Freezing circle watershed hydrological prediction method and system based on multi-source data fusion

Abstract

The invention provides a freezing circle watershed hydrological prediction method and system based on multi-source data fusion, and relates to the technical field of hydrological monitoring and sediment monitoring, and the method comprises the steps: obtaining a vibration signal of the interaction of a water flow and a riverbed and meteorological data in a region, and carrying out the preprocessing; the preprocessed meteorological data and vibration signals are input into a multi-source data fusion model, sequential feature extraction is conducted on the meteorological data and the vibration signals through a multi-layer perceptron in the multi-source data fusion model, the extracted sequential features are fused through a gating circulation unit, and multi-source data fusion features are obtained; and carrying out empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signal to obtain a plurality of signal characteristics such as instantaneous frequency and instantaneous amplitude of the vibration signal, inputting the signal characteristics and the multi-source data fusion characteristics into a hydrological parameter prediction model, and outputting corresponding predicted values of the water flow velocity, the flow and the sediment transport capacity.

Description

Technical Field

[0001] The present disclosure relates to the technical field of hydrological monitoring and sediment monitoring, and particularly to a method and system for hydrological prediction in cryosphere basins based on multi-source data fusion. Background Technique

[0002] The statements in this part only provide background technical information related to the present disclosure and do not necessarily constitute prior art.

[0003] The cryosphere is an important part of the Earth system, including glaciers, snow cover, permafrost, etc. The hydrological processes in cryosphere basins have a profound impact on the global water cycle, sea level change, and ecosystem balance. With the intensification of global climate change, the cryosphere is undergoing significant changes, such as glacier retreat, permafrost thaw, etc. These changes directly affect the water quantity, water quality, and stability of the aquatic ecosystem within the basin. Accurately monitoring the hydrological and sediment changes in cryosphere basins is crucial for understanding the impact of climate change on water resources, formulating reasonable water resource management strategies, and predicting and mitigating related natural disasters (such as floods, debris flows, etc.).

[0004] The application of traditional hydrological monitoring methods in cryosphere basins faces many challenges. Among them, contact measurement methods, such as current meters, sediment meters, etc., need to directly contact the water body. In the extreme environment of the cryosphere, the instrument equipment is easily affected by low temperature, freezing, water flow impact, and sediment abrasion, resulting in equipment damage, increased measurement errors, and even inability to work properly. In addition, on the glacier surface, subglacial rivers, and some inaccessible areas (such as the hinterland of high-altitude glaciers, rivers in steep canyons, etc.), the implementation of traditional measurement methods is extremely difficult, and even on-site measurement cannot be carried out. These limitations make it difficult for traditional monitoring methods to obtain comprehensive and accurate hydrological and sediment data in cryosphere basins, and cannot meet the needs of research and management of water resources and ecological environment changes in this area. Summary of the Invention

[0005] To solve the above problems, the present disclosure proposes a method and system for hydrological prediction in cryosphere basins based on multi-source data fusion. Based on vibration signals and meteorological data, multi-source data fusion of vibration signals and meteorological data is performed, and the multi-source data is processed based on a neural network structure to extract characteristic frequencies and amplitude information related to water flow and sediment movement, capture the time-dependent relationship related to hydrological processes contained in the vibration signals, and achieve accurate prediction and monitoring of hydrological and sediment parameters.

[0006] According to some embodiments, the present disclosure adopts the following technical solutions:

[0007] A method for hydrological prediction in cryosphere basins based on multi-source data fusion, comprising:

[0008] Obtain the vibration signals of the interaction between the water flow and the river bed and the meteorological data in the region, and perform preprocessing;

[0009] Input the preprocessed meteorological data and vibration signals into the multi-source data fusion model. Through the multi-layer perceptron in the multi-source data fusion model, extract the time-series features of the meteorological data and vibration signals respectively, and then fuse the extracted time-series features through the gated recurrent unit to obtain the multi-source data fusion features;

[0010] Perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signals to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signals. Input the signal features and multi-source data fusion features into the hydrological parameter prediction model, and output the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

[0011] According to some embodiments, the present disclosure adopts the following technical solutions:

[0012] A cryosphere basin hydrological prediction system based on multi-source data fusion, including:

[0013] A data acquisition module, configured to obtain the vibration signals of the interaction between the water flow and the river bed and the meteorological data in the region, and perform preprocessing;

[0014] A data fusion module, configured to input the preprocessed meteorological data and vibration signals into the multi-source data fusion model. Through the multi-layer perceptron in the multi-source data fusion model, extract the time-series features of the meteorological data and vibration signals respectively, and then fuse the extracted time-series features through the gated recurrent unit to obtain the multi-source data fusion features;

[0015] A hydrological prediction module, configured to perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signals to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signals. Input the signal features and multi-source data fusion features into the hydrological parameter prediction model, and output the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

[0016] According to some embodiments, the present disclosure adopts the following technical solutions:

[0017] A computer program product, including a computer program, where when the computer program is executed by a processor, it implements the cryosphere basin hydrological prediction method based on multi-source data fusion.

[0018] According to some embodiments, the present disclosure adopts the following technical solutions:

[0019] A non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, it implements the cryosphere basin hydrological prediction method based on multi-source data fusion.

[0020] According to some embodiments, the present disclosure adopts the following technical solutions:

[0021] An electronic device includes: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory so that the electronic device executes the implemented hydrological prediction method for cryosphere basins based on multi-source data fusion.

[0022] Compared with the prior art, the beneficial effects of the present disclosure are:

[0023] For the hydrological prediction method for cryosphere basins based on multi-source data fusion of the present disclosure, according to different environmental positions and consideration factors, various sensor layout methods are constructed. In the glacial front area where glacial meltwater activities are frequent, there are many streams and small rivers with fast and variable water flows, a dense linear layout is adopted, and sensors are arranged in sequence along the direction of meltwater flow; in the main river channel area, in the river sections with obvious changes in width and large bends, a combined layout of curve type and encryption in key areas is adopted; in the area around the lake, a fan-shaped radiation layout is adopted at the water outlet to monitor the change of lake water outflow, lake shore erosion and sediment carrying conditions; based on the unique water flow, terrain and sediment conditions in each area, this can obtain hydrological and sediment information more accurately and comprehensively, providing strong support for the research and management of cryosphere basins.

[0024] For the hydrological prediction method for cryosphere basins based on multi-source data fusion of the present disclosure, in the multi-source data fusion model, a neural network model combining a multi-layer perceptron and a gated recurrent unit is constructed for multi-source data fusion. The multi-source data fusion improves the monitoring accuracy and prediction ability, helps to deeply understand the hydrological process and environmental changes; provides strong support for climate change research and water resource management, and also plays a key role in disaster warning and ecological protection, effectively reducing disaster risks and maintaining ecological balance.

[0025] For the hydrological prediction method for cryosphere basins based on multi-source data fusion of the present disclosure, the vibration signals generated by the interaction between the water flow movement and the riverbed or the shore are studied. By analyzing the characteristics of these vibration signals, parameters such as the water velocity, water flow, and sediment transport volume of the water body can be indirectly obtained. The vibration monitoring technology has the potential of high-precision and large-scale monitoring, providing a new idea and method for the hydrological monitoring of cryosphere basins, overcoming the limitations of traditional monitoring methods in complex environments such as cryosphere basins, realizing non-contact measurement, avoiding equipment damage, and expanding the monitoring range; through advanced technical means, hydrological and sediment data can be accurately obtained, providing a reliable basis for research. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings forming a part of this disclosure are used to provide a further understanding of the disclosure. The illustrative embodiments and descriptions thereof of the disclosure are used to explain the disclosure and do not constitute an improper limitation of the disclosure.

[0027] Figure 1 It is a schematic diagram of the overall step implementation process of the method of the embodiment of the present disclosure;

[0028] Figure 2 It is a schematic diagram of the system framework of the embodiment of the present disclosure. Detailed implementation manners

[0029] The present disclosure will be further described below in conjunction with the accompanying drawings and embodiments.

[0030] It should be noted that the following detailed descriptions are all illustrative and are intended to provide a further description of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present disclosure belongs.

[0031] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0032] Embodiment

[0033] In one embodiment of the present disclosure, a cryosphere basin hydrological prediction method based on multi-source data fusion is provided, including:

[0034] Step 1: Obtain the vibration signal of the interaction between water flow and riverbed and the meteorological data in the region, and perform preprocessing;

[0035] Step 2: Input the preprocessed meteorological data and vibration signal into the multi-source data fusion model. The multi-layer perceptron in the multi-source data fusion model is used to extract the time series features of the meteorological data and the vibration signal respectively, and then the gated recurrent unit is used to fuse the extracted time series features to obtain the multi-source data fusion features;

[0036] Step 3: Perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signal to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signal. Input the signal features and the multi-source data fusion features into the hydrological parameter prediction model, and output the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

[0037] As an embodiment, the cryosphere basin hydrological prediction method based on multi-source data fusion of the present disclosure, for the hydrological monitoring of the cryosphere basin, by studying the vibration signals generated by the interaction between the water flow and the riverbed or the shore, and analyzing the characteristics of these vibration signals, parameters such as the water velocity, flow rate, and sediment transport volume of the water body can be indirectly obtained. The specific implementation process of the method is as follows:

[0038] Step 1: Obtain the vibration signals of the interaction between the water flow and the riverbed and the meteorological data in the region, and perform preprocessing;

[0039] Specifically, (1) Collection of vibration signals. The vibration signals of the interaction between the water flow and the riverbed are collected by vibration sensors and uploaded. Among them, different layout methods are constructed for vibration sensors in different geological regions. The layout methods of vibration sensors are as follows:

[0040] 1) In the glacier front area, due to the frequent glacial meltwater activities in this area, forming numerous streams and small channels, and the fast and changing water flow, an intensive linear layout method is adopted here. Sensors are arranged in sequence along the direction of glacial meltwater flow, and the spacing can be set to 5 - 10 meters to capture the subtle vibration signals generated by the interaction between the water flow flowing out of the glacier and the riverbed and surrounding substances, and monitor the initial runoff process of glacial meltwater and its impact on the downstream.

[0041] 2) In the main river channel area, especially in the river sections with obvious changes in the width and large curvature of the river channel, the dynamic conditions of the water flow are complex and diverse. A combined layout method of curve type and dense layout in key areas is adopted. Sensors are arranged on the inner and outer sides of the river channel bend respectively. Because the water flow velocity on the inner side is relatively slow and there may be sediment deposition, and the water flow velocity on the outer side is fast and has a strong scouring effect on the river bank; sensors are also densely arranged at the parts where the river channel suddenly widens or narrows. For example, within the range of 50 meters upstream to 100 meters downstream of the narrowing section, a sensor is arranged every 15 - 20 meters to accurately monitor the changes in water flow velocity, water level fluctuations, and sediment transport.

[0042] 3) For the area around the lake, sensors are arranged in a fan-shaped pattern radiating outward at the lake outlet. The fan angle can be determined to be 60° - 120° according to the width of the outlet and the water flow diffusion range. The sensor spacing gradually expands from 5 - 8 meters at the outlet to 15 - 20 meters outward, which can effectively monitor the flow rate changes when the lake water flows out, the erosion effect of the water flow on the lake bank, and the situation of carrying sediment; near the lake inlet, due to the possible sediment deposition and water flow impact brought by the river, a circular layout method is adopted. Within a range with a radius of 10 - 15 meters around the inlet, a sensor is arranged every 8 - 10 meters to monitor the characteristics of the water flow entering the lake and the sediment input volume.

[0043] Furthermore, (2) Data transmission and storage of the collected vibration signals, including:

[0044] Data transmission is performed on the collected vibration signals. In areas with good 4G / 5G signal coverage, the device directly transmits the monitoring data to the remote control center through the 4G / 5G module. However, in areas with poor signal coverage, relay transmission or heterogeneous network fusion methods need to be adopted. Wireless relay nodes are set up. These relay nodes can be powered by solar energy (considering the characteristics of sufficient sunlight but inconvenient power supply in the cryosphere region). Through wireless ad hoc network technology, the scattered sensor data is aggregated to an area with stronger signal, and then transmitted through the 4G / 5G network. When using the 4G / 5G network for transmission, the MQTT protocol is adopted. The MQTT protocol is a lightweight message transmission protocol suitable for resource-constrained Internet of Things devices. It is based on the publish / subscribe model and efficiently transmits sensor data. The device acts as a publisher and publishes the monitoring data to a specific topic, and the remote control center acts as a subscriber to receive the data of the corresponding topic.

[0045] For extremely remote areas without 4G / 5G signals, such as the interior of high-altitude glaciers or inaccessible valleys, satellite communication becomes a viable option. Satellite communication has the advantage of global coverage and is not restricted by the geographical environment. For satellite communication, the DVB-S2 protocol is adopted. The DVB-S2 protocol is a digital video broadcast standard widely used in satellite communication, with high spectral efficiency and transmission reliability. Before transmitting the data, the monitoring data is encapsulated and encoded to meet the format requirements of the DVB-S2 protocol, and then transmitted to the remote control center through the satellite link.

[0046] Furthermore, the collected data is stored. Considering the huge amount of monitoring data and the need for long-term preservation, a distributed storage system is selected. The Ceph distributed storage system is adopted, which has the characteristics of high reliability, high scalability and high performance. Ceph realizes redundant backup of data by storing the data distributedly on multiple nodes, improving the security of the data. At the same time, Ceph supports dynamic expansion and can flexibly add storage nodes according to the growth of the data volume.

[0047] Among them, a dedicated storage server is set up in the remote control center, equipped with a large-capacity hard disk array for storing the data of the Ceph distributed storage system. The storage server adopts technologies such as redundant power supply and hot-swappable hard disks to improve the reliability of the system and prevent data loss caused by hardware failures.

[0048] To improve data read and write speeds, solid-state drives (SSDs) are used as a cache layer. Frequently accessed data is stored in the SSD cache, reducing the number of accesses to the hard disk array and thus improving data read speeds. Furthermore, a data tiered storage strategy is implemented, storing data on storage media with different performance levels based on access frequency and importance, further optimizing storage system performance.

[0049] Create a hierarchical directory structure to store monitoring data. Use the monitoring site as the root directory, then create directories for year, month, and day. Store daily monitoring data in the corresponding day-of-month directory. For a given monitoring site, the data storage path is " / monitoring site ID / year / month / date / data file."

[0050] (3) Synchronous collection and time alignment of meteorological data

[0051] Specifically, to achieve precise synchronization between meteorological data and vibration signal data during synchronous acquisition and time alignment, meteorological sensors are deployed and a timestamp-based synchronization algorithm is employed. Each meteorological and vibration sensor is equipped with a high-precision clock module that captures time information accurate to the millisecond level. During data acquisition, the acquisition timestamp is also recorded.

[0052] Assume that the meteorological data collection timestamp is T m , the vibration signal data acquisition time stamp is T s In order to align the two, a linear interpolation algorithm is used to compensate for the time deviation.

[0053] Assume that the collection interval of meteorological data is ΔT m , the sampling interval of the vibration signal data is ΔT s , and ΔT m ≠ΔT s For a certain time t, if T m ≤t<T m +ΔT m And T s ≤t<T s +ΔT s , then the meteorological data needs to be interpolated to align it with the vibration signal data in time. m The value at the moment is M(T m ), in T m +ΔT m The value at the moment is M(T m +ΔT m ), then the interpolated meteorological data M(t) at time t is,

[0054]

[0055] In the above way, the meteorological data is adjusted in the time dimension to be consistent with the vibration signal data, ensuring the accuracy of subsequent fusion.

[0056] (4) Preprocess the collected meteorological data and vibration signals to achieve data preprocessing and quality control, including:

[0057] Specifically, for the preprocessing of meteorological data, the box - plot method is used to identify and process outliers. First, calculate the quartiles Q1, Q3 of meteorological data (such as temperature, precipitation, etc.), and the inter - quartile range IQR = Q3 - Q1. According to the rules of the box - plot, data less than Q1 - 1.5IQR or greater than Q3 + 1.5IQR is regarded as an outlier. For the identified outliers, the average value of adjacent data is used for replacement. The preprocessed meteorological data contains the results of outlier processing and time alignment. The outlier processing uses the box - plot method, and the time alignment is achieved through the linear interpolation algorithm. The processed meteorological data is more accurate in time and value, providing reliable support for subsequent multi - source data fusion and hydrological parameter prediction.

[0058] For the preprocessing of vibration signal data, the wavelet threshold denoising algorithm is used to denoise the vibration signal. Select a suitable wavelet basis (such as sym8 wavelet) to perform wavelet decomposition on the vibration signal. The decomposition level is determined according to the frequency characteristics of the vibration signal (generally, the level that can effectively separate noise and signal is selected through experiments). Let the vibration signal be s(t), and after n - layer wavelet decomposition, wavelet coefficients wj,k are obtained (j represents the decomposition scale, and k represents the position of the wavelet coefficient). For soft - threshold denoising, the threshold λ can be estimated according to the noise level, such as using the universal threshold (where σ is the estimated value of the noise standard deviation, and N is the signal length). The wavelet coefficients after soft - threshold processing are:

[0059]

[0060] where, wj,k is the wavelet coefficient; λ is the soft - threshold, which can be estimated according to the noise level; is the wavelet coefficient after soft - threshold processing.

[0061] Then, the denoised vibration signal is obtained through wavelet reconstruction where, wavelet reconstruction is achieved through three steps: obtaining coefficients by existing wavelet decomposition, obtaining coefficients by wavelet decomposition, and performing wavelet reconstruction operations.

[0062] Step 2: Input the preprocessed meteorological data and vibration signals into the multi - source data fusion model. Through the multi - layer perceptron in the multi - source data fusion model, extract the time - series features of the meteorological data and vibration signals respectively, and then fuse the extracted time - series features through the gated recurrent unit to obtain the multi - source data fusion features;

[0063] Furthermore, the multi-source data fusion model is a neural network model combining a multi-layer perceptron and a gated recurrent unit. A neural network model combining a multi-layer perceptron (MLP) and a gated recurrent unit (GRU) is constructed for multi-source data fusion. The MLP is used to extract features and perform preliminary fusion on meteorological data and vibration signal data, and the GRU is used to learn the temporal features of the fused data.

[0064] Specifically, the number of neurons in the input layer of the MLP is determined according to the number of features of the meteorological data and the vibration signal data. The meteorological data includes m features such as temperature, precipitation, and air pressure, and the vibration signal data is processed through spectrum analysis and other methods to obtain n features. Then the number of neurons in the input layer is m + n. Two hidden layers are set. The number of neurons in the first hidden layer is h1, and the number of neurons in the second hidden layer is h2 (h1 and h2 can be adjusted and optimized through experiments). The activation function selects the ReLU (Rectified Linear Unit) function f(x) = max(0, x) to introduce non-linear factors and enhance the expression ability of the model. The number of neurons in the output layer is determined according to the number of hydrological parameters to be predicted. Here, it is assumed that the subsequent prediction of water velocity v, flow rate q, and sediment transport volume s, then the number of neurons in the output layer is 3. The MLP is used for feature extraction and preliminary fusion of meteorological data and seismic signal data, providing support for subsequent hydrological parameter prediction. Through a fixed structure and processing method, the input data is processed, and finally the preliminarily fused features are output for subsequent analysis.

[0065] The input of the GRU layer is the output of the MLP. The number of GRU units is determined according to the length and complexity of the time series of the data, and is set to g. The update gate z t , reset gate r t and candidate hidden state are calculated as follows:

[0066] z t = σ(W z ·[h t-1 , x t +b z )

[0067] r t = σ(W r ·[h t-1 , x t +b r )

[0068]

[0069] where W z , Wr , where \(W\) is the weight matrix and \(b\) z , \(b\) r , \(b\) is the bias vector, \(\sigma\) is the sigmoid function, and \(\odot\) represents element-wise multiplication. The hidden state \(h\) of the GRU t The update formula is: The GRU layer effectively captures the temporal information in the data, which is of great significance for learning the laws of water flow and sediment changes over time in hydrological monitoring.

[0070] Furthermore, the training process of the multi-source data fusion model based on the combination of the multi-layer perceptron and the gated recurrent unit is as follows:

[0071] (1) Use the preprocessed meteorological data and vibration signal data as the input of the multi-source data fusion model to construct a training dataset. Divide the dataset into a training set and a validation set according to the ratio of 70%:30%.

[0072] (2) During the model training process, use the Adam (Adaptive Moment Estimation) optimization algorithm to adjust the model parameters. The Adam optimization algorithm combines the advantages of the Adagrad and RMSProp algorithms and adaptively adjusts the learning rate. Its calculation steps are as follows:

[0073] Initialize the parameter vector \(\theta\), the first moment estimation vector \(m\), and the second moment estimation vector \(v\), the learning rate \(\alpha\) (usually taken as 0.001), the exponential decay rate \(\beta_1 = 0.9\), \(\beta_2 = 0.999\), and the small constant \(\epsilon = 10\) -8 ;

[0074] For each training sample \((x\) i , \(y\) i )(where \(x\) i is the input data and \(y\) i is the corresponding true label).

[0075] Calculate the gradient (where \(J\) is the loss function).

[0076] Update the first moment estimation: \(m \leftarrow \beta_1m+(1 - \beta_1)g\) i .

[0077] Update the second moment estimation:

[0078] Correct the bias of the first moment estimation (where \(t\) is the current iteration number)

[0079] Correct the bias of the second moment estimation:

[0080] Update the parameters:

[0081] Meanwhile, to prevent overfitting, the L2 regularization method is adopted, and a regularization term is added to the loss function (where λ is the regularization coefficient, usually taking a relatively small value, such as 0.01), and at this time the loss function becomes Optimize the size of the model parameters to avoid overfitting caused by an overly complex model, and finally obtain a multi-source data fusion model after training and optimization.

[0082] Step 3: Perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signal to obtain multiple signal characteristics such as the instantaneous frequency and instantaneous amplitude of the vibration signal;

[0083] Specifically, for the preprocessed seismic signal perform empirical mode decomposition to decompose it into a finite number of intrinsic mode functions (IMFs) c i (t) and a residue function rn(t), satisfying:

[0084]

[0085] Among them, the specific decomposition process is as follows: <(

[0086] 1) First, find all local extreme points (maximum points and minimum points) of the signal x(t);

[0087] 2) Fit the upper envelope e max (t) and the lower envelope e min (t) through cubic spline interpolation functions respectively, and calculate the average value of the upper and lower envelopes

[0088] 3) Calculate the difference h(t) = x(t) - m(t), and judge whether h(t) meets the conditions of the intrinsic mode function, that is, the difference between the number of extreme points and the number of zero-crossing points of h(t) does not exceed 1, and at any time t, the local mean value of h(t) is 0. If not, take h(t) as the new x(t) and repeat the above steps until the conditions are met to obtain the first intrinsic mode function c1(t).

[0089] 4) Then calculate the residue signal r1(t) = x(t) - c1(t), take r1(t) as the new x(t) and repeat the above decomposition process to obtain c2(t), c3(t),... c n (t), until the residue signal r n (t) becomes a monotonic function or the number of its extreme points is less than 3.

[0090] Furthermore, perform Hilbert transform on each obtained intrinsic mode function c i (t) to obtain its analytic signal,

[0091] zi(t) = ci(t) + jH[ci(t)]

[0092] where H[c i (t)] is the Hilbert transform of c i (t), and its calculation formula is The instantaneous frequency of the analytic signal z i (t) is

[0093]

[0094] where Im[z i (t)] and Re[z i (t)] are the imaginary and real parts of zi(t), respectively.

[0095] Finally, construct the Hilbert spectrum

[0096] where is the instantaneous amplitude of c i (t), and δ[w - w i (t)] is the Dirac function. The Hilbert spectrum can clearly show the energy distribution of the signal in the time - frequency plane, thereby extracting the characteristic frequencies and amplitude information related to water flow and sediment movement.

[0097] Step 4: Input the signal features and the multi - source data fusion features into the hydrological parameter prediction model, and output the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

[0098] Specifically, establish a hydrological parameter prediction model based on the long short - term memory neural network (LSTM). The LSTM network is particularly suitable for processing data with time - series characteristics and can effectively capture the time - dependent relationships related to hydrological processes contained in seismic signals.

[0099] The LSTM network consists of multiple LSTM units, and each LSTM unit contains an input gate, a forget gate, an output gate, and a memory unit. Its calculation formulas are as follows:

[0100] Forget gate: f t = σ(W f ·[h t-1 , x t + b f )

[0101] Input gate: i t = σ(W i ·[h t-1 , x t + b i )

[0102] Candidate memory unit:

[0103] Output gate: o t = σ(W o · [h t-1 , x t + b o )

[0104] Hidden state: h t = o t Θ tanh(c t )

[0105] Where W f , W i , W c , W o , are weight matrices, b f , b i , b c , b o , are bias vectors, σ is the sigmoid function, and Θ represents element-wise multiplication. The input x(t) is the vibration signal feature processed by the Hilbert-Huang transform, and ht-1 is the hidden state at the previous moment.

[0106] Take the signal features (instantaneous frequency, instantaneous amplitude, etc.) obtained from Hilbert spectrum analysis and the multi-source data fusion features as the input of the model, and the predicted values of the corresponding water body flow velocity, flow rate, and sediment transport volume as the output.

[0107] Furthermore, the training process of the hydrological parameter prediction model based on the long short-term memory neural network (LSTM) is as follows:

[0108] In model training and optimization, take the signal features (such as instantaneous frequency, instantaneous amplitude, etc.) obtained from Hilbert spectrum analysis and the features after multi-source data fusion as the input of the model, and the measured values of the corresponding water body flow velocity, flow rate, and sediment transport volume as the output to construct a training dataset. Divide the dataset into a training set and a test set according to the ratio of 80%:20%.

[0109] Use the stochastic gradient descent (SGD) algorithm combined with the backpropagation through time (BPTT) algorithm to train the LSTM model. The loss function is selected as the mean squared error (MSE) function: Where y i is the actual measured value (flow velocity, flow rate, or sediment transport volume), is the model predicted value.

[0110] During the training process, to prevent overfitting, use the L2 regularization method and add a regularization term (λ is the regularization coefficient, usually taking a relatively small value, such as 0.01), and at this time the loss function becomes These are model parameters. The regularization term can constrain the magnitudes of the model parameters to avoid overfitting caused by an overly complex model.

[0111] Through the above steps, the non-contact hydrological and sediment monitoring strategy based on multi-source data fusion of the present disclosure can make full use of various technical means and algorithm models to achieve accurate monitoring of hydrological and sediment parameters, providing strong data support and decision-making basis for water resource management, operation of water conservancy projects, and ecological environment protection. The innovative designs in aspects such as sensor layout, data fusion, and signal processing improve the reliability and adaptability of the monitoring method.

[0112] Embodiment 2

[0113] In one embodiment of the present disclosure, a hydrological prediction system for cryosphere basins based on multi-source data fusion is provided, including:

[0114] A data acquisition module, configured to acquire vibration signals of the interaction between water flow and riverbed and meteorological data in the region, and perform preprocessing;

[0115] A data fusion module, configured to input the preprocessed meteorological data and vibration signals into a multi-source data fusion model, respectively extract temporal features of the meteorological data and vibration signals through a multi-layer perceptron in the multi-source data fusion model, and then fuse the extracted temporal features through a gated recurrent unit to obtain multi-source data fusion features;

[0116] A hydrological prediction module, configured to perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signals to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signals, input the signal features and multi-source data fusion features into a hydrological parameter prediction model, and output predicted values of corresponding water body velocity, flow rate, and sediment transport volume.

[0117] As an embodiment, the method executed by the hydrological prediction system for cryosphere basins based on multi-source data fusion is as shown in Embodiment 1, and the specific application process is as follows:

[0118] 1) Arrange vibration sensors in the local monitoring area

[0119] 2) Transmit and store the collected data

[0120] 3) Perform multi-source data fusion at the remote base station

[0121] 4) Perform signal processing and analysis at the remote base station to obtain prediction parameters.

[0122] Embodiment 3

[0123] In one embodiment of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the hydrological prediction method for cryosphere basins based on multi-source data fusion is implemented.

[0124] Embodiment 4

[0125] In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, and the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the hydrological prediction method for cryosphere basins based on multi-source data fusion is implemented.

[0126] Embodiment 5

[0127] In one embodiment of the present disclosure, an electronic device is provided, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory, so that the electronic device executes and implements the hydrological prediction method for cryosphere basins based on multi-source data fusion.

[0128] The present disclosure is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0130] Although the specific implementation manners of the present disclosure are described above in conjunction with the accompanying drawings, it is not a limitation to the protection scope of the present disclosure. Those skilled in the art should understand that, based on the technical solutions of the present disclosure, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present disclosure.

Claims

1. A hydrological prediction method for cryosphere basins based on multi-source data fusion, characterized in that Including: Obtain the vibration signals of the interaction between the water flow and the river bed and the meteorological data in the area, and perform preprocessing; Input the preprocessed meteorological data and vibration signals into the multi-source data fusion model. Use the multi-layer perceptron in the multi-source data fusion model to extract the time-series features of the meteorological data and vibration signals respectively, and then use the gated recurrent unit to fuse the extracted time-series features to obtain the multi-source data fusion features; Perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signals to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signals. Input the signal features and the multi-source data fusion features into the hydrological parameter prediction model to output the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

2. The hydrological prediction method for cryosphere basins based on multi-source data fusion according to claim 1, wherein The acquisition method of the vibration signals of the interaction between the water flow and the river bed is as follows: According to the set sensor layout method, in the glacier front area, adopt the dense linear layout method, and arrange sensors in sequence along the direction of glacier meltwater flow to capture the subtle vibration signals generated by the interaction between the water flow and the river bed and surrounding substances when the water flows out of the glacier; in the main river channel area of the river, adopt a combination of curve type and dense layout in key areas; for the area around the lake, arrange sensors in a fan-shaped radiation outward at the lake outlet; use the wireless ad hoc network technology to converge the scattered sensor data to the area with strong signal, and then through network transmission, realize the acquisition of the vibration signals of the interaction between the water flow and the river bed.

3. The hydrological prediction method for cryosphere basins based on multi-source data fusion according to claim 1, characterized in that, Synchronously collect and time-align the meteorological data and vibration signals in the area. Adopt the time-stamp-based synchronization algorithm, equip each meteorological sensor and vibration sensor with a high-precision clock module. The clock module obtains time information accurate to the millisecond level. When collecting data, record the time stamp of data collection at the same time, and use the linear interpolation algorithm to perform alignment compensation on the time deviation; then, use the box plot method to identify the outliers in the meteorological data, and use the wavelet threshold denoising algorithm to perform denoising processing on the vibration signals to realize the preprocessing of the data.

4. The cryosphere basin hydrological prediction method based on multi-source data fusion according to claim 1, characterized in that: The multi-source data fusion model is a neural network model based on the combination of a multi-layer perceptron and a gated recurrent unit. The multi-layer perceptron extracts features from the meteorological data and vibration signal data. The number of neurons in the input layer of the multi-layer perceptron is determined according to the number of features of the meteorological data and vibration signal data. The meteorological data includes m features such as temperature, precipitation, and air pressure, and the vibration signal data is n features, so the number of neurons in the input layer is m + n, and the time-series features of the meteorological data and vibration signal data are output. The input of the gated recurrent unit is the output of the multi-layer perceptron, which effectively captures the time-series features in the data for fusion to obtain a multi-source data fusion feature.

5. The hydrological prediction method for cryosphere basins based on multi-source data fusion according to claim 1, characterized in that Perform empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signal, including: performing empirical mode decomposition on the vibration signal to decompose it into a finite number of intrinsic mode functions and a residual function, performing Hilbert transform on each intrinsic mode function to obtain its analytic signal, calculating the instantaneous frequency of the analytic signal, constructing a Hilbert spectrum to calculate the instantaneous amplitude, obtaining the energy distribution of the signal in the time-frequency plane through the Hilbert spectrum, and extracting the characteristic frequencies and amplitude information related to water flow and sediment movement.

6. The hydrological prediction method for cryosphere basins based on multi-source data fusion according to claim 1, characterized in that The hydrological prediction model is a long short-term memory neural network, which consists of multiple LSTM units. Each LSTM unit includes an input gate, a forget gate, an output gate, and a memory unit. The instantaneous frequency, instantaneous amplitude, and multi-source data fusion features obtained from Hilbert spectrum analysis are used as the input of the hydrological prediction model to effectively process the time-dependent relationship related to the hydrological process contained in the vibration signal, and the predicted values of the corresponding water velocity, flow rate, and sediment transport volume are used as the output to achieve the preprocessing results of hydrological and sediment parameters, and further achieve the accurate monitoring of hydrological and sediment parameters.

7. A cryosphere basin hydrological prediction system based on multi-source data fusion, characterized in that, Including: A data acquisition module for acquiring the vibration signal of the interaction between water flow and river bed and the meteorological data in the region, and performing preprocessing; A data fusion module for inputting the preprocessed meteorological data and vibration signal into a multi-source data fusion model, respectively extracting the time series features of the meteorological data and vibration signal through a multi-layer perceptron in the multi-source data fusion model, and then fusing the extracted time series features through a gated recurrent unit to obtain multi-source data fusion features; A hydrological prediction module for performing empirical mode decomposition and Hilbert spectrum analysis on the preprocessed vibration signal to obtain multiple signal features such as the instantaneous frequency and instantaneous amplitude of the vibration signal, inputting the signal features and multi-source data fusion features into a hydrological parameter prediction model, and outputting the predicted values of the corresponding water velocity, flow rate, and sediment transport volume.

8. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the multi-source data fusion-based hydrological prediction method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, it implements the multi-source data fusion-based hydrological prediction method according to any one of claims 1-6.

10. An electronic device, characterized in that, Including: A processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory so that the electronic device executes the multi-source data fusion-based hydrological prediction method according to any one of claims 1-6.

Images