An ecological restoration method and system based on river monitoring

Abstract

The application discloses an ecological restoration method and system based on river monitoring, wherein the method comprises the following steps: acquiring water flow velocity data of a target river, and calculating a trend component and a fluctuation component of the water flow velocity data according to a discrete wavelet transform method; acquiring dissolved oxygen data, pH value data and turbidity data of the target river, and respectively pre-processing the dissolved oxygen data, the pH value data and the turbidity data to obtain a change rate of the dissolved oxygen data, a mapping relationship function of the pH value and aeration intensity, and a fractal dimension of the turbidity data; constructing a comprehensive adjustment factor according to the trend component, the fluctuation component, the change rate, the mapping relationship function and the fractal dimension; inputting the comprehensive adjustment factor into a pre-constructed LSTM model for prediction to obtain an ideal adjustment factor of the target river, and adjusting the aeration intensity of the target river according to the ideal adjustment factor to realize ecological restoration of the target river. The method provided by the application improves the accuracy of river ecological monitoring.

Description

Technical Field

[0001] This application relates to the field of environmental monitoring technology, and in particular to an ecological restoration method and system based on river monitoring. Background Technology

[0002] With the acceleration of urbanization and the increasing demands for water resource management, river ecological monitoring has become increasingly important. Through river ecological monitoring, potential ecological problems can be detected and warned in a timely manner, ensuring the health and stability of the aquatic ecosystem.

[0003] However, existing water quality monitoring technologies often focus only on a single water quality parameter, lacking a comprehensive analysis of the relationships between different parameters. This makes it impossible to fully consider the impact of various factors when formulating remediation strategies, thus affecting the formulation and implementation of remediation strategies. Summary of the Invention

[0004] This application aims to provide an ecological restoration method and system based on river monitoring, which solves the technical problem of refined monitoring of river ecology.

[0005] To address the aforementioned technical problems, embodiments of this application provide an ecological restoration method based on river monitoring, comprising:

[0006] Acquire water flow velocity data captured by several water flow velocity sensors located in the target river area, and calculate the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method;

[0007] Data on dissolved oxygen, pH, and turbidity captured by several water quality sensors located in the target river area are obtained. The dissolved oxygen, pH, and turbidity data are preprocessed to obtain the rate of change of dissolved oxygen, the mapping function between pH and aeration intensity, and the fractal dimension of turbidity data.

[0008] A comprehensive adjustment factor is constructed based on the trend component, the fluctuation component, the rate of change, the mapping function, and the fractal dimension. The comprehensive adjustment factor reflects the degree of influence of water flow parameters and water quality parameters on the aeration intensity of the target river channel.

[0009] The comprehensive adjustment factor is input into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target river channel;

[0010] The aeration intensity of the target river channel is adjusted according to the ideal adjustment factor to achieve ecological restoration of the target river channel.

[0011] As one preferred embodiment, the step of calculating the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method includes:

[0012] The water flow velocity data is decomposed at a preset scale based on a preset wavelet basis function to obtain approximation coefficients and detail coefficients.

[0013] The approximation coefficients of the lowest frequency are reconstructed to obtain a trend component, which reflects the trend of water flow velocity over a long time scale.

[0014] The fluctuation component is calculated based on the trend component, which reflects the fluctuation of the water flow velocity relative to the trend over a short time scale.

[0015] As one preferred embodiment, the preprocessing of the dissolved oxygen data, pH data, and turbidity data to obtain the rate of change of the dissolved oxygen data, the mapping function between the pH value and the aeration intensity, and the fractal dimension of the turbidity data includes:

[0016] Calculate the long-term and short-term rates of change of the dissolved oxygen data based on the dissolved oxygen data;

[0017] The mapping relationship function between pH value and aeration intensity was established using an SVM regression model.

[0018] The fractal dimension of the turbidity data was calculated using the box counting method.

[0019] As one preferred embodiment, the step of inputting the comprehensive adjustment factor into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target river channel includes:

[0020] A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value;

[0021] The training set is input into the pre-constructed initial LSTM model for training to obtain the LSTM model;

[0022] In the actual prediction process, the obtained real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

[0023] As one preferred embodiment, adjusting the aeration intensity of the target river channel according to the ideal adjustment factor includes:

[0024] The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows:

[0025]

[0026] Where I(t) is the adjusted aeration intensity, and I0 is the initial aeration intensity. It is the Sigmoid function. t0 is the ideal adjustment factor, t0 is the time parameter related to the biological clock of the target river ecosystem, and τ is the time constant.

[0027] Another embodiment of this application provides an ecological restoration system based on river monitoring, comprising:

[0028] The first acquisition module is used to acquire water flow velocity data captured by several water flow velocity sensors located in the target river area, and calculate the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method.

[0029] The second acquisition module is used to acquire dissolved oxygen data, pH data and turbidity data captured by several water quality sensors located in the target river area, and preprocess the dissolved oxygen data, pH data and turbidity data respectively to obtain the rate of change of dissolved oxygen data, the mapping relationship function between pH value and aeration intensity and the fractal dimension of turbidity data.

[0030] A construction module is used to construct a comprehensive adjustment factor based on the trend component, the fluctuation component, the rate of change, the mapping function, and the fractal dimension. The comprehensive adjustment factor reflects the degree of influence of water flow parameters and water quality parameters on the aeration intensity of the target river channel.

[0031] The prediction module is used to input the comprehensive adjustment factor into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor of the target river channel;

[0032] An adjustment module is used to adjust the aeration intensity of the target river channel according to the ideal adjustment factor in order to achieve ecological restoration of the target river channel.

[0033] As one preferred embodiment, the first acquisition module is specifically used for:

[0034] The water flow velocity data is decomposed at a preset scale based on a preset wavelet basis function to obtain approximation coefficients and detail coefficients.

[0035] The approximation coefficients of the lowest frequency are reconstructed to obtain a trend component, which reflects the trend of water flow velocity over a long time scale.

[0036] The fluctuation component is calculated based on the trend component, which reflects the fluctuation of the water flow velocity relative to the trend over a short time scale.

[0037] As one preferred embodiment, the second acquisition module is specifically used for:

[0038] Calculate the long-term and short-term rates of change of the dissolved oxygen data based on the dissolved oxygen data;

[0039] The mapping relationship function between pH value and aeration intensity was established using an SVM regression model.

[0040] The fractal dimension of the turbidity data was calculated using the box counting method.

[0041] As one preferred embodiment, the prediction module is specifically used for:

[0042] A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value;

[0043] The training set is input into the pre-constructed initial LSTM model for training to obtain the LSTM model;

[0044] In the actual prediction process, the obtained real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

[0045] As one preferred embodiment, the adjustment module is specifically used for:

[0046] The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows:

[0047]

[0048] Where I(t) is the adjusted aeration intensity, and I0 is the initial aeration intensity. It is the Sigmoid function. t0 is the ideal adjustment factor, t0 is the time parameter related to the biological clock of the target river ecosystem, and τ is the time constant.

[0049] Compared to the prior art, the beneficial effects of the embodiments of this application are at least one of the following:

[0050] (1) This application uses the discrete wavelet transform method to decompose the water flow velocity data into trend components and fluctuation components, which can more accurately understand the changing trend and short-term fluctuation of water flow velocity, and provide a more accurate basis for subsequent adjustments. At the same time, the dissolved oxygen, pH value and turbidity data are preprocessed to extract features such as change rate, mapping relationship function and fractal dimension. These features can more comprehensively reflect the water quality status and its changing trend, and improve the accuracy and sensitivity of water quality assessment.

[0051] (2) The comprehensive adjustment factor constructed in this application integrates multiple key features of flow parameters and water quality parameters, which can comprehensively reflect the influence of these factors on aeration intensity, avoid the limitations of single parameter adjustment, and improve the accuracy and scientific nature of the adjustment. At the same time, by inputting the comprehensive adjustment factor into the pre-constructed LSTM model for prediction, the ideal adjustment factor of the target river can be obtained. The long short-term memory capability of the LSTM model enables it to capture complex patterns in time series data, thereby improving the accuracy of prediction. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating an ecological restoration method based on river monitoring in one embodiment of this application.

[0053] Figure 2 This is a three-dimensional schematic diagram of an aerator in one embodiment of this application;

[0054] Figure 3 This is a schematic internal cross-sectional view of the aerator in one embodiment of this application;

[0055] Figure 4 This is a schematic diagram of an ecological restoration system based on river monitoring in one embodiment of this application;

[0056] Figure label:

[0057] Among them, 1. the air outlet of the micropores; 2. the air inlet pipe; 3. the main air vent pipe connected to the air inlet pipe. Detailed Implementation

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of this application more thorough and comprehensive. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0059] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0060] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0061] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in the specification of this application is for the purpose of describing specific embodiments only and is not intended to limit the application. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0062] One embodiment of this application provides an ecological restoration method based on river monitoring. For details, please refer to [link to relevant documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart of an ecological restoration method based on river monitoring according to one embodiment of this application, including steps S1-S5:

[0063] S1: Acquire water flow velocity data captured by several water flow velocity sensors located in the target river area, and calculate the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method;

[0064] In this step, a water flow velocity sensor based on the Doppler effect is used to measure the water flow velocity at fixed time intervals (e.g., every 5 minutes), and the measured value is v(t) (where t represents time). The water flow velocity sensor is based on the Doppler frequency shift principle; when a sound wave (or other wave) encounters a moving object (particles in the water flow), the frequency of the reflected wave changes. By measuring this frequency change, the water flow velocity can be calculated. The normal range for the water flow velocity is determined to be [v...]. min v max This range is determined based on long-term observation of the river channel or according to river channel design and ecological needs.

[0065] Preferably, in one embodiment of this application, calculating the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method includes:

[0066] Based on the preset wavelet basis function, the water flow velocity data is decomposed at a preset scale to obtain approximation coefficients and detail coefficients;

[0067] The approximate coefficients of the lowest frequency are reconstructed to obtain the trend component, which reflects the changing trend of water flow velocity over a long time scale.

[0068] The fluctuation component is calculated based on the trend component, which reflects the fluctuation of the water flow velocity relative to the trend over a short time scale.

[0069] In other words, to obtain the dynamic characteristics of water flow velocity, this application uses Discrete Wavelet Transform (DWT) to decompose the water flow velocity data v(t). Discrete Wavelet Transform is a mathematical tool that decomposes a signal into different frequency components. A suitable wavelet basis function (such as the db4 wavelet in the Daubechies wavelet system) is selected to decompose the water flow velocity data. For v(t), this application selects J scales for decomposition, and at each scale j = 1, 2, ..., J, the approximate coefficients a are obtained. j，k and detail coefficient d j，k Where k represents the translation amount, and its value range depends on the length of the data and the decomposition scale.

[0070] It should be noted that the approximation coefficient 'a' j，k This represents the low-frequency portion of the signal, reflecting trends in water flow velocity; for example, approximation coefficients at a larger scale may represent seasonal or long-term trends in water flow velocity. Detail coefficient d j，k This represents the high-frequency portion of the signal, reflecting fluctuations in water flow velocity over shorter timescales, such as changes in velocity caused by local topography, short-term weather variations, or other sudden factors. The low-frequency approximation coefficient 'a' is used to... j,k The trend component v of the water flow velocity is obtained by reconstruction. trend (t). The trend component describes the trend of water flow velocity over a longer time scale, such as seasonal water flow velocity trends or long-term velocity trends caused by factors such as river channel topography.

[0071] In order to obtain the trend component v of the water flow velocity trend (t), the approximation coefficients for the lowest frequency (i.e., the approximation coefficients a at the largest scale J). j,k The reconstruction process involves reconstructing a based on the inverse wavelet transform formula. j，k Converting back to the time domain signal yields v trend (t). Trend component v trendThe calculation of (t) is performed by approximating the lowest frequency coefficient a. j，k The reconstruction is performed using (J=3) the filter. The reconstruction process is the reverse of the decomposition process, using the same low-pass filter as the one used in the decomposition process. and high-pass filter

[0072] Specifically, the reconstruction formula is:

[0073] Furthermore, the fluctuation component is the result of subtracting the trend component from the original flow velocity data. The fluctuation component v of the flow velocity is calculated. fluct (t)=v(t)-v trend (t). The fluctuation component reflects the fluctuation of water flow velocity relative to the trend on a shorter time scale, such as velocity fluctuations caused by local water flow disturbances or short-term meteorological factors.

[0074] In this step, the fluctuation and trend components are normalized. Normalization makes the changes in water flow velocity at different time scales and data points comparable, facilitating subsequent analysis and comparison. The trend component v is normalized separately. trend (t) and fluctuation component v fluct (t) is normalized:

[0075] Trend component v trend The normalization formula for (t) is: Map it to the [0,1] interval for easier subsequent calculations and comparisons.

[0076] Fluctuation component v fluct The normalization formula for (t) is: Where v fluct，min and v fluct，max These are the lower and upper limits of the fluctuation components obtained from historical data statistics. This also maps the fluctuation components to the [0, 1] interval.

[0077] S2: Obtain dissolved oxygen, pH, and turbidity data captured by several water quality sensors located in the target river area. Preprocess the dissolved oxygen, pH, and turbidity data respectively to obtain the rate of change of dissolved oxygen, the mapping function between pH and aeration intensity, and the fractal dimension of turbidity data.

[0078] In this step, several water quality sensors are installed within the target river area. These sensors should be able to capture water quality parameters, including dissolved oxygen, pH, and turbidity, in real time or periodically. Data is collected from these sensors, ensuring its accuracy and completeness. The data can be transmitted to the data collection system via wired or wireless means.

[0079] Preferably, in one embodiment of this application, the dissolved oxygen data, pH data, and turbidity data are preprocessed to obtain the rate of change of dissolved oxygen data, the mapping function between pH value and aeration intensity, and the fractal dimension of turbidity data, including:

[0080] Calculate the long-term and short-term rates of change of dissolved oxygen data based on dissolved oxygen data.

[0081] A mapping function between pH value and aeration intensity was established using an SVM regression model.

[0082] The fractal dimension of turbidity data was calculated using the box counting method.

[0083] It is understandable that dissolved oxygen (DO) is measured by comparing dissolved oxygen values ​​at different time points and calculating the change by dividing by the time interval to obtain DO(t), and a target value DO is set. target and normal range [DO] min DO max [Among them, the time interval for calculating short-term dissolved oxygen is shorter than that for calculating long-term dissolved oxygen.]

[0084] In some implementations, the short-term rate of change ΔDO is calculated. s (t) = DO(t) - DO(t-Δt), where Δt = 10 minutes, is used to capture changes in dissolved oxygen content over a short period of time, such as fluctuations in dissolved oxygen caused by local biological activity or short-term water quality changes.

[0085] Calculate the long-term rate of change ΔDO l (t) = DO(t) - DO(tT), where T = 12 hours. This rate of change reflects the trend of dissolved oxygen changes over a longer period of time, such as when affected by diurnal rhythms or large-scale water quality changes.

[0086] In this step, the dissolved oxygen content (DO(t)) is normalized. This normalization process converts the dissolved oxygen content, short-term rate of change, and long-term rate of change to the same scale, facilitating subsequent data analysis and comparison. The formula is expressed as follows:

[0087] Specifically, for the short-term rate of change ΔDO s Normalize (t), the formula is expressed as: Where ΔDO s,min and ΔDO s，max These are the lower and upper limits of the short-term rate of change, determined based on historical data.

[0088] For the long-term rate of change ΔDO l Normalize (t), the formula is expressed as: Where ΔDO l，min and ΔDOl，max These are the lower and upper limits of the long-term rate of change, determined based on historical data.

[0089] In this step, an SVM regression model is used to establish the relationship between pH values ​​and aeration intensity. SVM is a machine learning model based on statistical learning theory and has good generalization ability. This application uses a pH sensor based on the principle of glass electrode or ion-selective electrode to measure the pH value of the water body at fixed time intervals (e.g., every 5 minutes). The normal range of pH value is set according to the water body type, water quality standards, and treatment objectives. For example, for some freshwater ecosystems, the normal range of pH value may be set to 6.5 to 8.5.

[0090] Historical pH data and corresponding aeration intensity adjustments were collected from previous experiments, monitoring, or manual experience records. The correspondence between data timestamps, pH values, and aeration intensity adjustments was ensured to be accurate, and outliers or noisy data were removed to guarantee data accuracy and reliability.

[0091] In some implementations, pH value is used as the input feature (independent variable), and the aeration intensity adjustment value is used as the output label (dependent variable). Since only the raw pH value is used as the feature, no additional feature transformation or feature engineering is required. However, in some complex cases, feature expansion or dimensionality reduction may be necessary.

[0092] Choosing an appropriate kernel function for an SVM regression model is crucial. Commonly used kernel functions include linear kernels, polynomial kernels, and radial basis functions (RBF).

[0093] In some implementations, the RBF kernel function can be selected, which can handle nonlinear relationships. Other model parameters, such as the penalty coefficient and the width parameter of the RBF kernel function, are determined using methods such as cross-validation. The selection of these parameters has a significant impact on the model's performance and generalization ability. The SVM model is trained using prepared historical data. The training objective is to minimize the error between the predicted aeration intensity adjustment value and the actual value. Commonly used error metrics include mean squared error (MSE). The mean squared error is expressed as:

[0094] Where y i This is the actual aeration intensity adjustment value. Here, is the predicted value, and n is the number of data points. The model's performance is evaluated using a test dataset, including prediction accuracy and generalization ability. If the model's performance is poor, it can be optimized by adjusting parameters, increasing the amount of data, or improving feature selection.

[0095] After training, a mapping function between pH value and aeration intensity is obtained. This function can predict a suitable adjustment value for aeration intensity based on the current pH value, taking into account the complex nonlinear influence of pH value on aeration intensity. For example, when pH(t) = 7, f p H(7) may return a value indicating the proportion of adjustment required to the aeration intensity at this pH value.

[0096] In this step, the box counting method is used to calculate the fractal dimension D of the turbidity data. T (t). Fractal dimension is an indicator that describes the complexity and irregularity of a geometric object. For turbidity data, fractal dimension can reflect the complexity of turbidity variations, such as the mixing of pollutants from different sources or the complex effects of water flow on turbidity.

[0097] Specifically, the turbidity sensor measures turbidity every 5 minutes, obtaining the measured value T(t). The turbidity sensor is based on the principle of light scattering or transmission, determining turbidity by measuring the degree to which suspended particles in the water scatter or transmit light. The fractal dimension D of the turbidity data is calculated using the box counting method. T (t). The basic steps of box counting are as follows:

[0098] 1) Determine the range of turbidity data and divide it into "boxes" of different sizes. Initially, you can choose a smaller box size (e.g., the side length of the box is a small positive number), and then gradually increase the box size. For example, for data with turbidity values ​​in the range [0, 100], you can start by dividing it into smaller box sizes (e.g., the side length of the box is 1).

[0099] 2) For each box size, count the number of boxes containing turbidity data points. The number of boxes containing data points typically decreases as the box side length increases. For example, count the number of boxes containing turbidity data points N(∈), where ∈ is the side length of the box. N(∈) will change as the box side length ∈ changes.

[0100] 3) Perform logarithmic processing on the number of boxes for different box side lengths, i.e., calculate log(number of boxes) and log(box side length). Use linear regression or other fitting methods to fit log(number of boxes) and log(box side length) to obtain the slope of the fitted line. This slope is the fractal dimension D of the turbidity data. In two-dimensional space, the fractal dimension D is usually between 1 and 2, where D=1 represents a perfectly regular geometric shape, and D=2 represents a completely filled plane.

[0101] Specifically, according to the definition of fractal dimension By performing logarithmic fitting on N(∈) under different ∈, the fractal dimension D of the turbidity data at each time t is calculated.T (t). Fractal dimension can reflect the complexity of turbidity changes; for example, if turbidity is caused by a mixture of suspended particles of various sizes, its fractal dimension may be high.

[0102] 4) Based on historical data, determine the lower bound D_min and upper bound D_max of the fractal dimension. These values ​​can be obtained by analyzing the fractal dimension distribution of historical turbidity data. Map the fractal dimension to a specified interval (e.g., 0 to 1) using a normalization formula for calculation in the comprehensive adjustment factor. The formula is expressed as:

[0103]

[0104] S3: Construct a comprehensive adjustment factor based on trend component, fluctuation component, rate of change, mapping relationship function and fractal dimension. The comprehensive adjustment factor reflects the degree of influence of flow parameters and water quality parameters on the aeration intensity of the target river channel.

[0105] In this step, the comprehensive adjustment factor C(t) is an index that comprehensively considers the influence of multiple factors such as water flow velocity, dissolved oxygen content, pH value, and turbidity on aeration intensity. The formula is expressed as:

[0106] C(t) = w1 × v n1 (t)×(1+w2×v n2 (t))+w3×DO n1 (t)×(1+w4×DO n2 (t)+w5×DO n3 (t)+w6×f pH (pH(t))+w7×D T，n (t)

[0107] Among them, w1×v n1 (t)×(1+w2×v n2 (t) represents the trend component of the water flow velocity v n1 (t) and its fluctuation component v n2 The effect of w(t) on the comprehensive adjustment factor C(t), w3×DO n1 (t)×(1+w4×DO n2 (t)+wx×DO n3 (t) represents the dissolved oxygen content DO. n1 (t) and its short-term rate of change DO n2 (t) and long-term rate of change DO n3 The effect of (t), w1×v n1 (t)×(1+w2×v n2 (t) represents the trend component of the water flow velocity v n1 (t) and its fluctuation component v n2The effect of w(t) on the comprehensive adjustment factor C(t), w3×DO n1 (t)×(1+w4×DO n2 (t)+w5×DO n3 (t) represents the dissolved oxygen content DO. n1 (t) and its short-term rate of change DO n2 (t) and long-term rate of change DO n3 The effect of (t), w6×f p H(pH(t)) represents the contribution of the mapping relationship between pH value and aeration intensity obtained through SVM regression to the comprehensive adjustment factor, w7×D T，n (t) represents the effect of the fractal dimension of turbidity on the overall adjustment factor. The weighting coefficients w1-w7 were determined based on experiments and actual river conditions.

[0108] S4: Input the comprehensive adjustment factor into the pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target channel;

[0109] Preferably, in one embodiment of this application, the comprehensive adjustment factor is input into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target channel, including:

[0110] A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value;

[0111] The training set is input into the pre-built initial LSTM model for training to obtain the LSTM model;

[0112] In the actual prediction process, the real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

[0113] In related technologies, LSTM is a special type of recurrent neural network (RNN) suitable for processing time series data. Internally, the LSTM network processes sequential information through a special gate structure (forget gate, input gate, and output gate), effectively capturing long-term dependencies in time series data. For example, it can learn the influence patterns of past comprehensive adjustment factors on current aeration intensity adjustments, such as the cumulative impact of changes in parameters like water flow velocity and dissolved oxygen content over a period of time on current aeration intensity adjustments.

[0114] Specifically, an LSTM network consists of multiple LSTM units, each containing an input gate i. t Forgotten Gate t Output gate o t and cell state C t Input gate i t =σ(W xi xt +W hi h t-1 +b i The current input x is determined. t (here x) t How much information in C(t) can update the cell state, where W xi and W hi It is the weight matrix, b i This is the bias term, where σ is the sigmoid function that maps values ​​to the interval (0, 1). The forget gate f t =σ(W xf x t +W hf h t-1 +b f The cell state C at the previous moment is determined. t-1 How much information can be forgotten? Cell state is represented as... here This represents element-wise multiplication. The cell state is updated through a forget gate and an input gate, and the tanh function maps values ​​to the interval (-1, 1). The output gate is... t =σ(W xo x t +W ho h t-1 +b o ) determines the cell state C t How much information can be output from the hidden state? Hidden state h t As the output of the LSTM unit, it also serves as the input h of the next LSTM unit. t-1 .

[0115] In this step, historical composite adjustment factor sequences and their corresponding ideal values ​​for the target river channel are collected. This historical data should cover a sufficient time span to capture long-term dependencies in the time series. The historical data is cleaned, denoised, and standardized to ensure accuracy and consistency. Based on the obtained historical composite adjustment factor sequences and their corresponding ideal values, a training set is constructed for training the LSTM model. Typically, the data needs to be divided into input sequences and output labels; the input sequences are continuous composite adjustment factor values, and the output labels are the corresponding ideal adjustment factor values.

[0116] Constructing the initial LSTM model includes setting the network structure (such as the number of hidden layers, the number of neurons, etc.), activation function, optimizer, etc. In some embodiments, it is assumed that the time window size n = 10, that is, the length of the input sequence is 10, and a suitable LSTM network structure is selected, for example, setting the number of hidden layer neurons to 20.

[0117] After building the initial model, the training set is input into the LSTM model for training. The model parameters are adjusted using the backpropagation algorithm to minimize the prediction error. During training, strategies such as cross-validation and early stopping can be used to prevent overfitting.

[0118] Specifically, the historical sequence of the comprehensive adjustment factor {C(tn), C(t-n+1), ..., C(t-1)} is used as the input x of the LSTM network. t After processing by the LSTM network, the output is an ideal adjustment factor that takes into account the influence of historical data. This not only incorporates the combined influence of various parameters on aeration intensity at the current moment but also integrates useful information from historical data, making aeration intensity adjustments more consistent with the dynamic changes in the river ecosystem. The output is obtained by inputting the historical sequence of the comprehensive adjustment factor C(t) into a pre-trained LSTM network. When training the LSTM network, a supervised learning method is used, which combines the historical integrated adjustment factor sequence with its corresponding ideal values ​​obtained through actual verification or simulation. Used as training data. For example, the sequence C(t) of the past 1000 time points and its corresponding ideal values. The dataset is divided into training, validation, and test sets in a ratio of 7:2:1.

[0119] In the actual prediction process, the comprehensive adjustment factor data of the target river channel is acquired in real time. The preprocessed real-time comprehensive adjustment factor data is then input into a trained LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

[0120] S5: Adjust the aeration intensity of the target river channel according to the ideal adjustment factor to achieve ecological restoration of the target river channel.

[0121] Preferably, in one embodiment of this application, adjusting the aeration intensity of the target river channel according to an ideal adjustment factor includes:

[0122] The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows:

[0123]

[0124] Where I(t) is the adjusted aeration intensity, and I0 is the initial aeration intensity. It is the Sigmoid function, which will Mapped to the interval (0, 1) t0 is the ideal adjustment factor, t0 is the time parameter related to the biological clock of the target river ecosystem, and τ is the time constant.

[0125] In this step, mapping using the Sigmoid function can... This is converted into a proportional factor suitable for adjusting the aeration intensity, making the adjustment smoother and more reasonable. This section considers the impact of time factors on aeration intensity adjustment. t0 is a time parameter related to the river ecosystem's biological clock; for example, if biological activity in the river peaks in the morning or evening, t0 can be set to the corresponding time. τ is a time constant used to control the time scale of adjustment. For example, when τ is small, the aeration intensity is more sensitive to time factors and adjusts faster; when τ is large, the response is relatively slower and the adjustment is more gradual.

[0126] The above scheme comprehensively, meticulously, and creatively considers the influence of multiple factors such as water flow velocity, dissolved oxygen content, water quality, and turbidity on aeration intensity. Furthermore, it utilizes various novel modeling techniques and reasonable mathematical formulas to construct a system that enables more intelligent and precise adjustment of aeration intensity based on the actual conditions and ecological needs of the river, thereby better achieving the goal of river ecological restoration.

[0127] In some embodiments of this application, the ecological restoration equipment selected is an aerator, and the aeration head of the aerator is made of a special microporous ceramic material. This material has extremely high porosity and a uniform microporous structure, which can generate extremely small bubbles. These tiny bubbles rise slowly in water and have a large contact area with the water, thereby greatly increasing the solubility of oxygen in the water and improving aeration efficiency. A schematic diagram of the aeration head is shown below. Figure 2 and Figure 3 As shown. Among them Figure 2 This is a three-dimensional schematic diagram of an aerator in one embodiment of this application. Figure 3 This is a schematic diagram of the internal cross-section of an aerator in one embodiment of this application, wherein each micro-pore is connected to the main air pipe through an air pore channel.

[0128] To enable the aeration head to automatically adjust the aeration intensity based on water flow velocity and water quality, this application incorporates a series of sensors on the aeration head. Among these, the water flow velocity sensor utilizes the Doppler effect principle to accurately measure the speed and direction of the water flow. The water quality sensors include dissolved oxygen, pH, and turbidity sensors, which can monitor changes in water quality in real time.

[0129] Specifically, the intelligent control system automatically adjusts the aeration intensity of the aeration heads based on sensor data. When the water flow is fast, the system increases the aeration rate of the aeration heads to ensure that the bubbles are fully mixed with the water; when the water flow is slow, the system appropriately reduces the aeration rate of the aeration heads to avoid over-aeration. Simultaneously, when the water quality is poor and the dissolved oxygen content is low, the system increases the aeration intensity to increase the oxygen supply; when the water quality is good and the dissolved oxygen content reaches the set value, the system reduces the aeration intensity to save energy. In this way, the intelligent aerator can always maintain optimal aeration performance, providing strong support for river ecological restoration.

[0130] Another embodiment of this application provides an ecological restoration system based on river monitoring. For details, please refer to [link to relevant documentation]. Figure 4 , Figure 4 The diagram illustrates an ecological restoration system based on river monitoring provided in one embodiment of this application. It includes: a first acquisition module 11, a second acquisition module 12, a construction module 13, a prediction module 14, and an adjustment module 15, wherein...

[0131] The first acquisition module 11 is used to acquire water flow velocity data captured by several water flow velocity sensors located in the target river area, and calculate the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method.

[0132] The second acquisition module 12 is used to acquire dissolved oxygen data, pH data and turbidity data captured by several water quality sensors set in the target river area, and to preprocess the dissolved oxygen data, pH data and turbidity data respectively to obtain the rate of change of dissolved oxygen data, the mapping relationship function between pH value and aeration intensity and the fractal dimension of turbidity data.

[0133] Module 13 is used to construct a comprehensive adjustment factor based on trend component, fluctuation component, rate of change, mapping relationship function and fractal dimension. The comprehensive adjustment factor reflects the degree of influence of water flow parameters and water quality parameters on the aeration intensity of the target river channel.

[0134] Prediction module 14 is used to input the comprehensive adjustment factor into the pre-built LSTM model for prediction to obtain the ideal adjustment factor of the target channel;

[0135] The adjustment module 15 is used to adjust the aeration intensity of the target river channel according to the ideal adjustment factor in order to achieve the ecological restoration of the target river channel.

[0136] Preferably, in one embodiment of this application, the first acquisition module is specifically used for:

[0137] Based on the preset wavelet basis function, the water flow velocity data is decomposed at a preset scale to obtain approximation coefficients and detail coefficients;

[0138] The approximate coefficients of the lowest frequency are reconstructed to obtain the trend component, which reflects the changing trend of water flow velocity over a long time scale.

[0139] The fluctuation component is calculated based on the trend component, which reflects the fluctuation of the water flow velocity relative to the trend over a short time scale.

[0140] Preferably, in one embodiment of this application, the second acquisition module is specifically used for:

[0141] Calculate the long-term and short-term rates of change of dissolved oxygen data based on dissolved oxygen data.

[0142] A mapping function between pH value and aeration intensity was established using an SVM regression model.

[0143] The fractal dimension of turbidity data was calculated using the box counting method.

[0144] Preferably, in one embodiment of this application, the prediction module is specifically used for:

[0145] A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value;

[0146] The training set is input into the pre-built initial LSTM model for training to obtain the LSTM model;

[0147] In the actual prediction process, the real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

[0148] Preferably, in one embodiment of this application, the adjustment module is specifically used for:

[0149] The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows:

[0150]

[0151] Where I(t) is the adjusted aeration intensity, and I0 is the initial aeration intensity. It is the Sigmoid function. t0 is the ideal adjustment factor, t0 is the time parameter related to the biological clock of the target river ecosystem, and τ is the time constant.

[0152] Compared with the prior art, the beneficial effects of the embodiments of this application are at least one of the following:

[0153] (1) This application uses the discrete wavelet transform method to decompose the water flow velocity data into trend components and fluctuation components, which can more accurately understand the changing trend and short-term fluctuation of water flow velocity, and provide a more accurate basis for subsequent adjustments. At the same time, the dissolved oxygen, pH value and turbidity data are preprocessed to extract features such as change rate, mapping relationship function and fractal dimension. These features can more comprehensively reflect the water quality status and its changing trend, and improve the accuracy and sensitivity of water quality assessment.

[0154] (2) The comprehensive adjustment factor constructed in this application integrates multiple key features of flow parameters and water quality parameters, which can comprehensively reflect the influence of these factors on aeration intensity, avoid the limitations of single parameter adjustment, and improve the accuracy and scientific nature of the adjustment. At the same time, by inputting the comprehensive adjustment factor into the pre-constructed LSTM model for prediction, the ideal adjustment factor of the target river can be obtained. The long short-term memory capability of the LSTM model enables it to capture complex patterns in time series data, thereby improving the accuracy of prediction.

[0155] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An ecological restoration method based on river monitoring, characterized in that, include: Acquire water flow velocity data captured by several water flow velocity sensors installed within a target river channel area, and calculate the trend component and fluctuation component of the water flow velocity data using the discrete wavelet transform method; wherein, the calculation of the trend component and fluctuation component of the water flow velocity data using the discrete wavelet transform method includes: The water flow velocity data is decomposed at a preset scale based on a preset wavelet basis function to obtain approximation coefficients and detail coefficients. The approximation coefficients of the lowest frequency are reconstructed to obtain a trend component, which reflects the trend of water flow velocity over a long time scale. The fluctuation component is calculated based on the trend component, and the fluctuation component reflects the fluctuation of the water flow velocity relative to the trend of change on a short time scale. Data on dissolved oxygen, pH, and turbidity captured by several water quality sensors located in the target river area are obtained. The dissolved oxygen, pH, and turbidity data are preprocessed to obtain the rate of change of dissolved oxygen, the mapping function between pH and aeration intensity, and the fractal dimension of turbidity data. A comprehensive adjustment factor is constructed based on the trend component, the fluctuation component, the rate of change, the mapping function, and the fractal dimension. The comprehensive adjustment factor reflects the degree of influence of water flow parameters and water quality parameters on the aeration intensity of the target river channel. The comprehensive adjustment factor is input into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target river channel; The aeration intensity of the target river channel is adjusted according to the ideal adjustment factor to achieve ecological restoration of the target river channel. The adjustment of the aeration intensity of the target river channel according to the ideal adjustment factor includes: The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows: ， in, The adjusted aeration intensity, The initial aeration intensity, It is the Sigmoid function. For the ideal adjustment factor, For time parameters related to the biological clock of the target river ecosystem, It is a time constant.

2. The ecological restoration method based on river monitoring as described in claim 1, characterized in that, The preprocessing of the dissolved oxygen data, pH data, and turbidity data to obtain the rate of change of the dissolved oxygen data, the mapping function between the pH value and the aeration intensity, and the fractal dimension of the turbidity data includes: Calculate the long-term and short-term rates of change of the dissolved oxygen data based on the dissolved oxygen data; The mapping function between pH value and aeration intensity was established using an SVM regression model. The fractal dimension of the turbidity data was calculated using the box counting method.

3. The ecological restoration method based on river monitoring as described in claim 1, characterized in that, The step of inputting the comprehensive adjustment factor into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor for the target river channel includes: A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value; The training set is input into the pre-constructed initial LSTM model for training to obtain the LSTM model; In the actual prediction process, the obtained real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

4. An ecological restoration system based on river monitoring, characterized in that, include: The first acquisition module is used to acquire water flow velocity data captured by several water flow velocity sensors located within the target river channel area, and to calculate the trend component and fluctuation component of the water flow velocity data according to the discrete wavelet transform method; wherein, the first acquisition module is specifically used for: The water flow velocity data is decomposed at a preset scale based on a preset wavelet basis function to obtain approximation coefficients and detail coefficients. The approximation coefficients of the lowest frequency are reconstructed to obtain a trend component, which reflects the trend of water flow velocity over a long time scale. The fluctuation component is calculated based on the trend component, and the fluctuation component reflects the fluctuation of the water flow velocity relative to the trend of change on a short time scale. The second acquisition module is used to acquire dissolved oxygen data, pH data and turbidity data captured by several water quality sensors located in the target river area, and preprocess the dissolved oxygen data, pH data and turbidity data respectively to obtain the rate of change of dissolved oxygen data, the mapping relationship function between pH value and aeration intensity and the fractal dimension of turbidity data. A construction module is used to construct a comprehensive adjustment factor based on the trend component, the fluctuation component, the rate of change, the mapping function, and the fractal dimension. The comprehensive adjustment factor reflects the degree of influence of water flow parameters and water quality parameters on the aeration intensity of the target river channel. The prediction module is used to input the comprehensive adjustment factor into a pre-constructed LSTM model for prediction to obtain the ideal adjustment factor of the target river channel; An adjustment module is used to adjust the aeration intensity of the target river channel according to the ideal adjustment factor to achieve ecological restoration of the target river channel. Specifically, the adjustment module is used for: The adjusted aeration intensity is calculated based on the ideal adjustment factor and the aeration intensity adjustment formula, which is expressed as follows: ， in, The adjusted aeration intensity, The initial aeration intensity, It is the Sigmoid function. For the ideal adjustment factor, For time parameters related to the biological clock of the target river ecosystem, It is a time constant.

5. The ecological restoration system based on river monitoring as described in claim 4, characterized in that, The second acquisition module is specifically used for: Calculate the long-term and short-term rates of change of the dissolved oxygen data based on the dissolved oxygen data; The mapping function between pH value and aeration intensity was established using an SVM regression model. The fractal dimension of the turbidity data was calculated using the box counting method.

6. The ecological restoration system based on river monitoring as described in claim 4, characterized in that, The prediction module is specifically used for: A training set is constructed based on the obtained historical comprehensive adjustment factor sequence and its corresponding ideal value; The training set is input into the pre-constructed initial LSTM model for training to obtain the LSTM model; In the actual prediction process, the obtained real-time comprehensive adjustment factor is input into the LSTM model for prediction to obtain the ideal adjustment factor of the target river channel.

Images