CN-119691586-B - Dam osmotic pressure space-time early warning method and device

CN119691586BCN 119691586 BCN119691586 BCN 119691586BCN-119691586-B

Abstract

The invention provides a dam osmotic pressure space-time early warning method and device, wherein the method comprises the steps of collecting original osmotic pressure time sequence data of a plurality of monitoring points in the monitoring process of a dam; the method comprises the steps of preprocessing time sequence data, utilizing the processed time sequence data of a plurality of monitoring points to perform characterization learning of seepage pressure time sequence, establishing a multi-task prediction model, and establishing an early warning model on the basis of the multi-task prediction model. By utilizing the scheme of the invention, the relation among different seepage pressure monitoring data can be effectively processed, and the dam seepage pressure time-space early warning is realized.

Inventors

LI KE
CHENG ZHENGFEI
YU JIALIN
PU GUOQING
WU GUOHUA
YU HONGLING

Assignees

四川华电泸定水电有限公司
中国水利水电建设工程咨询有限公司
水电水利规划设计总院有限公司

Dates

Publication Date: 20260512
Application Date: 20241108

Claims (10)

1. The dam osmotic pressure space-time early warning method is characterized by comprising the following steps of: Collecting original osmotic pressure time series data of a plurality of monitoring points in the dam monitoring process; Preprocessing the time sequence data; Performing characterization learning of seepage pressure time series by using the processed time series data of the plurality of monitoring points, and establishing a multi-task prediction model; establishing an early warning model on the basis of the multitask prediction model; The method for performing characterization learning of the seepage pressure time sequence by using the processed time sequence data of the plurality of monitoring points comprises the following steps of: Dividing the time sequence data of the plurality of monitoring points by utilizing a sliding window to obtain a subsequence; extracting the characteristics of the subsequences and establishing labels corresponding to the subsequences to obtain a training sample set; Constructing a diffusion convolutional neural network by using the training sample set to obtain a multi-task prediction model; The building of the early warning model based on the multitasking prediction model comprises the following steps: determining a loss function, the loss function comprising an uncertainty trade-off loss function; the uncertainty trade-off loss function is specifically configured to: The prediction results for all M monitoring points are represented using the following equation: ; The model has M outputs corresponding to M learning tasks; Assuming that the input is X, the model output is Y is output in a model Is the mean value, Is a normal distribution of variance; The multitasking possibility is described as: Wherein, the The uncertainty of the Mth task is represented by the observation noise parameter of the model; Taking the maximum value of the log likelihood values of the model parameters and the observed noise parameters to obtain a minimum target, namely obtaining a multi-task uncertainty trade-off loss function, wherein the expression is as follows: Wherein, the Representing the loss of the Mth task, coefficient terms Is the relative weight of the Mth task, as Is used for the increase of (a), And, in addition, Preventing as regularizer Too much increase and train the network to predict log variance, s=log ; And establishing an early warning model based on the loss function.
2. The method of dam osmotic pressure space-time pre-warning according to claim 1, wherein the preprocessing the time-series data comprises: preprocessing the time series data by adopting linear interpolation to clean; the data after washing were normalized using Z-score normalization.
3. The dam osmotic pressure space-time early warning method according to claim 2, wherein different input features in the multi-task prediction model have different weights.
4. A dam osmotic pressure space-time pre-warning method according to claim 3, wherein the method further comprises: Weights for each input feature in the multitasking prediction model are determined based on an attention mechanism.
5. The dam osmotic pressure space-time early warning method according to claim 3, wherein the diffusion convolution neural network comprises a graph convolution and a recurrent neural network, the graph convolution is used for simulating the spatial relationship of monitoring points, and the recurrent neural network simulates the time dependence.
6. The method of dam osmotic pressure space-time pre-warning according to claim 5, further comprising: the graph convolution is achieved by constructing an adjacency matrix through pearson correlations between multiple monitoring sequences.
7. The method of dam osmotic pressure space-time pre-warning according to any one of claims 1 to 6, further comprising: Acquiring current osmotic pressure time series data of a plurality of monitoring points; And carrying out real-time monitoring and early warning on the dam according to the current osmotic pressure time sequence data and the early warning model.
8. A dam osmotic pressure space-time early warning device, characterized in that the device comprises: The data collection module is used for collecting original osmotic pressure time series data of a plurality of monitoring points in the dam monitoring process; The preprocessing module is used for preprocessing the time sequence data; The multi-task prediction model construction module is used for carrying out characterization learning of the seepage pressure time sequence by utilizing the processed time sequence data of the plurality of monitoring points and establishing a multi-task prediction model; The early warning model building module is used for building an early warning model on the basis of the multitask prediction model; The method for performing characterization learning of the seepage pressure time sequence by using the processed time sequence data of the plurality of monitoring points comprises the following steps of: Dividing the time sequence data of the plurality of monitoring points by utilizing a sliding window to obtain a subsequence; extracting the characteristics of the subsequences and establishing labels corresponding to the subsequences to obtain a training sample set; Constructing a diffusion convolutional neural network by using the training sample set to obtain a multi-task prediction model; The building of the early warning model based on the multitasking prediction model comprises the following steps: determining a loss function, the loss function comprising an uncertainty trade-off loss function; the uncertainty trade-off loss function is specifically configured to: The prediction results for all M monitoring points are represented using the following equation: ; The model has M outputs corresponding to M learning tasks; Assuming that the input is X, the model output is Y is output in a model Is the mean value of the two values, Is a normal distribution of variance; The multitasking possibility is described as: Wherein, the The uncertainty of the Mth task is represented by the observation noise parameter of the model; Taking the maximum value of the log likelihood values of the model parameters and the observed noise parameters to obtain a minimum target, namely obtaining a multi-task uncertainty trade-off loss function, wherein the expression is as follows: Wherein, the Representing the loss of the Mth task, coefficient terms Is the relative weight of the Mth task, as Is used for the increase of (a), And, in addition, Preventing as regularizer Too much increase and train the network to predict log variance, s=log ; And establishing an early warning model based on the loss function.
9. The dam osmotic pressure space-time pre-warning device according to claim 8, wherein the device further comprises: And the early warning module is used for acquiring the current osmotic pressure time series data of the plurality of monitoring points and carrying out real-time monitoring and early warning on the dam according to the current osmotic pressure time series data and the early warning model.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor performs the steps of the dam osmotic pressure spatiotemporal warning method of any of claims 1 to 7.

Description

Dam osmotic pressure space-time early warning method and device Technical Field The invention relates to the field of dam seepage safety monitoring in water conservancy and hydropower engineering, in particular to a dam seepage pressure space-time early warning method and device. Background The seepage safety of the dam directly relates to the stability of the structure of the dam, and in order to ensure the safety of the dam, sensor equipment is usually installed in the dam, so that the seepage condition is monitored in real time. With the increase of monitoring range and equipment number, efficient and accurate seepage pressure prediction becomes particularly important. However, how to effectively process the relation between different seepage pressure monitoring data and establish a multi-task seepage pressure prediction model with high efficiency and accuracy and multiple monitoring points is a problem to be solved urgently. In addition, in order for the field operation and maintenance personnel to clearly know the state of the current point location, whether the predicted result exceeds the expected value needs to be evaluated. Once the predicted value exceeds the expected range, early warning needs to be sent out in time, so that the accuracy and the instantaneity of the earth-rock dam seepage early warning are improved, and the safe and stable operation of the earth-rock dam is further ensured. At present, research on the prediction of the time series of the seepage pressure mainly focuses on the prediction of a single measuring point. However, dam osmotic pressure monitoring involves multiple points and requires a large number of sequences for predictive analysis. If a prediction model is built for each measuring point separately, the calculation, storage and time cost will be increased significantly, and it is difficult to meet the requirements of actual engineering. Disclosure of Invention The invention provides a dam osmotic pressure space-time early warning method and device, which are used for effectively processing the relation between different osmotic pressure monitoring data, establishing a multi-task osmotic pressure prediction model with high efficiency and accuracy and multiple monitoring points, and realizing dam osmotic pressure space-time early warning. Therefore, the invention provides the following technical scheme: in one aspect, the invention provides a dam osmotic pressure space-time early warning method, which comprises the following steps: Collecting original osmotic pressure time series data of a plurality of monitoring points in the dam monitoring process; Preprocessing the time sequence data; Performing characterization learning of seepage pressure time series by using the processed time series data of the plurality of monitoring points, and establishing a multi-task prediction model; and establishing an early warning model based on the multitask prediction model. Optionally, the preprocessing the time series data includes: preprocessing the time series data by adopting linear interpolation to clean; the data after washing were normalized using Z-score normalization. Optionally, the performing characterization learning of the seepage pressure time sequence by using the processed time sequence data of the plurality of monitoring points, and establishing the multi-task prediction model includes: Dividing the time sequence data of the plurality of monitoring points by utilizing a sliding window to obtain a subsequence; extracting the characteristics of the subsequences and establishing labels corresponding to the subsequences to obtain a training sample set; And constructing a diffusion convolutional neural network by using the training sample set to obtain a multi-task prediction model. Optionally, different input features in the multi-task prediction model have different weights. Optionally, the method further comprises determining a weight for each input feature in the multitasking prediction model based on an attention mechanism. Optionally, the diffusion convolutional neural network comprises a graph convolution and a recurrent neural network, wherein the graph convolution is used for simulating the spatial relation of monitoring points, and the recurrent neural network simulates the time dependence. Optionally, the method further comprises constructing an adjacency matrix by pearson correlation between the plurality of monitored sequences, effecting graph convolution. Optionally, the building an early warning model based on the multitasking prediction model includes: Determining a loss function, the loss function comprising an uncertainty trade-off loss function and/or a quantile regression loss function; And establishing an early warning model based on the loss function. Optionally, the method further comprises: Acquiring current osmotic pressure time series data of a plurality of monitoring points; And carrying out real-time monitoring and early warning on the dam accor