A method and device for generating a water chlorophyll a prediction model based on a space-time transmission mechanism and an agent model, and an electronic device

Abstract

This invention provides a method, apparatus, and electronic device for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanisms and a surrogate model. It relates to the field of water composition analysis and prediction, and includes: determining the optimal time delay for transport from upstream to downstream and the upstream optimal time-delay chlorophyll a concentration; constructing a flow gating function based on a set flow threshold and generating flow-gated features by combining the upstream optimal time-delay chlorophyll a concentration; determining a spatiotemporally coupled feature set based on the optimal time delay, the upstream optimal time-delay chlorophyll a concentration, and the flow-gated features; calculating the importance of each feature in the full feature set based on a time-series prediction surrogate model, recursively eliminating the features with the lowest importance until the number of features in the full feature set is reduced to a preset value, obtaining the optimal feature subset; and training a water chlorophyll a prediction model based on the optimal feature subset. This invention improves prediction accuracy and provides technical support for eutrophication early warning.

Description

Technical Field

[0001] This invention relates to the field of water composition analysis and prediction technology, and in particular to a method, apparatus and electronic device for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanism and surrogate model. Background Technology

[0002] Eutrophication of water bodies seriously threatens drinking water safety. As a key indicator of algal biomass, the accurate prediction of chlorophyll a is of great significance for eutrophication control. In river network and lake-reservoir systems with upstream and downstream hydraulic connections, pollutant transport exhibits a time-space delay effect. Accurate prediction of chlorophyll a concentration changes can provide decision support for water source management.

[0003] Existing water chlorophyll a prediction technologies have the following shortcomings:

[0004] (1) Although the mechanism model method (such as CE-QUAL-W2, EFDC) has physical meaning, the parameter calibration is difficult, the computational complexity is high, and it is difficult to capture nonlinear dynamic changes.

[0005] (2) Traditional machine learning methods (such as SVM and RF) are difficult to handle long-term time dependencies, feature engineering relies on human experience, and has limited ability to fit complex nonlinear relationships.

[0006] (3) Although deep learning methods (such as CNN and RNN) can process time-series data, they still have three key technical shortcomings: First, insufficient representation of spatiotemporal mechanisms. Existing methods predict based on single-point historical data, ignoring the transmission lag from upstream to downstream, resulting in prediction lag and poor early warning timeliness. Second, low efficiency of feature selection. Using all features leads to redundancy and overfitting, and traditional feature selection does not consider the adaptability to the prediction model. Third, difficulty in multi-scale information fusion. Chlorophyll a changes include multiple time-scale patterns such as daily, weekly, and seasonal, which are difficult for existing methods to capture simultaneously.

[0007] Therefore, a method, device, and electronic equipment for generating a water chlorophyll a prediction model based on spatiotemporal transmission mechanism and proxy model are proposed. Summary of the Invention

[0008] This specification provides a method, device, and electronic equipment for generating a water chlorophyll a prediction model based on spatiotemporal transmission mechanism and surrogate model, which improves prediction accuracy and provides technical support for eutrophication early warning.

[0009] This specification provides a method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanisms and surrogate models, including:

[0010] Spatiotemporal data of upstream monitoring points and downstream monitoring points are acquired. The spatiotemporal data of upstream monitoring points includes upstream chlorophyll a concentration, and the spatiotemporal data of downstream monitoring points includes downstream chlorophyll a concentration.

[0011] Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the optimal time delay for transport from upstream to downstream and the optimal upstream time delay chlorophyll a concentration are determined.

[0012] A flow gating function is constructed based on a set flow threshold, and a flow gating feature is generated by combining the upstream optimal time-delay chlorophyll a concentration.

[0013] Based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics, a spatiotemporal coupling feature set is determined;

[0014] Ecological mechanism features, environmental stress features, time cycle features, and time-series derived features are constructed, and combined with the aforementioned spatiotemporal coupling feature set to form a complete feature set;

[0015] The importance of each feature in the full feature set is calculated based on the time series prediction proxy model. The feature with the lowest importance is recursively eliminated until the number of features in the full feature set is reduced to a preset value, thus obtaining the optimal feature subset.

[0016] A water chlorophyll a prediction model is obtained by training based on the optimal feature subset.

[0017] Optionally, determining the optimal time delay for transport from upstream to downstream based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes:

[0018] Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the Pearson correlation coefficients under different time lags were calculated.

[0019] The time delay that maximizes the absolute value of the Pearson correlation coefficient is determined as the optimal time delay.

[0020] Optionally, the calculation of the Pearson correlation coefficient under different time lags based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes:

[0021] ρ(τ)=Σ[(C up (t-τ)-μ up (C) down (t)-μ down )] / [σ up ×σ down ×n]

[0022] Where ρ(τ) is the Pearson correlation coefficient under different time delays, and Cup (t-τ) represents the upstream chlorophyll a concentration under different time lags, μ up C is the mean of the upstream sequence. down (t) represents the downstream chlorophyll a concentration, μ down σ is the mean of the downstream sequence. up σ represents the standard deviation of the upstream sequence. down τ is the standard deviation of the downstream sequence, n is the number of samples, and the maximum value of τ is determined by the distance between the upstream and downstream and the average flow velocity in the slow-flow zone.

[0023] Optionally, the step of constructing a flow gating function based on a set flow threshold and generating flow gating features by combining the upstream optimal time-delay chlorophyll a concentration includes:

[0024] F gated (t)=C up (t- )×G[Q(t- )]

[0025] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- ) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function.

[0026] Optionally, the flow gating function includes a sigmoid smoothing function or a step function;

[0027] When the flow gating function is the Sigmoid smoothing function:

[0028]

[0029] Where k is the steepness parameter, The flow threshold that minimizes the prediction error, determined through grid search;

[0030] When the flow gating function is the step function:

[0031] G(Q) = .

[0032] Optionally, the calculation of the importance of each feature in the full feature set based on the time series prediction proxy model includes:

[0033]

[0034] in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. valFor the validation set data, the expected value E is the sample mean.

[0035] Optionally, the step of training the water chlorophyll a prediction model based on the optimal feature subset includes:

[0036] The water chlorophyll a prediction model is a two-branch parallel neural network model, specifically:

[0037] The intermediate processing branch is used to process the raw time series to capture short-term change patterns;

[0038] The long-term processing branch is used to process downsampled time series to capture macro long-term trends;

[0039] The intelligent fusion layer is used to integrate the outputs of the intermediate processing branch and the long-term processing branch.

[0040] This specification provides a device for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanisms and surrogate models, comprising:

[0041] The monitoring and acquisition module is used to acquire spatiotemporal data from upstream monitoring points and downstream monitoring points. The spatiotemporal data from upstream monitoring points includes upstream chlorophyll a concentration, and the spatiotemporal data from downstream monitoring points includes downstream chlorophyll a concentration.

[0042] The first calculation module is used to determine the optimal time delay for transmission from upstream to downstream and the optimal time delay chlorophyll a concentration upstream, based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration.

[0043] The second calculation module is used to construct a flow gating function based on a set flow threshold, and generate flow gating features by combining the upstream optimal time-delay chlorophyll a concentration.

[0044] The third calculation module is used to determine the spatiotemporal coupling feature set based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics.

[0045] The construction module is used to construct ecological mechanism features, environmental stress features, time period features, and time-series derived features, which, combined with the spatiotemporal coupling feature set, form a complete feature set;

[0046] The fourth calculation module is used to calculate the importance of each feature in the full feature set based on the time series prediction surrogate model, recursively eliminate the features with the lowest importance, until the number of features in the full feature set is reduced to a preset value, and obtain the optimal feature subset.

[0047] The model generation module is used to train a water chlorophyll a prediction model based on the optimal feature subset.

[0048] Optionally, the first computing module includes:

[0049] Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the Pearson correlation coefficients under different time lags were calculated.

[0050] The time delay that maximizes the absolute value of the Pearson correlation coefficient is determined as the optimal time delay.

[0051] Optionally, the calculation of the Pearson correlation coefficient under different time lags based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes:

[0052] ρ(τ)=Σ[(C up (t-τ)-μ up (C) down (t)-μ down )] / [σ up ×σ down ×n]

[0053] Where ρ(τ) is the Pearson correlation coefficient under different time delays, and C up (t-τ) represents the upstream chlorophyll a concentration under different time lags, μ up C is the mean of the upstream sequence. down (t) represents the downstream chlorophyll a concentration, μ down σ is the mean of the downstream sequence. up σ represents the standard deviation of the upstream sequence. down τ is the standard deviation of the downstream sequence, n is the number of samples, and the maximum value of τ is determined by the distance between the upstream and downstream and the average flow velocity in the slow-flow zone.

[0054] Optionally, the second computing module includes:

[0055] F gated (t)=C up (t- )×G[Q(t- )]

[0056] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- ) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function.

[0057] Optionally, the flow gating function includes a sigmoid smoothing function or a step function;

[0058] When the flow gating function is the Sigmoid smoothing function:

[0059]

[0060] Where k is the steepness parameter, The flow threshold that minimizes the prediction error, determined through grid search;

[0061] When the flow gating function is the step function:

[0062] G(Q) = .

[0063] Optionally, the fourth computing module includes:

[0064]

[0065] in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. val For the validation set data, the expected value E is the sample mean.

[0066] Optionally, the model generation module includes:

[0067] The water chlorophyll a prediction model is a two-branch parallel neural network model, specifically:

[0068] The intermediate processing branch is used to process the raw time series to capture short-term change patterns;

[0069] The long-term processing branch is used to process downsampled time series to capture macro long-term trends;

[0070] The intelligent fusion layer is used to integrate the outputs of the intermediate processing branch and the long-term processing branch.

[0071] This specification also provides an electronic device, wherein the electronic device includes:

[0072] A processor; and a memory storing computer-executable instructions, which, when executed, cause the processor to perform any of the methods described above.

[0073] This specification also provides a computer-readable storage medium that stores one or more programs that, when executed by a processor, implement any of the methods described above.

[0074] This invention accurately characterizes the pollutant transport relationship between upstream and downstream water bodies by parameterizing the spatiotemporal transport mechanism, including determining the optimal time delay and constructing flow-gated features, thus improving the timeliness of predictions. Through recursive feature elimination based on a surrogate model (SG-RFE), a lightweight model with 20-50% of the parameters of the target model is used for feature selection, improving feature selection efficiency. A dual-branch parallel TCN-LSTM architecture is employed, with the intermediate branch processing the original sequence and the long-term branch processing the downsampled sequence, achieving effective fusion of multi-scale information and improving prediction accuracy. The model adaptive update mechanism and the four-level eutrophication early warning mechanism enhance the system's practicality and reliability. Attached Figure Description

[0075] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0076] Figure 1 A schematic diagram illustrating the principle of a method for generating a water chlorophyll a prediction model based on spatiotemporal transmission mechanism and proxy model, provided in the embodiments of this specification;

[0077] Figure 2 This is a comparison chart of the data after processing by the data quality control process and the original data provided in the embodiments of this specification;

[0078] Figure 3 A schematic diagram of the correlation coefficient curve for determining the optimal time delay provided in the embodiments of this specification;

[0079] Figure 4 This is a diagram illustrating the proxy model architecture provided in the embodiments of this specification.

[0080] Figure 5 This is a comparison chart of prediction results from different methods provided in the embodiments of this specification;

[0081] Figure 6 A schematic diagram of a water chlorophyll a prediction model generation device based on spatiotemporal transmission mechanism and proxy model provided in the embodiments of this specification;

[0082] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this specification;

[0083] Figure 8 This is a schematic diagram of a computer-readable medium provided for embodiments of this specification. Detailed Implementation

[0084] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0085] The following is in conjunction with the appendix Figure 1-8 Exemplary embodiments of the invention will be described more fully here. However, exemplary embodiments can be implemented in many forms and should not be construed as limiting the invention to the embodiments set forth herein. Rather, these exemplary embodiments are provided to make the invention more comprehensive and complete, and to facilitate a full communication of the inventive concept to those skilled in the art. The same reference numerals in the figures denote the same or similar elements, components, or parts, and therefore repeated descriptions of them are omitted.

[0086] Subject to the technical concept of this invention, the features, structures, characteristics or other details described in a particular embodiment may be combined in one or more other embodiments in a suitable manner.

[0087] In the description of specific embodiments, the features, structures, characteristics, or other details described in this invention are intended to enable those skilled in the art to fully understand the embodiments. However, it is not excluded that those skilled in the art can practice the technical solutions of this invention without one or more of the specific features, structures, characteristics, or other details.

[0088] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0089] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0090] The terms “and / or” or “and / or” include all combinations of any one or more of the listed items.

[0091] Figure 1 This is a schematic diagram illustrating the principle of a method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanisms and surrogate models, as provided in the embodiments of this specification. The method may include:

[0092] This embodiment takes Reservoir J, which has upstream and downstream hydraulic connections, as an example to construct a chlorophyll a prediction model using the method of this invention. The upstream intake of Reservoir J ​​is about 6 kilometers away from the downstream monitoring point, and the water body is in a slow-flowing zone with an average flow velocity of about 0.02 m / s.

[0093] S110: Acquire spatiotemporal data of upstream monitoring points and downstream monitoring points, wherein the spatiotemporal data of upstream monitoring points includes upstream chlorophyll a concentration and the spatiotemporal data of downstream monitoring points includes downstream chlorophyll a concentration;

[0094] In the specific implementation of this specification, water quality and quantity monitoring stations are deployed at the upstream water intake and downstream monitoring point, respectively. The equipment includes chlorophyll a sensor, flow meter, multi-parameter water quality meter, etc. At the same time, a small weather station is deployed, and the monitoring data includes wind speed and direction, air temperature, rainfall, and photosynthetically active radiation.

[0095] The spatiotemporal data includes at least flow rate, total nitrogen, total phosphorus, water temperature, pH value, and dissolved oxygen. The data is sampled once per hour, automatically collected by sensors deployed at monitoring points, and transmitted in real time to the data center via a 4G network.

[0096] Strict quality control is performed on the acquired raw data, including:

[0097] Outlier identification and correction: The Local Outlier Factor (LOF) algorithm is used for identification, with a threshold of 1.5. For each data point x, its local outlier factor is calculated. Where lrd is the locally reachable density, Let x be the set of k nearest neighbors. Preferably, k can be 20. When LOF(x) > 1.5, it is determined to be an outlier and corrected by the arithmetic mean of the data before and after it in the previous 24 hours.

[0098] Missing value imputation: For missing data caused by equipment failure or transmission interruption, time series linear interpolation is used to imput the missing data to ensure the continuity of the data sequence.

[0099] Taking downstream chlorophyll a as an example, after data quality control processing, it is compared with the original data. Figure 2 As shown, the data quality has been improved.

[0100] S120: Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, determine the optimal time delay for transport from upstream to downstream and the upstream optimal time delay chlorophyll a concentration;

[0101] Optionally, S120 includes:

[0102] Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the Pearson correlation coefficients under different time lags were calculated.

[0103] The time delay that maximizes the absolute value of the Pearson correlation coefficient is determined as the optimal time delay.

[0104] In the specific implementation of this specification, for the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, different time delays τ∈[0, τ] are calculated. max The Pearson correlation coefficient ρ(τ) is used to determine the optimal time delay, which maximizes |ρ(τ)|. .

[0105] Specifically, based on the upstream and downstream distance L = 6000m and the average flow velocity in the slow-flow zone v = 0.02m / s, the theoretical transport time is calculated to be 83.3 hours, rounded to 84 hours. The upstream chlorophyll a concentration sequence C... up (t) and downstream chlorophyll a concentration sequence C down (t), calculate the Pearson correlation coefficient for different time delays τ∈[0,84], and the results are as follows. Figure 3 As shown, the correlation coefficient reaches its maximum value of 0.15 at τ=28 hours, therefore the optimal time delay is determined. =28 hours. Similarly, a time-delay analysis was performed on the upstream water intake flow rate Q(t) to obtain τ. Q =49 hours (maximum correlation coefficient 0.11); for chlorophyll-flux product C up Analyzing (t)×Q(t), we obtain τ CQ =35 hours (maximum correlation coefficient 0.08).

[0106] Optionally, the calculation of the Pearson correlation coefficient under different time lags based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes:

[0107] ρ(τ)=Σ[(C up (t-τ)-μ up (C) down (t)-μ down )] / [σ up ×σ down ×n]

[0108] Where ρ(τ) is the Pearson correlation coefficient under different time delays, and C up (t-τ) represents the upstream chlorophyll a concentration under different time lags, μ up C is the mean of the upstream sequence. down (t) represents the downstream chlorophyll a concentration, μ down σ is the mean of the downstream sequence.up σ represents the standard deviation of the upstream sequence. down τ is the standard deviation of the downstream sequence, n is the number of samples, and the maximum value of τ is determined by the distance between the upstream and downstream and the average flow velocity in the slow-flow zone.

[0109] S130: Construct a flow gating function based on a set flow threshold, and generate flow gating features by combining the upstream optimal time-delay chlorophyll a concentration;

[0110] Optionally, S130 includes:

[0111] F gated (t)=C up (t- )×G[Q(t-τ*)]

[0112] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- ) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function.

[0113] Optionally, the flow gating function includes a sigmoid smoothing function or a step function;

[0114] When the flow gating function is the Sigmoid smoothing function:

[0115]

[0116] Where k is the steepness parameter, The flow threshold that minimizes the prediction error, determined through grid search;

[0117] When the flow gating function is the step function:

[0118] G(Q) = .

[0119] In the specific implementation of this specification, the flow threshold Q th The determination was made using a grid search method, specifically by calculating the 75th percentile of historical flow data: Q. p75 =18,500 m³ / h, in [0.8×Q p75 1.2×Q p75 Within the range [14800, 22200], a search is performed with a step size of 1000 m³ / h. For each candidate threshold, the MAE of the validation set is calculated, and the threshold that minimizes the MAE is selected: Q. th =19800m³ / h.

[0120] Constructing a Sigmoid smoothing flow gate function: Where the steepness parameter k=0.001. Generate flow gating feature: F gated (t)=C up (t-28)×G(Q(t-28)).

[0121] Four upstream time delay characteristics are formed: C up (t-28): upstream chlorophyll a concentration 28 hours later; Q(t-49): upstream water intake flow rate 49 hours later; Cup(t-35)×Q(t-35): chlorophyll-flow rate product 35 hours later.

[0122] S140: Based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics, determine the spatiotemporal coupling feature set;

[0123] In the specific embodiments described in this specification, the spatiotemporal coupling feature set includes at least the upstream optimal time-delay chlorophyll a concentration C. up (t-τ*), the set flow threshold Q(t-τ*) under the optimal time delay The interaction characteristics between the upstream optimal time-delay chlorophyll a concentration and the set flow threshold under the optimal time delay (C) up (t- )×Q(t- ), flow gating feature F gated (t), that is, {C up (t- ), Q(t- ), C up (t- )×Q(t- ), F gated (t)}.

[0124] S150: Construct ecological mechanism features, environmental stress features, time cycle features, and time-series derived features, and combine them with the aforementioned spatiotemporal coupling feature set to form a complete feature set;

[0125] In the specific implementation methods described in this specification, the ecological mechanism characteristics (a total of 8) are as follows:

[0126] Nitrogen-to-phosphorus ratio: NP ratio =N / (P+0.001)

[0127] Logarithm of nitrogen-phosphorus ratio: NP ratio_log =log(1+NP ratio )

[0128] Redfield ratio deviation: redfield deviation =|NPr atio -16|

[0129] Nutritional restriction index: nutrient limitation =min(N / 16,P)

[0130] Synergistic effect of nutrients: NPP roduct =N×P

[0131] Logarithm of synergistic effect of nutrients: NP product_log =log(1+NP_product)

[0132] Nitrogen availability: N availability =N / (N+1.0)

[0133] Phosphorus availability: P availability =P / (P+0.1)

[0134] Environmental stress characteristics (6 in total):

[0135] Temperature stress: temp stress =|T-25|, where 25℃ is the optimum temperature.

[0136] Optimal temperature range: temp optimal =1 if T∈[20,30]; else, 0

[0137] pH stress: pH stress =max(0,7.8-pH)+max(0,pH-8.4)

[0138] pH optimal range: pH optimal =1if pH∈[7.8,8.4]; else, 0

[0139] Dissolved oxygen stress: DO stress =max(0,6.0-DO)

[0140] Dissolved oxygen supersaturation: DO supersaturation =max(0,DO-12.0)

[0141] Time cycle characteristics (6 in total):

[0142] Daily cycle: hour sin =sin(2π×hour / 24),hour cos =cos(2π×hour / 24)

[0143] Seasonal cycle: doy sin =sin(2π×dayofyear / 365),doy cos =cos(2π×dayofyear / 365)

[0144] Weekly cycle: DOW sin =sin(2π×dayofweek / 7),dow cos =cos(2π×dayofweek / 7)

[0145] Temporal derived features (69 in total):

[0146] Calculations were performed for each of the original environmental factors (total nitrogen, total phosphorus, water temperature, pH, dissolved oxygen, chlorophyll a (historical data), permanganate index, and turbidity, a total of 8):

[0147] First-order difference: Δx(t) = x(t) - x(t-1) (8 characteristics)

[0148] Moving Average (MA) 6h MA 12h MA 24h MA 48h MA 72h (40 features)

[0149] Exponentially Weighted Moving Average (EWMA) 12h EWMA 48h (16 features)

[0150] In addition, five interaction features were calculated, including temperature-light interaction, pH-temperature interaction, dissolved oxygen-temperature interaction, turbidity-light ratio, and wind speed-temperature mixing.

[0151] Based on all the above features, a complete feature set is formed, which includes 16 original features, 4 upstream time lag features, 8 ecological mechanism features, 6 environmental stress features, 6 time cycle features, and 69 time-series derived features, for a total of 109 features.

[0152] S160: Calculate the importance of each feature in the full feature set based on the time series prediction surrogate model, recursively eliminate the feature with the lowest importance, until the number of features in the full feature set is reduced to a preset value, and obtain the optimal feature subset;

[0153] In the specific implementation of this specification, the recursive feature elimination process is as follows:

[0154] Initial number of features: 109

[0155] Number of target features: 30

[0156] Elimination strategy: When the number of remaining features > 100, eliminate 10% in each round;

[0157] When the number of remaining features is between 50 and 100, 5 features are eliminated each round;

[0158] When the number of remaining features is less than 50, one feature is eliminated each round;

[0159] After recursive elimination, the final 30 selected features include:

[0160] Upstream time delay characteristics (4): All retained

[0161] Ecological mechanism characteristics (5): NP ratio redfield deviation nutrient limitation NP product N availability

[0162] Environmental stress characteristics (3): temp stress pH stress DO stress

[0163] Time periodicity features (0)

[0164] Time-derived features (18): first-order difference of permanganate index, first-order difference of turbidity, first-order difference of dissolved oxygen, first-order difference of pH, first-order difference of water temperature, first-order difference of chlorophyll a, average of chlorophyll a over the past 72 hours, average of chlorophyll a over the past 24 hours, average of chlorophyll a over the past 12 hours, average of chlorophyll a over the past 6 hours, average of total phosphorus over the past 72 hours, average of water temperature over the past 72 hours, average of air temperature over the past 72 hours, average of air temperature over the past 24 hours, average of air temperature over the past 6 hours, 12-hour moving weighted average of air temperature, 12-hour moving weighted average of chlorophyll a, 48-hour moving weighted average of chlorophyll a.

[0165] Optionally, the calculation of the importance of each feature in the full feature set based on the time series prediction proxy model includes:

[0166]

[0167] in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. val For the validation set data, the expected value E is the sample mean.

[0168] In the specific implementation of this specification, the time series prediction proxy model is a lightweight proxy model, and its architecture includes: (a) intermediate branch: temporal convolutional layer + LSTM layer; (b) long-term branch: pooling layer + temporal convolutional layer + LSTM layer; (c) fusion layer: after feature fusion of the outputs of the two branches, the prediction result is generated through a fully connected layer.

[0169] S170: A water chlorophyll a prediction model is obtained by training based on the optimal feature subset.

[0170] Optionally, S170 includes:

[0171] The water chlorophyll a prediction model is a two-branch parallel neural network model, specifically:

[0172] The intermediate processing branch is used to process the raw time series to capture short-term change patterns;

[0173] The long-term processing branch is used to process downsampled time series to capture macro long-term trends;

[0174] The intelligent fusion layer is used to integrate the outputs of the intermediate processing branch and the long-term processing branch.

[0175] In the specific implementation of this specification, the dual-branch parallel neural network model includes a multi-scale TCN-LSTM prediction model, the training and evaluation process of which includes: (1) constructing a target prediction model, using a dual-branch parallel TCN-LSTM architecture, such as Figure 4 As shown, it includes: mid-term processing branch, long-term processing branch, intelligent fusion layer, and output mapping layer; (2) Use the selected feature subset to train the model, and adopt the mean square error loss function and Adam optimizer; (3) Use the validation set to evaluate the model performance and adjust the hyperparameters; (4) Use the test set to evaluate the final performance, calculate at least one of R², MAE, and RMSE, and determine the optimal model when the selected indicator meets the preset threshold. The threshold is determined according to the water body functional zoning and management requirements.

[0176] The architecture of the lightweight proxy model includes:

[0177] (a) Intermediate branch: Temporal convolutional layer + LSTM layer;

[0178] (b) Long-term branch: pooling layer + temporal convolutional layer + LSTM layer;

[0179] (c) Fusion layer: After feature fusion of the two branch outputs, the prediction result is generated through a fully connected layer.

[0180] The specific model training was divided into training, validation, and test sets at 70%, 15%, and 15% respectively. Mean squared error (MSE) was used as the loss function, and the Adam optimizer was employed. The initial learning rate was 5e-4, the batch size was 16, the maximum number of epochs was 200, and an early stopping strategy was implemented: training stopped when the MAE on the validation set failed to improve for 20 consecutive epochs; learning rate decay was applied: the learning rate was multiplied by 0.5 after 10 consecutive epochs without improvement. The training process stopped early at epoch 87, with a final learning rate of 6.25e-5.

[0181] Specific model evaluation, performance on the test set of the model predicting chlorophyll a concentration for the next 24 hours:

[0182] R²=0.60

[0183] MAE = 1.76 μg / L

[0184] RMSE = 2.96 μg / L

[0185] The results of the comparative experiment are shown in Table 1:

[0186] Table 1 Comparative Experimental Results of Different Training Methods

[0187]

[0188] Prediction process curves of different methods are as follows Figure 5 As shown.

[0189] The above results demonstrate that the method of the present invention achieves good results. In practical applications, the parameters need to be adjusted according to the specific characteristics of the water body.

[0190] This example illustrates the adaptive update mechanism after model deployment.

[0191] After model deployment, the system continuously monitors prediction performance. A sliding window of N=10 (10 prediction periods) is set, and the response threshold MAE is determined based on historical data trends. th It is 2.5 μg / L.

[0192] In March 2024, due to changes in hydrological conditions caused by new upstream construction projects, the MAE (Mean Energy Emission Level) for 10 consecutive cycles... 10 Reaching 2.62 μg / L, exceeding the threshold, triggering adaptive updates:

[0193] Parameter fine-tuning, specifically:

[0194] Fix the first 80% of the parameters of the convolutional layer and the LSTM layer

[0195] Fine-tuning only the parameters of the fusion layer and the output layer

[0196] Using data from the last 1000 hours

[0197] The learning rate is set to 5e-5 (1 / 10 of the original).

[0198] Fine-tuning over 15 epochs

[0199] Result: MAE 10 The concentration was reduced to 2.35 μg / L, which meets the requirements.

[0200] If the requirements are still not met after fine-tuning the parameters, feature reselection is performed; if the requirements are still not met after feature reselection, full retraining is performed.

[0201] This example illustrates the practical application of the four-level early warning mechanism.

[0202] Early warning level classification criteria:

[0203] Blue early warning: 10 < Chl-a ≤ 15 μg / L, lasting for 12 - 24 hours

[0204] Yellow early warning: 15 < Chl-a ≤ 25 μg / L, lasting for 24 - 48 hours

[0205] Orange early warning: 25 < Chl-a ≤ 40 μg / L, lasting for 48 - 72 hours

[0206] Red early warning: Chl-a > 40 μg / L, or lasting for more than 72 hours

[0207] On August 15, 2023, the system predicted that the chlorophyll a concentration would reach a peak of 23 μg / L on August 17 within the next 7 days, lasting for 36 hours, triggering a yellow early warning.

[0208] Early warning information includes:

[0209] Early warning number: YW - 20230815 - 001

[0210] Early warning level: Yellow

[0211] Predicted peak value: 23 ± 3 μg / L

[0212] Peak time: 2023 - 08 - 17 14:00

[0213] Duration: 36 hours

[0214] Suggested measures: Strengthen monitoring, prepare emergency supplies, and consider emergency water intake plans

[0215] Actual monitoring shows that the chlorophyll a concentration reached 21.5 μg / L at 16:00 on August 17, which is basically consistent with the prediction, and the early warning is effective.

[0216] The method of the present invention can also be applied to other water body systems:

[0217] (1) River system: For a river with unidirectional flow, τ max can be directly calculated based on the flow velocity and the distance of the monitoring section, and the flow gate threshold can be set relatively low.

[0218] (2) Lake system: For a large lake, the influence of multiple upstream inflows needs to be considered. The optimal time delay of each inflow can be calculated separately to construct a multi-source time delay characteristic.

[0219] (3) River network system: For a complex river network, the main upstream nodes can be selected for time delay analysis, or this method can be extended using a graph neural network.

[0220] In terms of feature engineering, adjustments can be made according to the specific characteristics of the water body:

[0221] For shallow lakes, it can increase the characteristics related to sediment release.

[0222] For reservoirs, stratified correlation features can be added.

[0223] For tidal river sections, tidal-related features can be added.

[0224] In terms of model architecture, the following variations are possible:

[0225] Use GRU instead of LSTM to reduce computational cost;

[0226] Use Transformer instead of TCN-LSTM to enhance long-range dependency modeling;

[0227] A third branch is added to handle ultra-long-term trends.

[0228] This invention accurately characterizes the pollutant transport relationship between upstream and downstream water bodies by parameterizing the spatiotemporal transport mechanism, including determining the optimal time delay and constructing flow-gated features, thus improving the timeliness of predictions. Through recursive feature elimination based on a surrogate model (SG-RFE), a lightweight model with 20-50% of the parameters of the target model is used for feature selection, improving feature selection efficiency. A dual-branch parallel TCN-LSTM architecture is employed, with the intermediate branch processing the original sequence and the long-term branch processing the downsampled sequence, achieving effective fusion of multi-scale information and improving prediction accuracy. The model adaptive update mechanism and the four-level eutrophication early warning mechanism enhance the system's practicality and reliability.

[0229] Figure 6 A schematic diagram of a water chlorophyll a prediction model generation device based on spatiotemporal transport mechanism and surrogate model provided in this specification embodiment, the device may include:

[0230] The monitoring and acquisition module 10 is used to acquire spatiotemporal data of upstream monitoring points and downstream monitoring points. The spatiotemporal data of upstream monitoring points includes upstream chlorophyll a concentration, and the spatiotemporal data of downstream monitoring points includes downstream chlorophyll a concentration.

[0231] The first calculation module 20 is used to determine the optimal time delay for transmission from upstream to downstream and the optimal time delay chlorophyll a concentration upstream, based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration.

[0232] The second calculation module 30 is used to construct a flow gating function based on a set flow threshold, and generate flow gating features by combining the upstream optimal time-delay chlorophyll a concentration.

[0233] The third calculation module 40 is used to determine the spatiotemporal coupling feature set based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics.

[0234] Module 50 is used to construct ecological mechanism features, environmental stress features, time period features, and time-series derived features, and combine them with the spatiotemporal coupling feature set to form a complete feature set;

[0235] The fourth calculation module 60 is used to calculate the importance of each feature in the full feature set based on the time series prediction surrogate model, recursively eliminate the feature with the lowest importance, until the number of features in the full feature set is reduced to a preset value, and obtain the optimal feature subset.

[0236] The model generation module 70 is used to train a water chlorophyll a prediction model based on the optimal feature subset.

[0237] Optionally, the first computing module 20 includes:

[0238] Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the Pearson correlation coefficients under different time lags were calculated.

[0239] The time delay that maximizes the absolute value of the Pearson correlation coefficient is determined as the optimal time delay.

[0240] Optionally, the calculation of the Pearson correlation coefficient under different time lags based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes:

[0241] ρ(τ)=Σ[(C up (t-τ)-μ up (C) down (t)-μ down )] / [σ up ×σ down ×n]

[0242] Where ρ(τ) is the Pearson correlation coefficient under different time delays, and C up (t-τ) represents the upstream chlorophyll a concentration under different time lags, μ up C is the mean of the upstream sequence. down (t) represents the downstream chlorophyll a concentration, μ down σ is the mean of the downstream sequence. up σ represents the standard deviation of the upstream sequence. down τ is the standard deviation of the downstream sequence, n is the number of samples, and the maximum value of τ is determined by the distance between the upstream and downstream and the average flow velocity in the slow-flow zone.

[0243] Optionally, the second computing module 30 includes:

[0244] F gated (t)=C up (t- )×G[Q(t-τ*)]

[0245] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- ) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function.

[0246] Optionally, the flow gating function includes a sigmoid smoothing function or a step function;

[0247] When the flow gating function is the Sigmoid smoothing function:

[0248]

[0249] Where k is the steepness parameter, The flow threshold that minimizes the prediction error, determined through grid search;

[0250] When the flow gating function is the step function:

[0251] G(Q) = .

[0252] Optionally, the fourth computing module 60 includes:

[0253]

[0254] in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. val For the validation set data, the expected value E is the sample mean.

[0255] Optionally, the model generation module 70 includes:

[0256] The water chlorophyll a prediction model is a two-branch parallel neural network model, specifically:

[0257] The intermediate processing branch is used to process the raw time series to capture short-term change patterns;

[0258] The long-term processing branch is used to process downsampled time series to capture macro long-term trends;

[0259] The intelligent fusion layer is used to integrate the outputs of the intermediate processing branch and the long-term processing branch.

[0260] The functions of the apparatus in this embodiment have been described in the above method embodiments. Therefore, for any parts not detailed in this embodiment, please refer to the relevant descriptions in the foregoing embodiments, which will not be repeated here.

[0261] Based on the same inventive concept, embodiments of this specification also provide an electronic device.

[0262] The following describes embodiments of the electronic device of the present invention, which can be considered as specific implementations of the methods and apparatus embodiments of the present invention described above. Details described in the embodiments of the electronic device of the present invention should be considered as supplements to the methods or apparatus embodiments described above; details not disclosed in the embodiments of the electronic device of the present invention can be implemented with reference to the methods or apparatus embodiments described above.

[0263] Figure 7 This is a schematic diagram of an electronic device provided as an embodiment of this specification. Refer to the following... Figure 7 The electronic device 300 according to this embodiment of the present invention will be described. Figure 7 The electronic device 300 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0264] like Figure 7 As shown, the electronic device 300 is presented in the form of a general-purpose computing device. The components of the electronic device 300 may include, but are not limited to: at least one processing unit 310, at least one storage unit 320, a bus 330 connecting different system components (including storage unit 320 and processing unit 310), a display unit 340, etc.

[0265] The storage unit stores program code that can be executed by the processing unit 310, causing the processing unit 310 to perform the steps described in the processing method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 310 can perform, for example... Figure 1 The steps are shown.

[0266] The storage unit 320 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 3201 and / or a cache storage unit 3202, and may further include a read-only memory unit (ROM) 3203.

[0267] The storage unit 320 may also include a program / utility 3204 having a set (at least one) program module 3205, such program module 3205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0268] Bus 330 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0269] Electronic device 300 can also communicate with one or more external devices 400 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable viewers to interact with electronic device 300, and / or with any device that enables electronic device 300 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 350. Furthermore, electronic device 300 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 360. Network adapter 360 can communicate with other modules of electronic device 300 via bus 330. It should be understood that, although... Figure 7 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 300, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0270] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described in this invention can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this invention can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the method described above according to this invention. When the computer program is executed by a data processing device, it enables the computer-readable medium to implement the method described above, i.e.: as... Figure 1 The method shown.

[0271] Figure 8 This is a schematic diagram of a computer-readable medium provided for embodiments of this specification.

[0272] accomplish Figure 1The computer program of the method shown can be stored on one or more computer-readable media. A computer-readable medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0273] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0274] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the audience's computing device, partially on the audience's device, as a standalone software package, partially on the audience's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the audience's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0275] In summary, the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that in practice, general-purpose data processing devices such as microprocessors or digital signal processors (DSPs) can be used to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0276] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the present invention is not inherently related to any specific computer, virtual device, or electronic device, and various general-purpose devices can also implement the present invention. The above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0277] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0278] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanisms and surrogate models, characterized in that, include: Spatiotemporal data of upstream monitoring points and downstream monitoring points are acquired. The spatiotemporal data of upstream monitoring points includes upstream chlorophyll a concentration, and the spatiotemporal data of downstream monitoring points includes downstream chlorophyll a concentration. Based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration, the optimal time delay for transport from upstream to downstream and the optimal upstream time delay chlorophyll a concentration are determined; wherein, determining the optimal time delay for transport from upstream to downstream based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes: calculating the Pearson correlation coefficient under different time delays based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration; the time delay that makes the absolute value of the Pearson correlation coefficient reach its maximum value is determined as the optimal time delay; A flow gating function is constructed based on a set flow threshold, and flow gating features are generated by combining the upstream optimal time-delay chlorophyll a concentration, including: F gated (t)=C up (t- )×G[Q(t- )] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- G(Q) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function. Based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics, a spatiotemporal coupling feature set is determined; Ecological mechanism features, environmental stress features, time cycle features, and time-series derived features are constructed, and combined with the aforementioned spatiotemporal coupling feature set to form a complete feature set; The importance of each feature in the full feature set is calculated based on a time series prediction surrogate model. The features with the lowest importance are recursively eliminated until the number of features in the full feature set is reduced to a preset value, thus obtaining the optimal feature subset. The calculation of the importance of each feature in the full feature set based on the time series prediction surrogate model includes: in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. val For the validation set data, the expected value E is the sample mean; A water chlorophyll a prediction model is obtained by training based on the optimal feature subset.

2. The method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanism and surrogate model as described in claim 1, characterized in that, The calculation of Pearson correlation coefficients under different time lags based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes: ρ(τ)=Σ[(C up (t-t)-m up )(C down (t)-m down )] / [s up ×s down ×n] Where ρ(τ) is the Pearson correlation coefficient under different time delays, and C up (t-τ) represents the upstream chlorophyll a concentration under different time lags, μ up C is the mean of the upstream sequence. down (t) represents the downstream chlorophyll a concentration, μ down σ is the mean of the downstream sequence. up σ represents the standard deviation of the upstream sequence. down τ is the standard deviation of the downstream sequence, n is the number of samples, and the maximum value of τ is determined by the distance between the upstream and downstream and the average flow velocity in the slow-flow zone.

3. The method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanism and surrogate model as described in claim 2, characterized in that, The flow gating function includes a Sigmoid smoothing function or a step function; When the flow gating function is the Sigmoid smoothing function: Where k is the steepness parameter, The flow threshold that minimizes the prediction error, determined through grid search; When the flow gating function is the step function: G(Q)= 。 4. The method for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanism and surrogate model as described in claim 1, characterized in that, The water chlorophyll a prediction model trained based on the optimal feature subset includes: The water chlorophyll a prediction model is a two-branch parallel neural network model, specifically: The intermediate processing branch is used to process the raw time series to capture short-term change patterns; The long-term processing branch is used to process downsampled time series to capture macro long-term trends; The intelligent fusion layer is used to integrate the outputs of the intermediate processing branch and the long-term processing branch.

5. A device for generating a water chlorophyll a prediction model based on spatiotemporal transport mechanism and surrogate model, characterized in that, include: The monitoring and acquisition module is used to acquire spatiotemporal data from upstream monitoring points and downstream monitoring points. The spatiotemporal data from upstream monitoring points includes upstream chlorophyll a concentration, and the spatiotemporal data from downstream monitoring points includes downstream chlorophyll a concentration. The first calculation module is used to determine the optimal time delay for upstream-to-downstream transmission and the optimal upstream time delay chlorophyll a concentration based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration; wherein, determining the optimal time delay for upstream-to-downstream transmission based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration includes: calculating the Pearson correlation coefficient under different time delays based on the upstream chlorophyll a concentration and the downstream chlorophyll a concentration; and determining the time delay that maximizes the absolute value of the Pearson correlation coefficient as the optimal time delay; The second calculation module is used to construct a flow gating function based on a set flow threshold, and, in conjunction with the upstream optimal time-delay chlorophyll a concentration, generate flow gating features, including: F gated (t)=C up (t- )×G[Q(t- )] Among them, F gated (t) represents the flow gating characteristic, C up (t- ) represents the upstream optimal time-delay chlorophyll a concentration, Q(t- G(Q) is the set flow threshold under the optimal time delay, and G(Q) is the flow gating function. The third calculation module is used to determine the spatiotemporal coupling feature set based on the optimal time delay and the upstream optimal time delay chlorophyll a concentration and flow gating characteristics. The construction module is used to construct ecological mechanism features, environmental stress features, time period features, and time-series derived features, which, combined with the spatiotemporal coupling feature set, form a complete feature set; The fourth calculation module is used to calculate the importance of each feature in the full feature set based on the time series prediction surrogate model, recursively eliminate the features with the lowest importance, until the number of features in the full feature set is reduced to a preset value, thus obtaining the optimal feature subset; wherein, the calculation of the importance of each feature in the full feature set based on the time series prediction surrogate model includes: in, Let L be the importance of each feature in the full feature set, L be the mean squared error loss function, M be the surrogate model, and D be the value of each feature. val For the validation set data, the expected value E is the sample mean; The model generation module is used to train a water chlorophyll a prediction model based on the optimal feature subset.

6. An electronic device, wherein, The electronic device includes: A processor; and a memory storing computer-executable instructions, which, when executed, cause the processor to perform the method according to any one of claims 1-4.

7. A computer-readable storage medium, wherein, The computer-readable storage medium stores one or more programs that, when executed by a processor, implement the method of any one of claims 1-4.

Images