Hmax of the drainage network.

10. A drainage network liquid level and risk prediction method according to claim 1, characterized in that: The determining whether there is a risk in the drainage network liquid level includes: Retrieve the Ph1, Ph2, and Ph3 values ​​obtained during the forecast rainfall period, and determine whether the Ph1, Ph2, and Ph3 values ​​are all lower than the Level III threshold; if so, there is no risk in the drainage network liquid level corresponding to the forecast rainfall period; if not, there is a risk in the drainage network liquid level; use the highest risk level obtained as the predicted liquid level risk level, record the predicted liquid level risk levels to form a drainage network liquid level risk prediction table, and obtain all liquid level risk periods during the forecast rainfall period based on the drainage network liquid level risk prediction table.

Images