Multi-sensor water level automatic monitoring and early warning method, device and equipment and storage medium

Abstract

The invention discloses a multi-sensor water level automatic monitoring and early warning method, device and equipment and a storage medium, and belongs to the field of hydraulic engineering, and the method comprises the steps: monitoring water level source data through monitoring station water level multi-sensor equipment, and carrying out the preprocessing (data interpolation and abnormal value correction) of the water level source data, so as to achieve the continuity of the data; the method comprises the steps that firstly, source data are obtained, CEEMDAN decomposition is conducted on the source data, effective signals are extracted, the extracted effective signals serve as characteristic values to be substituted into a GRU to be trained, predicted IMF component results are overlapped to obtain a post-rental predicted value, WMAE, WRMSE and WR2 serve as water level early warning precision evaluation factors to monitor water level abnormity, and finally early warning data are stored in a ground database; according to the method, accurate prediction of the large water level is achieved, the characteristics of source water level data are reserved, the method has obvious advantages in the aspect of intelligent water level prediction, and a reliable scientific basis is provided for establishing a scientific reservoir water level prediction and early warning decision model and strengthening system management and operation data.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy projects, and in particular to a multi-sensor water level automatic monitoring and early warning method, device, equipment and storage medium. Background Art

[0002] With the frequent occurrence of extreme weather, the accelerated urbanization process and the increasing demand for efficiency and accuracy of hydrological monitoring in urban reservoirs, traditional hydrological monitoring and changes in climate environment cannot provide accurate forecasts of water levels, and the forecast response is low, posing a serious threat to the safety of life and property of people around the reservoirs.

[0003] With the continuous development of 5G technology, especially the application of wireless communication technology and the development of multi-sensors, hydrological monitoring has entered a new stage of intelligence and automation. The importance of water level monitoring in flood control, drainage and water resources management is becoming increasingly prominent. Urban areas are often densely populated, and once disasters such as floods occur, the consequences will be disastrous. In order to improve the efficiency and accuracy of water level monitoring, reduce labor costs, and realize real-time monitoring and early warning of hydrological dynamics, the present invention has developed a multi-sensor water level automatic monitoring and early warning method, device, equipment and storage medium, established a CEEMDAN-GRU water level intelligent prediction model, and realized accurate prediction of water level monitoring stations, which provides a scientific and effective method for establishing a scientific reservoir water level prediction and early warning decision-making model, strengthening system management and operation data, and providing strong support for water resources management, ecological protection and disaster response. Summary of the Invention

[0004] The present invention addresses the problems of low water level monitoring efficiency and accuracy, as well as high labor costs, and achieves real-time monitoring and early warning of hydrological dynamics. Water level source data is monitored by multi-sensor equipment at monitoring stations. The water level source data is preprocessed (data interpolation and outlier correction) to achieve data continuity. The source data is then decomposed by CEEMDAN to extract valid signals. The extracted valid signals are then used as eigenvalues for GRU training. The predicted IMF component results are superimposed to obtain post-lease prediction values. Water level anomalies are monitored using WMAE, WRMSE, and WR2 as water level early warning accuracy evaluation factors. Finally, the early warning data is stored in a ground database. The present invention can be widely applied to the field of water level monitoring and early warning technology.

[0005] Technical solution: To solve the above technical problems, according to one aspect of the present invention, more specifically, a multi-sensor water level automatic monitoring and early warning method includes the following steps:

[0006] Step 1: The water level multi-sensor at the monitoring station detects the water level and obtains effective water level source data;

[0007] Step 2: After obtaining the water level source data in step 1, preprocess the data; the present invention uses data interpolation and outlier correction to obtain valid water level data;

[0008] Step 3: After the preprocessing in step 2, the water level source data is decomposed into effective signals using CEEMDAN;

[0009] Step 4: Set the water level source data as the original signal, that is, w(t), and add N groups of WN noise ρ with a mean of 0 to w(t) i (t), then the mathematical model of the signal at time i is:

[0010] w i (t)=w(t)+ρ i (t)

[0011] In the mathematical model of the signal, i = 1, 2, ..., N is the number of repetitions; t is a certain moment of the signal;

[0012] Step 5: i (t) Perform CEEMDAN decomposition, the corresponding first modal component Then calculate the first IMF of N groups of signals, the mathematical model is:

[0013]

[0014] Step 6: Based on step 5, use w(t) to subtract IMF1 to obtain the residual component r1. The mathematical model is:

[0015] r1=w(t)-IMF1

[0016] Step 7: Solve IMF2 in sequence and continue to add ρ to r1 i (t), forming a new signal to be decomposed r'1(t), the mathematical model is:

[0017] r'1(t)=r1(t)+ρ i (t)

[0018] Step 8: According to step 7, iterate and update again to obtain IMF2 and r2 as:

[0019]

[0020] r2=w(t)-IMF2

[0021] Step 9: Repeat steps 7 and 8 until the signal is monotonic. After decomposition, k IMF components are obtained. The mathematical model of the signal w(t) is:

[0022]

[0023] Furthermore, effective signal feature extraction includes the following steps:

[0024] Step 10: The obtained signal is effectively extracted; the correlation coefficient method R is used to determine the high-frequency and low-frequency signals, and the IMF component of each CEEMDAN decomposition is calculated. When R is the first minimum point, the IMFR is the last high-frequency noise w(t) and the mathematical model of R between the IMF is:

[0025]

[0026] Step 11: According to step 10, when R K The first minimum point is removed, and the remaining IMF components are reconstructed into the denoised signal. The mathematical model is:

[0027]

[0028] is the sum of the modal components from K+1 to m.

[0029] Furthermore, the effective signal is brought into the GRU as a feature for training, which includes the following steps:

[0030] Step 12: The GRU prediction model includes input gate, forget gate and output gate units. The present invention combines the input gate and forget gate into an update gate to improve prediction accuracy and solve the gradient problem and the problem of long convergence speed.

[0031] Step 13: The mathematical model of the GRU module gate recharge before and after update is:

[0032] Reset gate: r t =σ(W r [h t-1 ,x t ]+b r )

[0033] Update gate: z t =σ(W z [h t-1 ,x t ]+b z ).

[0034] Furthermore, after the water level is predicted, the water level warning accuracy evaluation factor is used to perform water level warning, including the following steps:

[0035] Step 14: The mathematical formulas of water level evaluation factors, namely water level mean absolute error WMAE, water level root mean square error WRMSE and water level determination coefficient WR2, are as follows:

[0036]

[0037] Among them, w k is the original water level time series, is the water level prediction result, is the average value of the raw water level time series data, and n is the number of raw water level time series data. The resulting value of the water level evaluation factor is in the millimeter level. If it exceeds the millimeter level, the ground control department must be notified in time to issue a water level warning. The relevant water level warning data is stored in the ground database.

[0038] According to another aspect of the present invention, more specifically, a multi-sensor water level automatic monitoring and early warning device includes:

[0039] The water level detection module is used to detect the water level through multiple water level sensors at the monitoring station to obtain effective water level source data;

[0040] A data preprocessing module, connected to the water level detection module, is used to preprocess the acquired water level source data, and obtain valid water level data by using data interpolation and outlier correction;

[0041] A signal decomposition module, connected to the data preprocessing module, for performing CEEMDAN effective signal decomposition on the preprocessed water level source data;

[0042] The effective signal extraction module is connected to the signal decomposition module, and uses the correlation coefficient method R to determine the high-frequency and low-frequency signals, and calculates the IMF components of each CEEMDAN decomposition. When R reaches the first minimum point, the first several IMFs are eliminated, and the remaining IMF components are reconstructed into the denoised signal;

[0043] A prediction module is connected to the effective signal extraction module, adopts a GRU prediction model, merges the input gate and the forget gate into an update gate, and uses the denoised signal as a feature value for training to predict water level changes;

[0044] An early warning module is connected to the prediction module and uses the water level mean absolute error (WMAE), the water level root mean square error (WRMSE) and the water level determination coefficient (WR2) as water level early warning accuracy evaluation factors. When the evaluation factor result value exceeds the millimeter level, a water level early warning message is generated.

[0045] The data storage module is connected to the early warning module and is used to store water level early warning related data into the ground database.

[0046] According to another aspect of the present invention, more specifically, a multi-sensor water level automatic monitoring and early warning device includes the aforementioned multi-sensor water level automatic monitoring and early warning device, and further includes:

[0047] A power supply module is used to provide power support for the monitoring and early warning device and other components of the equipment;

[0048] A communication module, connected to the monitoring and early warning device, is used to realize data transmission between the device and external equipment or systems, so as to send water level early warning information to the ground control department, etc.;

[0049] The control module is connected to the monitoring and early warning device, the power module and the communication module respectively, and is used to coordinate the work of various components of the equipment and control the flow and processing flow of data.

[0050] According to another aspect of the present invention, more specifically, a multi-sensor water level automated monitoring and early warning storage medium stores: program code for implementing a multi-sensor water level automated monitoring and early warning method, including code for implementing water level source data acquisition, data preprocessing, CEEMDAN effective signal decomposition, effective signal extraction, GRU prediction model training, and water level early warning accuracy evaluation;

[0051] Water level source data, preprocessed water level data, IMF component data after CEEMDAN decomposition, GRU prediction result data, and water level warning related data, among which the water level warning related data includes the calculation results of water level mean absolute error WMAE, water level root mean square error WRMSE, water level determination coefficient WR2, and warning information generated based on these results.

[0052] The beneficial effects of the multi-sensor water level automatic monitoring and early warning method, device, equipment and storage medium of the present invention are:

[0053] This method, on the one hand, extracts effective signals through CEEMDAN decomposition, preserving the complete trend characteristics of the water level source data. On the other hand, it uses the effective signals as features to train a deep learning GRU prediction model to predict water level changes. Finally, the performance of the prediction model is evaluated using WMAE, WRMSE, and WR2, thereby achieving accurate prediction of large water levels. This method retains the characteristics of the source water level data and has significant advantages in intelligent water level prediction. It provides a reliable scientific basis for establishing scientific reservoir water level prediction and early warning decision-making models, and strengthening system management and operational data. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0055] Figure 1 It is a flow chart of the method of the present invention;

[0056] Figure 2 is a decomposition diagram of a CEEMDAN signal according to an embodiment of the method of the present invention;

[0057] Figure 3is the distribution of CEEMDAN-GRU prediction values of an embodiment of the method of the present invention;

[0058] Figure 4 This is a comparison of the results of CEEMDAN-GRU and GRU in the embodiment of the method of the present invention;

[0059] Figure 5 This is an evaluation of the CEEMDAN-GRU prediction model results of the method embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0061] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Reference Figure 1-Figure 5 As shown, a multi-sensor water level automatic monitoring and early warning method includes the following steps:

[0063] Step 1: The water level multi-sensor at the monitoring station detects the water level and obtains effective water level source data;

[0064] Step 2: After obtaining the water level source data in step 1, preprocess the data; the present invention uses data interpolation and outlier correction to obtain valid water level data;

[0065] Step 3: After the preprocessing in step 2, the water level source data is decomposed into effective signals using CEEMDAN;

[0066] Step 4: Set the water level source data as the original signal, that is, w(t), and add N groups of WN noise ρ with a mean of 0 to w(t) i (t), then the mathematical model of the signal at time i is:

[0067] w i (t)=w(t)+ρ i (t)

[0068] In the mathematical model of the signal, i = 1, 2, ..., N is the number of repetitions; t is a certain moment of the signal;

[0069] Step 5: i (t) Perform CEEMDAN decomposition, the corresponding first modal component Then calculate the first IMF of N groups of signals, the mathematical model is:

[0070]

[0071] Step 6: Based on step 5, use w(t) to subtract IMF1 to obtain the residual component r1. The mathematical model is:

[0072] r1=w(t)-IMF1

[0073] Step 7: Solve IMF2 in sequence and continue to add ρ to r1 i (t), forming a new signal to be decomposed r'1(t), the mathematical model is:

[0074] r'1(t)=r1(t)+ρ i (t)

[0075] Step 8: According to step 7, iterate and update again to obtain IMF2 and r2 as:

[0076]

[0077] r2=w(t)-IMF2

[0078] Step 9: Repeat steps 7 and 8 until the signal is monotonic. After decomposition, k IMF components are obtained. The mathematical model of the signal w(t) is:

[0079]

[0080] Furthermore, effective signal feature extraction includes the following steps:

[0081] Step 10: The obtained signal is effectively extracted; the correlation coefficient method R is used to determine the high-frequency and low-frequency signals, and the IMF component of each CEEMDAN decomposition is calculated. When R is the first minimum point, the IMFR is the last high-frequency noise w(t) and the mathematical model of R between the IMF is:

[0082]

[0083] Step 11: According to step 10, when R K The first minimum point is removed, and the remaining IMF components are reconstructed into the denoised signal. The mathematical model is:

[0084]

[0085] is the sum of the modal components from K+1 to m.

[0086] Furthermore, the effective signal is brought into the GRU as a feature for training, which includes the following steps:

[0087] Step 12: The GRU prediction model includes input gate, forget gate and output gate units. The present invention combines the input gate and forget gate into an update gate to improve prediction accuracy and solve the gradient problem and the problem of long convergence speed. Figure 3 is the CEEMDAN-GRU prediction value distribution of the embodiment of the method of the present invention, Figure 4 The results of CEEMDAN-GRU and GRU in the embodiment of the method of the present invention are compared.

[0088] Step 13: The mathematical model of the GRU module gate recharge before and after update is:

[0089] Reset gate: r t =σ(W r [h t-1 ,x t ]+b r )

[0090] Update gate: z t =σ(W z [h t-1 ,x t ]+b z ).

[0091] Furthermore, after the water level is predicted, the water level warning accuracy evaluation factor is used to perform water level warning, including the following steps: Step 14: The water level evaluation factors are the water level mean absolute error WMAE, the water level root mean square error WRMSE and the water level determination coefficient WR2. The mathematical formula is:

[0092]

[0093]

[0094] Among them, w k is the original water level time series, is the water level prediction result, is the average value of the raw water level time series data, and n is the number of raw water level time series data. The resulting value of the water level evaluation factor is in the millimeter level. If it exceeds the millimeter level, the ground control department must be notified in time to issue a water level warning. The relevant water level warning data is stored in the ground database.

[0095] According to another aspect of the present invention, more specifically, a multi-sensor water level automatic monitoring and early warning device includes:

[0096] The water level detection module is used to detect the water level through multiple water level sensors at the monitoring station to obtain effective water level source data;

[0097] A data preprocessing module, connected to the water level detection module, is used to preprocess the acquired water level source data, and obtain valid water level data by using data interpolation and outlier correction;

[0098] A signal decomposition module, connected to the data preprocessing module, for performing CEEMDAN effective signal decomposition on the preprocessed water level source data;

[0099] The effective signal extraction module is connected to the signal decomposition module, and uses the correlation coefficient method R to determine the high-frequency and low-frequency signals, and calculates the IMF components of each CEEMDAN decomposition. When R reaches the first minimum point, the first several IMFs are eliminated, and the remaining IMF components are reconstructed into the denoised signal;

[0100] A prediction module is connected to the effective signal extraction module, adopts a GRU prediction model, merges the input gate and the forget gate into an update gate, and uses the denoised signal as a feature value for training to predict water level changes;

[0101] An early warning module is connected to the prediction module and uses the water level mean absolute error (WMAE), the water level root mean square error (WRMSE) and the water level determination coefficient (WR2) as water level early warning accuracy evaluation factors. When the evaluation factor result value exceeds the millimeter level, a water level early warning message is generated.

[0102] The data storage module is connected to the early warning module and is used to store water level early warning related data into the ground database.

[0103] According to another aspect of the present invention, more specifically, a multi-sensor water level automatic monitoring and early warning device includes the aforementioned multi-sensor water level automatic monitoring and early warning device, and further includes:

[0104] A power supply module is used to provide power support for the monitoring and early warning device and other components of the equipment;

[0105] A communication module, connected to the monitoring and early warning device, is used to realize data transmission between the device and external equipment or systems, so as to send water level early warning information to the ground control department, etc.;

[0106] The control module is connected to the monitoring and early warning device, the power module and the communication module respectively, and is used to coordinate the work of various components of the equipment and control the flow and processing flow of data.

[0107] According to another aspect of the present invention, more specifically, a multi-sensor water level automated monitoring and early warning storage medium stores: program code for implementing a multi-sensor water level automated monitoring and early warning method, including code for implementing water level source data acquisition, data preprocessing, CEEMDAN effective signal decomposition, effective signal extraction, GRU prediction model training, and water level early warning accuracy evaluation;

[0108] Water level source data, preprocessed water level data, IMF component data after CEEMDAN decomposition, GRU prediction result data, and water level warning related data, among which the water level warning related data includes the calculation results of water level mean absolute error WMAE, water level root mean square error WRMSE, water level determination coefficient WR2, and warning information generated based on these results.

[0109] Simulation experiment:

[0110] Table 1 evaluates the CEEMDAN-GRU prediction accuracy for vertical water level monitoring stations, using a precipitation value of 400 mm as a benchmark. Compared to existing GRU methods, CEEMDAN-GRU achieves significant improvements in accuracy. Station 2 achieves the highest improvements in WRMSE (60.40%), WMAE (33.73%), and WR2 (4.89%). Station 5 achieves the lowest improvements in WRMSE (12.11%), WMAE (2.35%), and WR2 (0.94%). Compared to the GRU model, CEEMDAN-GRU significantly improves accuracy, with the prediction curve closer to the station's measured water level, improving prediction accuracy.

[0111] Table 1. Evaluation of accuracy indexes.

[0112]

[0113] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0114] It should be understood that the above description of the preferred embodiment is relatively detailed and cannot be regarded as limiting the scope of protection of the patent of the present invention. Under the guidance of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which all fall within the scope of protection of the present invention. The scope of protection requested by the present invention shall be based on the attached claims.

Claims

1. A multi-sensor water level automatic monitoring and early warning method, characterized in that: The following steps are involved: Step 1: The water level multi-sensor at the monitoring station detects the water level and obtains effective water level source data; Step 2: After obtaining the water level source data in step 1, preprocess the data; use data interpolation and outlier correction to obtain valid water level data; Step 3: After the preprocessing in step 2, the water level source data is decomposed into effective signals using CEEMDAN; Step 4: Set the water level source data as the original signal, that is, w(t), and add N groups of WN noise ρ with a mean of 0 to w(t) i (t), then the mathematical model of the signal at time i is: w i (t)=w(t)+ρ i (t) In the mathematical model of the signal, i = 1, 2, ..., N is the number of repetitions; t is a certain moment of the signal; Step 5: i (t) Perform CEEMDAN decomposition, the corresponding first modal component Then calculate the first IMF of N groups of signals, the mathematical model is: Step 6: Based on step 5, use w(t) to subtract IMF1 to obtain the residual component r1. The mathematical model is: r1=w(t)-IMF1 Step 7: Solve IMF2 in sequence and continue to add ρ to r1 i (t), forming a new signal to be decomposed r'1(t), the mathematical model is: r'1(t)=r1(t)+ρ i (t) Step 8: According to step 7, iterate and update again to obtain IMF2 and r2 as: r2=w(t)-IMF2; Step 9: Repeat steps 7 and 8 until the signal is monotonic. After decomposition, k IMF components are obtained. The mathematical model of the signal w(t) is:

2. A multi-sensor water level automatic monitoring and early warning method according to claim 1, characterized in that: Effective signal feature extraction includes the following steps: Step 10: The obtained signal is effectively extracted; the correlation coefficient method R is used to determine the high-frequency and low-frequency signals, and the IMF component of each CEEMDAN decomposition is calculated. When R is the first minimum point, the IMFR is the last high-frequency noise w(t) and the mathematical model of R between the IMF is: Step 11: According to step 10, when R K The first minimum point is removed, and the remaining IMF components are reconstructed into the denoised signal. The mathematical model is: is the sum of the modal components from K+1 to m.

3. A multi-sensor water level automatic monitoring and early warning method according to claim 2, characterized in that: Bringing effective signals as features into GRU for training includes the following steps: Step 12: The GRU prediction model includes input gate, forget gate, and output gate units. The input gate and forget gate are merged into an update gate to improve prediction accuracy and solve the gradient problem and the problem of slow convergence speed. Step 13: The mathematical model of the GRU module gate recharge before and after update is: Reset gate: r t =σ(W r [h t-1 ,x t ]+b r ) Update gate: z t =σ(W z [h t-1 ,x t ]+b z ).

4. A multi-sensor water level automatic monitoring and early warning method according to claim 3, characterized in that: After the water level is predicted, the water level warning accuracy evaluation factor is used to perform water level warning, including the following steps: Step 14: The mathematical formulas of water level evaluation factors, namely water level mean absolute error WMAE, water level root mean square error WRMSE and water level determination coefficient WR2, are as follows: Among them, w k is the original water level time series, is the water level prediction result, is the average value of the original water level time series data, and n is the number of original water level time series data; the result value of the water level evaluation factor is in the millimeter level. If it exceeds the millimeter level, the ground control department must be notified in time to issue a water level warning. The water level warning related data is stored in the ground database.

5. A multi-sensor water level automatic monitoring and early warning device, characterized in that: Includes: The water level detection module is used to detect the water level through multiple water level sensors at the monitoring station to obtain effective water level source data; A data preprocessing module, connected to the water level detection module, is used to preprocess the acquired water level source data, and obtain valid water level data by using data interpolation and outlier correction; A signal decomposition module, connected to the data preprocessing module, for performing CEEMDAN effective signal decomposition on the preprocessed water level source data; The effective signal extraction module is connected to the signal decomposition module, and uses the correlation coefficient method R to determine the high-frequency and low-frequency signals, and calculates the IMF components of each CEEMDAN decomposition. When R reaches the first minimum point, the first several IMFs are eliminated, and the remaining IMF components are reconstructed into the denoised signal; A prediction module is connected to the effective signal extraction module, adopts a GRU prediction model, merges the input gate and the forget gate into an update gate, and uses the denoised signal as a feature value for training to predict water level changes; An early warning module is connected to the prediction module and uses the water level mean absolute error (WMAE), the water level root mean square error (WRMSE) and the water level determination coefficient (WR2) as water level early warning accuracy evaluation factors. When the evaluation factor result value exceeds the millimeter level, a water level early warning message is generated. The data storage module is connected to the early warning module and is used to store water level early warning related data into the ground database.

6. A multi-sensor water level automatic monitoring and early warning device, comprising the multi-sensor water level automatic monitoring and early warning device according to claim 5, characterized in that: Also included are: A power supply module is used to provide power support for the monitoring and early warning device and other components of the equipment; A communication module, connected to the monitoring and early warning device, is used to realize data transmission between the device and external equipment or systems, so as to send water level early warning information to the ground control department; The control module is connected to the monitoring and early warning device, the power module and the communication module respectively, and is used to coordinate the work of various components of the equipment and control the flow and processing flow of data.

7. A multi-sensor water level automatic monitoring and early warning storage medium, characterized in that: The storage includes: program codes for realizing a multi-sensor water level automatic monitoring and early warning method, including codes for realizing water level source data acquisition, data preprocessing, CEEMDAN effective signal decomposition, effective signal extraction, GRU prediction model training and water level early warning accuracy evaluation; Water level source data, preprocessed water level data, IMF component data after CEEMDAN decomposition, GRU prediction result data, and water level warning related data, among which the water level warning related data includes the calculation results of water level mean absolute error WMAE, water level root mean square error WRMSE, water level determination coefficient WR2, and warning information generated based on these results.

Images