Reverse reservoir water level prediction method based on osmotic pressure data fusion

Abstract

The invention discloses a reservoir water level reverse prediction method based on osmotic pressure data fusion, and the method comprises the steps: firstly, carrying out the preprocessing of osmotic pressure data and water level data, and carrying out the data alignment and missing value filling; then, calculating the correlation between the osmometers and the water level of the reservoir, and weighting different osmometer data according to the correlation to form fused osmometer data; and finally, training the gradient lifting regression model by using the fused data to predict the water level of the reservoir, and measuring the performance of the prediction model by taking a mean square error as an evaluation index. The reservoir level prediction method provides accurate reservoir level prediction, has strong adaptability and reliability, is especially suitable for reservoir safety management and disaster prevention and reduction, and provides a scientific basis for reservoir management.

Description

Technical Field

[0001] The present invention relates to reservoir water level prediction and data processing technologies, and particularly to a reservoir water level regression prediction method based on multi-piezometer data fusion. Background Art

[0002] As an important water conservancy infrastructure, reservoirs play a crucial role in flood control, irrigation, water resource allocation, and ecological protection. The safe and stable operation of reservoirs is directly related to social and economic development and the safety of people's lives and property. Therefore, accurately predicting reservoir water level changes is crucial for reservoir management. Traditionally, reservoir water level prediction mainly relies on meteorological data and hydrological models. However, due to the uncertainty of meteorological prediction and the complexity of hydrological models, the prediction results often have large errors.

[0003] In recent years, piezometers, as an effective tool for monitoring internal seepage and water level changes in reservoirs, have been gradually applied to reservoir water level prediction. Piezometers can accurately reflect the seepage state inside the reservoir dam, thus providing important preliminary data support for water level changes. However, due to the large number of piezometers, high data dimensions, and possible noise, how to effectively fuse the data of multiple piezometers for accurate prediction has become a key issue in current technologies.

[0004] In the prior art, most methods rely on the data of a single piezometer or simple weighted averaging of multi-point data. This way may ignore the spatial correlation and temporal characteristics between different piezometers, thereby affecting the accuracy of water level prediction. To address this issue, fusing the data of multiple piezometers and accurately predicting the water level through regression analysis methods have become a research hotspot. To achieve this goal, a series of key technical issues such as interpolation processing, anomaly detection, weighted fusion, and regression modeling of piezometer data need to be solved. Summary of the Invention

[0005] The purpose of the present invention is to provide a reservoir water level regression prediction method based on multi-piezometer data fusion, which can improve the accuracy and reliability of water level prediction through a reasonable weighting mechanism and regression analysis method.

[0006] The technical solution adopted by the present invention to solve its technical problems is:

[0007] A reverse prediction method for reservoir water level based on piezometric data fusion, comprising the following steps:

[0008] Step S1: Read the pressure data of multiple piezometers and the reservoir water level data, and align the data according to time and fill in the missing values;

[0009] Step S2: Calculate the correlation between each piezometer and the reservoir water level, and analyze the correlation degree between the pressure data and the water level data of different piezometers;

[0010] Step S3: Weight the data of each piezometer according to the calculated correlation to form weighted pressure data;

[0011] Step S4: Fuse the weighted pressure data to generate a single fused piezometric data as the input feature of the regression model;

[0012] Step S5: Use the gradient boosting regression model to train the model with the fused piezometric data and predict the reservoir water level;

[0013] Step S6: Evaluate the model performance, use the mean square error as the evaluation index, and optimize the model parameters;

[0014] Step S7: Apply the trained model to real-time data prediction to provide accurate reservoir water level warning information.

[0015] Further technology of the present invention:

[0016] Preferably, by using the Pearson correlation coefficient, calculate the correlation between each piezometer and the reservoir water level, and its value range is [-1, 1]. The calculation method is as follows:

[0017]

[0018] where x i and y i respectively represent the i-th value in the piezometric data and the water level data, and respectively represent the means of the piezometric data and the water level data, and n is the length of the data set.

[0019] Preferably, specifically in step S3, calculate the weight parameter of each piezometric data according to the correlation of the piezometric data and normalize it:

[0020]

[0021] where w i is the weight of the piezometric data i, r i is the correlation between the piezometric data i and the reservoir water level data. By square transformation (|r i | 2 ), the influence of the piezometer with high correlation is enhanced, and the sum of all weights is ensured to be 1 through the normalization operation.

[0022] Preferably, the weighted data fusion formula in step S4 is as follows:

[0023]

[0024] Among them, is the osmotic pressure data after fusion, and P i is the pressure data of the osmotic pressure gauge i.

[0025] Preferably, step S5 is specifically as follows:

[0026] S51. Divide the fused pressure data and the water level data W into a training set and a test set, where 80% is used for training and 20% is used for testing;

[0027] S52. Model training: The gradient boosting regression model updates the model by gradually fitting the residuals. The update formula for each round is:

[0028]

[0029] Among them, is the prediction result of the (m - 1)th round, is the model prediction result of the mth round;

[0030] S53. Model prediction: The prediction process is expressed as:

[0031]

[0032] Among them, is the predicted water level of the ith sample, is the final prediction result of the ith sample, obtained through the weighted sum of M trees, is the initial prediction value, which is set to the mean value of the water level W here. η = 0.1 is the learning rate, which controls the contribution of each tree to the final prediction result, and M = 100 is the total number of trees.

[0033] Advantages of the present invention:

[0034] Through efficient data preprocessing, feature extraction and weighted fusion, using the gradient boosting regression model for water level prediction, the accuracy and stability of reservoir water level prediction are significantly improved. By performing weighted fusion on the data of multiple osmotic pressure gauges, the reliability of the data is ensured, and the influence of the failure or deviation of a single osmotic pressure gauge on the prediction result is reduced. This method can not only accurately capture the complex relationship between the osmotic pressure gauge data and the change of the reservoir water level, but also has strong robustness and adaptability, can handle the dynamic changes in different environments, and ensure the stability and real-time nature of the prediction result. It is particularly suitable for fields such as reservoir water level monitoring, flood warning and disaster prevention and mitigation, provides a scientific decision-making basis for reservoir safety management, and improves the efficiency and safety of water resource scheduling and management. Description of the Drawings

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0036] Figure 1 It is a schematic flow chart of a reservoir water level reverse prediction method based on seepage pressure data fusion disclosed in an embodiment of the present invention. Specific embodiments

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will further elaborate on the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0038] Combined with Figure 1 As shown, the present invention provides a reservoir water level reverse prediction method based on seepage pressure data fusion. The specific embodiments include:

[0039] Step S1: Read the pressure data of multiple piezometers and the reservoir water level data and align the data according to time and fill in the missing values.

[0040] S11. Extract the pressure data of multiple piezometers related to the reservoir and the water level data of the reservoir from the data platform and ensure that the time ranges are consistent.

[0041] Data set = {pressure data 渗压计1 , pressure data 渗压计2 ..., water level data

[0042] Among them, the pressure data represents the measured values of each piezometer during this time period, and the water level data represents the reservoir water level at the corresponding time point.

[0043] S12. Align the timestamps of each data set to ensure that the pressure data of all piezometers and the reservoir water level data have corresponding records at the same time point;

[0044] Use the following interpolation formula to fill in the missing data:

[0045]

[0046] Here P 渗压计j (t k ) is the piezometer data filled after interpolation

[0047] S13. Merge the pressure data and water level data of all piezometers according to the time stamp to form a unified data table. Each row records the data of multiple piezometers and the water level data at a certain time point. The final data form is as follows:

[0048] <![CDATA[t1]]> <![CDATA[P 渗压计1 (t1)]]> … <![CDATA[P 渗压计m (t1)]]> <![CDATA[W(t1) <!-- 3 -->]]> <![CDATA[t2]]> <![CDATA[P 渗压计1 (t2)]]> … <![CDATA[P 渗压计m (t2)]]> <![CDATA[W(t2)]]> … … … … … <![CDATA[t n > <![CDATA[P 渗压计1 (t n )]]> … <![CDATA[P 渗压计m (t n )]]> <![CDATA[W(t n )]]>

[0049] Among them, m represents the number of reservoir piezometers, and t i represents the time stamp, and P 渗压计j (t i ) represents the pressure value of piezometer j at time t i , and W(t i ) represents the water level value of the reservoir at the current time.

[0050] Step S2: Calculate the correlation between each piezometer and the reservoir water level, and analyze the correlation degree between the pressure data and water level data of different piezometers.

[0051] S21. Calculate the linear correlation between the pressure data and water level data of each piezometer by using the Pearson correlation coefficient. The Pearson correlation coefficient can be used to measure the linear relationship between two variables, and its value range is [-1, 1]. The calculation method is as follows:

[0052]

[0053] Among them, x i and y i respectively represent the i-th value in the piezometric data and the water level data. and respectively represent the means of the piezometric data and the water level data, and n is the length of the data set.

[0054] Step S3: Weight the data of each piezometer according to the calculated correlation to form weighted pressure data. Calculate the weight parameter of each piezometric data and normalize it by using the correlation of the piezometric data:

[0055]

[0056] Among them, w i is the weight of piezometric data i, and r i is the correlation between piezometric data i and the reservoir water level data. The influence of piezometers with high correlation is enhanced by square transformation (|r i | 2 ), and the normalization operation is used to ensure that the sum of all weights is 1.

[0057] Step S4: Merge the weighted pressure data to generate a single weighted and fused piezometric data as the input feature of the regression model. The weighted data fusion formula is as follows:

[0058]

[0059] Among them, is the osmotic pressure data after fusion, P i is the pressure data of the osmotic pressure gauge i.

[0060] Step S5: Use the gradient boosting regression model to train the model with the weighted osmotic pressure data and predict the reservoir water level;

[0061] S51. Divide the fused pressure data and the water level data W into a training set and a test set, where 80% is used for training and 20% is used for testing.

[0062] S52. Model training: The gradient boosting regression model updates the model by gradually fitting the residuals. The update formula for each round is:

[0063]

[0064] Among them, is the prediction result of the (m - 1)-th round, is the model prediction result of the m-th round.

[0065] S53. Model prediction: The prediction process of the gradient boosting regression model used in this method is expressed as:

[0066]

[0067] Among them, is the predicted water level (model output) of the i-th sample, is the final prediction result of the i-th sample, obtained by the weighted sum of M trees, is the initial predicted value, which is set to the mean value of the water level W here. η = 0.1 is the learning rate, which controls the contribution of each tree to the final prediction result. M = 100 is the total number of trees. In this method, max_depth = 3 is set to control the maximum depth of the decision tree. A smaller depth will result in a simpler tree model, improve the generalization ability of the model, and avoid overfitting.

[0068] Step S6: Evaluate the model performance, use the mean squared error (MSE) as the evaluation index, and optimize the model parameters to improve the prediction accuracy. The mean squared error (MSE) measures the difference between the model prediction value and the true value. The smaller the value, the better the prediction effect of the model. Its calculation formula is:

[0069]

[0070] where y i is the true value (i.e., the water level data), and n is the number of samples.

[0071] Step S7: In this embodiment, we apply the trained model to the prediction of the reservoir water level for real-time data. By predicting the real-time data of Qianshanshan Reservoir and Jiudouchuan Reservoir in Xuancheng City over a period of half a year, we can accurately predict the changes in the reservoir water level and provide timely early warning information for the reservoir water level. After model training and optimization, the mean square error (MSE) on the test set reached 0.00059 and 0.00132 respectively. These results indicate that this method can accurately predict the changes in the reservoir water level and effectively reflect the performance of the model in practical applications. Through the water level prediction provided by this method, real-time monitoring and early warning of the reservoir water level can be achieved, further improving the ability of reservoir safety management and disaster prevention and mitigation.

[0072] The present invention discloses a reverse prediction method for reservoir water level based on piezometric data fusion. This method first reads the pressure data from multiple piezometers, performs time alignment and missing value filling on it with the reservoir water level data to ensure the temporal consistency and integrity of the data. After data processing, the present invention calculates the correlation between each piezometer and the reservoir water level data, obtains the weights of different piezometers, and weights the piezometer data according to the correlation coefficient to form the fused pressure data. Then, a gradient boosting regression model is used to train the fused pressure data, and the water level data is used as the target variable for water level prediction. This method uses the mean square error (MSE) as the evaluation index and further improves the prediction accuracy through model optimization to ensure the robustness and accuracy of the model.

[0073] The reservoir water level prediction method of the present invention integrates multiple steps such as data missing value processing, feature weighting, regression modeling, and model optimization, significantly improving the accuracy and stability of reservoir water level prediction, and is particularly suitable for fields such as reservoir safety monitoring, flood control and drainage. Through the accurate prediction of the reservoir water level, the present invention can provide a scientific basis for reservoir management and achieve real-time early warning of the reservoir water level, enhancing the ability of disaster prevention and mitigation.

[0074] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by this claim book, they should fall within the protection scope of the present invention.

Claims

1. A reverse prediction method for reservoir water level based on osmotic pressure data fusion, characterized in that, It includes the following steps: Step S1: Read the pressure data of multiple piezometers and the reservoir water level data, and align the data according to time and fill in the missing values; Step S2: Calculate the correlation between each piezometer and the reservoir water level, and analyze the correlation degree between the pressure data and the water level data of different piezometers; Step S3: Weight the data of each piezometer according to the calculated correlation to form the weighted pressure data; Step S4: Fuse the weighted pressure data to generate a single fused piezometric data as the input feature of the regression model; Step S5: Use the gradient boosting regression model, and use the fused piezometric data to train the model to predict the reservoir water level; Step S6: Evaluate the model performance, use the mean square error as the evaluation index, and optimize the model parameters; Step S7: Apply the trained model to real-time data prediction to provide accurate reservoir water level warning information.

2. The reverse prediction method of reservoir water level based on osmotic pressure data fusion according to claim 1, characterized in that By using the Pearson correlation coefficient, calculate the correlation between each piezometer and the reservoir water level, and its value range is [-1, 1]. The calculation method is as follows: where x i and y i represent the i-th values in the osmotic pressure data and the water level data respectively, and represent the means of the osmotic pressure data and the water level data respectively, and n is the length of the data set.

3. A reverse prediction method for reservoir water level based on osmotic pressure data fusion according to claim 1, characterized in that Specifically for Step S3, calculate the weight parameter of each piezometric data according to the correlation of the piezometric data and normalize it: Among them, w i is the weight of the seepage pressure data i, and r i is the correlation between the seepage pressure data i and the reservoir water level data. The influence of the piezometers with high correlation is enhanced through square transformation (|r i | 2 ), and the sum of all weights is ensured to be 1 through normalization operation.

4. A reverse prediction method for reservoir water level based on osmotic pressure data fusion according to claim 3, characterized in that The weighted data fusion formula for Step S4 is as follows: Among them, is the integrated osmotic pressure data, P i is the pressure data of the osmotic pressure gauge i.

5. A reverse prediction method for reservoir water level based on seepage pressure data fusion according to claim 4, characterized in that Specifically for Step S5: S51. Divide the fused pressure data and water level data W into a training set and a test set, with 80% for training and 20% for testing; S52. Model training: The gradient boosting regression model updates the model by gradually fitting the residuals. The update formula for each round is: Among them, is the prediction result of the (m - 1)-th round, is the model prediction result of the m-th round; S53. Model prediction: The prediction process is expressed as: wherein, is the predicted water level of the i-th sample, is the final prediction result of the i-th sample, obtained by the weighted sum of M trees, is the initial predicted value, which is set here to the mean value of the water level W, η = 0.1 is the learning rate, controlling the contribution of each tree to the final prediction result, and M = 100 is the total number of trees.

Images