A method for formulating dynamic dam safety monitoring indicators considering multiple influencing factors

Abstract

The present invention relates to a method for formulating dynamic dam safety monitoring indicators that considers multiple influencing factors, comprising: selecting a dam monitoring data sequence; constructing a prediction model for effect size; calculating the residuals between the fitted data sequence obtained by the prediction model and the original data sequence to obtain the residual standard deviation σ; determining, based on the calculation results of the prediction model, the minimum standard deviation multiple n that can envelop all credible monitoring data to obtain nσ; determining the multiple m of the nominal parameter and the standard deviation σ; comparing mσ, nσ, and 3σ, taking the maximum value as the endpoint of the normal value interval, and determining the dam safety monitoring indicator. The present invention fully considers the influence of multiple factors such as the probability distribution of observation values, the accuracy and operational stability of monitoring instruments, and the rich experience of dam safety engineers to determine reasonable dam safety monitoring indicators, which have strong guiding significance for monitoring and forecasting warning or dangerous states of dam operation.

Description

Technical Field

[0001] The present invention belongs to the field of dam safety monitoring, and in particular relates to a method for formulating dynamic dam safety monitoring indicators taking multiple influencing factors into consideration. Background Art

[0002] The devastating consequences of numerous dam failures have led people to realize that for special structures like dams, their operational and safety status must be accurately and quickly determined so that measures can be taken to ensure their normal operation. One effective method is to develop quantitative indicators based on qualitative analysis and existing monitoring data to monitor and forecast dams in alert or dangerous operating conditions.

[0003] For concrete dams, on the one hand, safety monitoring data is used to build models and derive monitoring indicators. On the other hand, safety monitoring indicators are used to analyze the measured data to determine the dam's operating status. The results of this analysis will influence subsequent decision-making. Therefore, a reasonable method for formulating dam safety monitoring indicators is a key factor in achieving dam safety monitoring through dam safety monitoring.

[0004] Traditionally, dam safety monitoring indicators are developed based on monitoring data modeling and the Raida criterion. This involves first modeling credible historical data, then determining the standard deviation of the residual between the original monitoring data and the fitted series. The predicted value, plus or minus three standard deviations, is then used as the warning interval. This method is based on the consideration of various instrumental errors and assumes that multiple observations under the same conditions follow a normal distribution. Using a threshold of plus or minus three standard deviations, based on the normal distribution, guarantees a 99.7% probability that the data will fall within the interval, thus ensuring that it can be used as a monitoring indicator.

[0005] However, there are many problems in the actual work when using the above indicator formulation method, such as:

[0006] (1) There is a high probability of false positives for data with a high observation frequency, with a probability of about 0.3%;

[0007] (2) When the modeling effect is good, the standard deviation is very small, resulting in a too small normal value interval;

[0008] (3) The standard deviations obtained by different analytical methods are inconsistent.

[0009] Therefore, in order to effectively avoid the shortcomings of traditional methods, the present invention studies a method for formulating dynamic dam safety monitoring indicators that takes multiple influencing factors into consideration. Summary of the Invention

[0010] The purpose of the present invention is to address the above-mentioned problems and provide a method for formulating dynamic dam safety monitoring indicators that takes into account multiple influencing factors. It comprehensively considers the influence of multiple factors such as the probability distribution of observation values, the accuracy and operating stability of monitoring instruments, and the rich experience of dam safety engineers to determine a reasonable range of dam safety monitoring indicators, thereby avoiding the problem of a small normal value range caused by a very small standard deviation when the modeling effect is good in traditional methods.

[0011] The technical solution of the present invention is a method for formulating dynamic dam safety monitoring indicators taking into account multiple influencing factors, comprising the following steps:

[0012] Step 1: Determine the dam safety monitoring points for which indicators need to be formulated and select the dam monitoring data series;

[0013] Step 2: Based on the reliable measurement values ​​and environmental quantity data series, a prediction model of effect size is constructed;

[0014] Step 3: Calculate the residuals of the fitted data sequence obtained by the prediction model and the original data sequence, and calculate the standard deviation σ of the residuals;

[0015] Step 4: Calculate 3σ based on the Laida criterion;

[0016] Step 5: Verify whether the data in the dam monitoring data series is abnormal. Based on the calculation results of the prediction model, determine the minimum standard deviation multiple that can envelop all credible monitoring data. n ;

[0017] Step 6: Minimum standard deviation multiple obtained in step 5 n and standard deviation σ, we can calculate n σ;

[0018] Step 7: Analyze the characteristics of the measuring instruments for dam monitoring data, use nominal parameters to describe their accuracy and operational stability, and relate the nominal parameters to the standard deviation σ, calculate the ratio of the nominal parameters to the standard deviation σ, and determine the multiple of the nominal parameters to the standard deviation σ m ;

[0019] Step 8: Compare m σ、 n The size of σ and 3σ is X = max(m, n, 3), where X is the characteristic multiple of the final residual standard deviation, and max() represents the maximum value function;

[0020] Step 9: Determine the dam safety monitoring index using the interval (E-Xσ, E+Xσ) as the normal value interval, where E represents the mean of the monitoring data series;

[0021] Step 10: Update dam safety monitoring indicators;

[0022] Step 10.1: If the dam operating performance changes or reaches a specified interval, re-execute steps 1-9 to update the prediction model and the normal value range of the indicator;

[0023] Step 10.2: After the dam has been in operation for a period of time, the value of n in step 5 and the value of m in step 7 are updated based on the latest monitoring data and the stability of the monitoring instrument operation.

[0024] Furthermore, the effect quantity is a deformation value, a seepage amount, or a stress value.

[0025] Preferably, step 2 specifically includes the following sub-steps:

[0026] Step 2.1: Obtain historical monitoring data vectors / time-value data pairs for a certain period of time, including reservoir water level, rainfall, effect size data, and temperature data. The temperature data is dam body temperature, boundary temperature, or air temperature.

[0027] Step 2.2: If the temperature data is boundary temperature or dam body temperature data, data processing is performed. First, the data is spatially interpolated into a fixed-size matrix and feature extraction is performed using a BP neural network to obtain a temperature feature vector. If the data is air temperature, the air temperature value is directly used as the feature value.

[0028] Step 2.3: Use the water level, temperature, and rainfall data at each moment as the independent variable vector as the input of the LSTM network, and the effect size vector as the final output to train and obtain the LSTM prediction model;

[0029] Step 2.4: Call the LSTM prediction model, construct the feature vector of the predicted moment as input, and realize the prediction of the monitoring effect size at the corresponding moment.

[0030] Preferably, the BP neural network comprises an input layer, a hidden layer, and an output layer, and the neural layers are connected to each other in a fully connected manner.

[0031] Preferably, the hidden state of the LSTM network unit is expressed as:

[0032]

[0033] In the formula zt i Indicates the i The hidden state of the moment, α 、 β are all constants, f 1. f 0 is the activation function, t i 、H i 、T i Respectively represent iTime, water level, and temperature at each moment.

[0034] Preferably, the time interval between updates in step 10.1 and step 10.2 is 1 year.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] (1) The present invention provides a method for formulating dynamic dam safety monitoring indicators that takes into account multiple influencing factors. It fully considers the influence of multiple factors such as the probability distribution of observation values, the accuracy and operation stability of monitoring instruments, and the rich experience of dam safety engineers, and determines reasonable dam safety monitoring indicators. It has a strong guiding significance for the monitoring and forecasting of warning or dangerous states of dam operation.

[0037] (2) According to the principle of taking the maximum value, the present invention selects the maximum value of the three methods of data modeling, abnormality verification and instrument analysis as the monitoring indicator, which can effectively avoid the problem of the normal value interval being too small due to the small standard deviation when the modeling effect is good in the traditional method.

[0038] (3) The method of the present invention can effectively avoid false alarms of monitoring indicators by setting scientific and reasonable normal value intervals of monitoring indicators.

[0039] (4) The method of the present invention dynamically updates the characteristic parameters regularly, so that the monitoring indicators can follow and adapt to the dynamic changes of the measuring points during the dam operation period and always remain within a reasonable range. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention will be further described below with reference to the accompanying drawings and examples.

[0041] Figure 1 Schematic diagram of the flow of the method for formulating dam safety monitoring indicators according to an embodiment of the present invention.

[0042] Figure 2 Schematic diagram of a prediction model according to an embodiment of the present invention.

[0043] Figure 3 This is a graph showing deformation data of the dam crest measuring point according to an embodiment of the present invention.

[0044] Figure 4 3σ monitoring interval curve diagram of the embodiment of the present invention.

[0045] Figure 5 For the embodiment of the present invention n σ monitoring interval graph.

[0046] Figure 6 For the embodiment of the present invention m σ curve diagram of the monitoring interval. DETAILED DESCRIPTION

[0047] like Figure 1 As shown in the figure, the method for formulating dynamic dam safety monitoring indicators considering multiple influencing factors includes the following steps:

[0048] (1) Selection of monitoring data series: Select the process line of the A measuring point of the dam crest deformation of a certain gravity dam, manually remove invalid values ​​and identify whether the measured values ​​are normal.

[0049] (2) Data modeling: Constructing LSTM prediction model, such as Figure 2 shown.

[0050] The hidden state of the LSTM network unit is:

[0051]

[0052] In the formula zt i Indicates the i The hidden state of the moment, α 、 β are all constants, f 1. f 0 is the activation function, t i 、H i 、T i Respectively represent i Time, water level, and temperature at each moment.

[0053] Figure 2 The BP neural network shown in the figure includes an input layer, a hidden layer, and an output layer, and the neural layers are connected to each other using a fully connected method. The BP neural network performs feature extraction to obtain a temperature feature vector, which serves as the input of the LSTM network unit.

[0054] After manually eliminating invalid values ​​and identifying whether the marked measured values ​​are normal, the deformation process line of a gravity dam crest and the fitting data obtained by modeling in the embodiment are as follows: Figure 3 As shown, it can be seen that the modeling effect is good.

[0055] (3) Calculation of residual standard deviation σ: The value of the residual standard deviation obtained through modeling calculation is 0.248.

[0056] (4) According to the Laida criterion, calculate

[0057]

[0058] (5) Abnormal verification: Experienced dam safety experts verify whether the data in the sequence is abnormal and mark it. Based on the model analysis results, the minimum standard deviation multiple that can be included in all credible monitoring data is calculated. n , calculated n It is 4.8.

[0059] (6) Feedback analysis: calculated

[0060]

[0061] (7) Instrument analysis: By analyzing the type, accuracy, working environment, reading method, etc. of the measuring instrument at measuring point A, a nominal parameter is used to describe its accuracy and operating stability, and the nominal parameter is associated with the residual standard deviation to obtain its multiple of the residual standard deviation. m , calculated m is 4.1.

[0062] Eigenvalue analysis: calculated

[0063]

[0064] (8) Development of dam safety monitoring indicators: By calculating monitoring indicators through data modeling, anomaly verification and instrument analysis, we can obtain:

[0065] Data modeling: The monitoring interval is set with the traditional ±0.744 or ±3σ, such as Figure 4 As shown in the figure, when monitoring in this interval, it is found that 5 measured values ​​fall outside the monitoring area. In fact, these 5 measured values ​​are normal values, which will cause false alarms.

[0066] Abnormal verification: Considering the impact of reliable historical measurements on monitoring indicators, the envelope obtained by reverse calculation is ±1.190, that is, ±4.8σ interval, such as Figure 5 As shown, the monitoring threshold value formed in this way can include all normal measured values.

[0067] Instrument analysis: Considering factors such as instrument accuracy and operational stability, the monitoring index is obtained as ±1.017, that is, ±4.1σ interval. Figure 6 As shown, it can be seen that there are still 2 normal measurement values ​​outside the monitoring area.

[0068] (9) By taking the maximum value X=max(3, n , m ) principle, the interval of ±1.190 or ±4.8σ was selected as the monitoring index of measuring point A, which avoided false reporting of normal measured values.

[0069] The embodiment fully considers the influence of multiple factors such as the probability distribution of observation values, the accuracy and operational stability of monitoring instruments, and the rich experience of dam safety engineers, and determines a reasonable dam safety monitoring indicator. This effectively avoids the problem of a small normal value interval caused by a very small standard deviation when the modeling effect is good in traditional methods, and largely avoids false alarms of monitoring indicators. It has guiding value for monitoring and forecasting of warning or dangerous states of dam operation, and has promotion and use value in the hydropower industry.

Claims

1. A method for formulating dynamic dam safety monitoring indicators considering multiple influencing factors, characterized by: The following steps are involved: Step 1: Determine the dam safety monitoring points for which indicators need to be formulated and select the dam monitoring data series; Step 2: Based on the reliable measurement values ​​and environmental quantity data series, a prediction model of effect size is constructed; Step 3: Calculate the residuals of the fitted data sequence obtained by the prediction model and the original data sequence, and calculate the standard deviation σ of the residuals; Step 4: Calculate 3σ based on the Laida criterion; Step 5: Verify whether the data in the dam monitoring data series is abnormal. Based on the calculation results of the prediction model, determine the minimum standard deviation multiple that can envelop all credible monitoring data. n ; Step 6: Minimum standard deviation multiple obtained in step 5 n and standard deviation σ, we can calculate n σ; Step 7: Analyze the characteristics of the measuring instruments for dam monitoring data, use nominal parameters to describe their accuracy and operational stability, and relate the nominal parameters to the standard deviation σ, calculate the ratio of the nominal parameters to the standard deviation σ, and determine the multiple of the nominal parameters to the standard deviation σ m ; Step 8: Compare m σ、 n The size of σ and 3σ, take X = max( m , n , 3), where X is the characteristic multiple of the final residual standard deviation, and max( ) represents the maximum value function; Step 9: Determine the dam safety monitoring index using the interval (E-Xσ, E+Xσ) as the normal value interval, where E represents the mean of the monitoring data sequence.

2. The method for formulating dynamic dam safety monitoring indicators according to claim 1, characterized in that: The effect quantity is a deformation value, a seepage amount, or a stress value.

3. The method for formulating dynamic dam safety monitoring indicators according to claim 2, characterized in that: Step 2 specifically includes the following sub-steps: Step 2.1: Obtain historical monitoring data vectors / time-value data pairs for a certain period of time, including reservoir water level, rainfall, effect size data, and temperature data. The temperature data is dam body temperature, boundary temperature, or air temperature. Step 2.2: If the temperature data is boundary temperature or dam body temperature data, data processing is performed. First, the data is spatially interpolated into a matrix and feature extraction is performed using a BP neural network to obtain a temperature feature vector. If the data is air temperature, the air temperature value is directly used as the feature value. Step 2.3: Use the water level, temperature, and rainfall data at each moment as the independent variable vector as the input of the LSTM network, and the effect size vector as the final output to train and obtain the LSTM prediction model; Step 2.4: Call the LSTM prediction model, construct the feature vector of the predicted moment as input, and realize the prediction of the monitoring effect size at the corresponding moment.

4. The method for formulating dynamic dam safety monitoring indicators according to claim 3 is characterized in that: The BP neural network comprises an input layer, a hidden layer, and an output layer, and the neural layers are connected to each other in a fully connected manner.

5. The method for formulating dynamic dam safety monitoring indicators according to claim 3 is characterized in that: The LSTM network includes n LSTM network units, the hidden state expression of the LSTM network unit is: ； In the formula zt i Indicates the i The hidden state of the moment, α 1. α 2. β are all constants, f 1. f 0 is the activation function, t i 、H i 、T i Respectively represent i Time, water level, and temperature at each moment.

6. The method for formulating dynamic dam safety monitoring indicators according to claim 3, 4 or 5, characterized in that: The indicator formulation method also includes updating the dam safety monitoring indicators: if the dam operating status changes or reaches a specified interval, steps 1-9 are re-executed to update the prediction model and the indicator normal value range.

7. The method for formulating dynamic dam safety monitoring indicators according to claim 3, 4 or 5, characterized in that: The indicator formulation method is to formulate the indicators in step 5 after the dam has been in operation for a period of time based on the latest monitoring data and the stability of the monitoring instrument operation. n The value of and step 7 m The value of is updated.

Images