A watershed pollution monitoring method based on watershed non-point source pollution monitoring model

Abstract

The present invention discloses a method for monitoring watershed pollution based on a watershed non-point source pollution monitoring model, and belongs to the technical field of watershed pollution monitoring. The present invention first collects historical pollution values ​​of the watershed to construct a historical pollution sequence, and further obtains a historical pollution increment sequence. According to the historical pollution increment sequence and the historical runoff, the correlation and flow impact ratio are obtained and a corresponding sequence is constructed, and the pollution attenuation ratio is calculated at the same time. By collecting the real-time pollution increments of each upstream watershed, a reference increment sequence of the watershed to be monitored is obtained in combination with the pollution attenuation ratio. Finally, the real-time pollution increment is processed using the pollution monitoring model, and the features are enhanced with the help of the correlation and impact ratio sequences. Combined with the reference increment sequence, the predicted pollution increment of the watershed to be monitored is obtained. The present invention solves the problem of low accuracy in predicting pollution increments in the watershed to be monitored in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of watershed pollution monitoring, and in particular to a watershed pollution monitoring method based on a watershed non-point source pollution monitoring model. Background Art

[0002] With the acceleration of industrialization and urbanization, the problem of water pollution in river basins has become increasingly serious, among which non-point source pollution has become an important factor affecting the quality of water environment. In the means of river basin pollution monitoring, the numerical simulation method of water quality model is often used. This method accurately simulates the transport and diffusion process of pollutants in rivers by constructing hydrodynamic models and water quality models. However, in the actual application process, this method has encountered many insurmountable obstacles. It is extremely difficult to accurately determine the various parameters required by the model. For example, the roughness parameter of the river channel is affected by multiple factors such as the bottom topography of the river channel, the material of the riverbed, and the long-term scouring of the water flow. The roughness may vary significantly in different river basins or even in different sections of the same river basin, making it difficult to accurately measure. Another example is the degradation rate parameter of the pollutant, which not only depends on the chemical properties of the pollutant itself, but also is closely related to the temperature, dissolved oxygen content, and microbial community structure of the water body. These environmental factors are in dynamic change all the time, making the value of the degradation rate parameter full of uncertainty. In addition, the upstream basin and the downstream basin are far apart, and the transport and diffusion path is long, which is difficult to simulate through the numerical value of the water quality model.

[0003] When faced with such complex non-point source pollution inputs in the upstream basin, the numerical simulation results based on the water quality model often deviate greatly from the actual situation. This deviation makes it extremely difficult to effectively predict the pollution increment in the downstream basin to be monitored through the model. Inaccurate simulation results cannot provide a reliable decision-making basis for basin pollution control and prevention, resulting in the possibility of missing the best pollution control opportunity due to insufficient estimation of downstream pollution increments, or wasting resources due to overestimation in actual work, which seriously restricts the efficient implementation of basin water environment protection and governance work, and is difficult to meet the urgent needs of maintaining the health of the basin ecosystem and sustainable development. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a watershed pollution monitoring method based on a watershed non-point source pollution monitoring model, which solves the problem of low accuracy in predicting the pollution increment in the watershed to be monitored in the prior art.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a watershed pollution monitoring method based on a watershed non-point source pollution monitoring model, comprising:

[0006] Collect historical pollution values ​​of upstream basins and the basin to be monitored, and construct historical pollution sequences of each basin;

[0007] Based on the historical pollution sequence of each river basin, the historical pollution increment sequence of each river basin is constructed;

[0008] According to the historical pollution increment sequence and historical runoff of each basin, the correlation and flow impact ratio of the upstream basin are obtained, and the correlation sequence and impact ratio sequence are constructed;

[0009] Calculate the pollution attenuation ratio of each upstream basin based on the difference between the historical total pollution increment of the upstream basin and the historical pollution increment of the basin to be monitored;

[0010] Collect the real-time pollution increment of each upstream basin, and obtain the reference increment sequence of the basin to be monitored based on the pollution attenuation ratio of each upstream basin;

[0011] The pollution monitoring model is used to process the real-time pollution increment of each upstream basin. The characteristics are enhanced through the correlation sequence and the impact ratio sequence. Combined with the reference increment sequence, the predicted pollution increment of the basin to be monitored is obtained.

[0012] Furthermore, the process of constructing the historical pollution increment sequence specifically includes:

[0013] Subtract the historical pollution values ​​at adjacent moments in the historical pollution sequence of the upstream basin to obtain the historical pollution increment, and construct each historical pollution increment into the historical pollution increment sequence of the upstream basin;

[0014] Subtract the pollution values ​​at adjacent moments in the historical pollution sequence of the monitored basin to obtain the historical pollution increment, and construct each historical pollution increment into a historical pollution increment sequence of the monitored basin.

[0015] Furthermore, the specific process of obtaining the relevance of the upstream basin includes:

[0016] Calculate the correlation between the historical pollution increment sequence of each upstream basin and the historical pollution increment sequence of the basin to be monitored;

[0017] The ratio of the historical runoff of each upstream basin to the historical runoff of the basin to be monitored is taken as the flow impact ratio;

[0018] The correlation degree of each upstream basin is taken as an element to construct a correlation degree sequence;

[0019] The flow impact ratio of each upstream basin is used as an element to construct an impact ratio sequence.

[0020] Furthermore, the formula for calculating the correlation is: , where r is the correlation, z up,i is the i-th historical pollution increment in the historical pollution increment sequence of the upstream basin, is the mean of the historical pollution increment sequence of the upstream basin, z s,iis the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, is the mean of the historical pollution increment sequence of the basin to be monitored, I is the sequence length, and i is a positive integer.

[0021] Furthermore, the specific process of calculating the pollution attenuation ratio of each upstream basin includes:

[0022] Add up the historical pollution increments of each upstream basin at the same time to obtain the total historical pollution increment of each upstream basin;

[0023] The total pollution attenuation coefficient is calculated based on the difference between the historical total pollution increment of the upstream and the historical pollution increment of the basin to be monitored:

[0024] At the same time, the ratio of the historical pollution increment in the upstream basin to the historical total pollution increment is taken as the incremental contribution value;

[0025] Multiply each incremental contribution value by the total pollution attenuation coefficient to obtain the pollution attenuation ratio;

[0026] The pollution attenuation ratio of each upstream basin at each moment is taken as an element to construct the pollution attenuation sequence of the upstream basin.

[0027] Furthermore, the formula for calculating the total pollution attenuation coefficient is: , where θ i is the ith total pollution attenuation coefficient, G up,i is the total historical pollution increment of the i-th upstream, z s,i It is the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, and i is a positive integer.

[0028] Furthermore, the specific process of obtaining the reference increment sequence of the river basin to be monitored includes: multiplying the real-time pollution increment of the upstream river basin by the characteristic value in the pollution attenuation sequence of the corresponding river basin to obtain the attenuation amount; adding the attenuation amounts of all upstream river basins for the same characteristic value to obtain the total attenuation amount; adding the real-time pollution increments of each upstream river basin to obtain the real-time total pollution increment, and subtracting the total attenuation amount from the real-time total pollution increment to obtain the reference increment of the river basin to be monitored; using each reference increment of the river basin to be monitored as an element to construct a reference increment sequence, wherein the characteristic value is the maximum value, minimum value or mean value in the pollution attenuation sequence.

[0029] Furthermore, the pollution monitoring model includes: a first convolutional layer, a second convolutional layer, a third convolutional layer, a first multiplier C1, a second multiplier C2, a first LSTM layer, a second LSTM layer, a first Concat layer, a second Concat layer, a first fully connected layer, a second fully connected layer and a weighted layer.

[0030] Furthermore, the input end of the first convolutional layer is used to input the correlation sequence;

[0031] The input end of the second convolutional layer is used to input the real-time pollution increment sequence, where the elements in the real-time pollution increment sequence are the real-time pollution increments of each upstream basin;

[0032] The input end of the third convolutional layer is used to input the influence ratio sequence;

[0033] A first input terminal of the first multiplier C1 is connected to the output terminal of the first convolutional layer, a second input terminal thereof is connected to the output terminal of the second convolutional layer, and an output terminal thereof is connected to the input terminal of the first LSTM layer;

[0034] A first input terminal of the second multiplier C2 is connected to the output terminal of the second convolutional layer, a second input terminal thereof is connected to the output terminal of the third convolutional layer, and an output terminal thereof is connected to the input terminal of the second LSTM layer;

[0035] The first input end of the first Concat layer is connected to the output end of the first LSTM layer, the second input end of the first Concat layer is used to input the reference increment sequence, and the output end of the first Concat layer is connected to the input end of the first fully connected layer;

[0036] The first input end of the second Concat layer is connected to the output end of the second LSTM layer, the second input end is used to input the reference increment sequence, and the output end is connected to the input end of the second fully connected layer;

[0037] The input end of the weighted layer is connected to the output end of the first fully connected layer and the output end of the second fully connected layer respectively, and the output end thereof serves as the output end of the pollution monitoring model.

[0038] Furthermore, the first convolution layer is used to perform a convolution operation on the correlation sequence to obtain a correlation feature; the second convolution layer is used to perform a convolution operation on the real-time pollution increment sequence to obtain a real-time pollution increment feature; the third convolution layer is used to perform a convolution operation on the influence ratio sequence to obtain an influence ratio feature;

[0039] The first multiplier C1 is used to multiply the correlation feature and the real-time pollution increment feature element by element to obtain the first pollution association feature; the second multiplier C2 is used to multiply the impact ratio feature and the real-time pollution increment feature element by element to obtain the second pollution association feature;

[0040] The first LSTM layer is used to extract the first deep pollution association feature from the first pollution association feature; the second LSTM layer is used to extract the second deep pollution association feature from the second pollution association feature;

[0041] The first Concat layer is used to concatenate the first deep pollution correlation feature and the reference incremental sequence to obtain a first concatenation vector; the second Concat layer is used to concatenate the second deep pollution correlation feature and the reference incremental sequence to obtain a second concatenation vector;

[0042] The first fully connected layer is used to map the first splicing vector to obtain a first predicted increment; the second fully connected layer is used to map the second splicing vector to obtain a second predicted increment; the weighted layer is used to weight the first predicted increment and the second predicted increment to obtain the predicted pollution increment of the watershed to be monitored.

[0043] The beneficial effects of the present invention are as follows: the present invention obtains the correlation and flow impact ratio of the upstream basin, reflects the follow-up characteristics of the upstream basin and the basin to be monitored in terms of pollution increment, and the proportion of the runoff in the upstream basin to the runoff in the basin to be monitored. Then, according to the difference between the historical total pollution increment of the upstream basin and the historical pollution increment of the basin to be monitored, the pollution attenuation ratio of each upstream basin is calculated to reflect the attenuation of the pollution value, thereby calculating the reference increment of the basin to be monitored, which is convenient for the pollution monitoring model to refer to and improves the accuracy of predicting pollution increment. When the pollution monitoring model processes the real-time pollution increment of each upstream basin, it combines the correlation sequence and the impact ratio sequence for feature enhancement, which can fully integrate multi-dimensional information and further improve the accuracy of predicting pollution increment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a watershed pollution monitoring method based on a watershed non-point source pollution monitoring model;

[0045] Figure 2 This is a schematic diagram of the pollution monitoring model. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, a watershed pollution monitoring method based on a watershed non-point source pollution monitoring model comprises:

[0048] Collect historical pollution values ​​of upstream basins and the basin to be monitored, and construct historical pollution sequences of each basin;

[0049] Based on the historical pollution sequence of each river basin, the historical pollution increment sequence of each river basin is constructed;

[0050] According to the historical pollution increment sequence and historical runoff of each basin, the correlation and flow impact ratio of the upstream basin are obtained, and the correlation sequence and impact ratio sequence are constructed;

[0051] Calculate the pollution attenuation ratio of each upstream basin based on the difference between the historical total pollution increment of the upstream basin and the historical pollution increment of the basin to be monitored;

[0052] Collect the real-time pollution increment of each upstream basin, and obtain the reference increment sequence of the basin to be monitored based on the pollution attenuation ratio of each upstream basin;

[0053] The pollution monitoring model is used to process the real-time pollution increment of each upstream basin. The characteristics are enhanced through the correlation sequence and the impact ratio sequence. Combined with the reference increment sequence, the predicted pollution increment of the basin to be monitored is obtained.

[0054] In this embodiment, the process of constructing the historical pollution increment sequence specifically includes:

[0055] Subtract the historical pollution values ​​at adjacent moments in the historical pollution sequence of the upstream basin to obtain the historical pollution increment, and construct each historical pollution increment into the historical pollution increment sequence of the upstream basin;

[0056] Subtract the pollution values ​​at adjacent moments in the historical pollution sequence of the basin to be monitored to obtain the historical pollution increment, and construct each historical pollution increment into the historical pollution increment sequence of the basin to be monitored. The formula for the historical pollution increment is: , where z t is the historical pollution increment at time t, c t is the historical pollution value at time t, c t-1 is the historical pollution value at time t-1, and t is the number of the time.

[0057] In actual river basin pollution monitoring, more attention is paid to the changes in pollution rather than just the pollution value at a certain moment. For example, when a new pollution source or sudden pollution incident appears in the upstream river basin, the abnormal increase in pollution value can be quickly detected through the pollution increment sequence, and potential pollution risks can be discovered in time so that countermeasures can be taken quickly. In addition, by obtaining pollution increments, it is possible to intuitively show how pollution accumulates or decreases over a period of time, and can clearly reflect the dynamic changes of pollution in the time dimension.

[0058] In this embodiment, the pollution value includes: pH, chemical oxygen demand, biochemical oxygen demand, ammonia nitrogen concentration, heavy metals, etc. The pollution value is one or more of the indicators.

[0059] In this embodiment, the specific process of obtaining the relevance of the upstream watershed includes:

[0060] Calculate the correlation between the historical pollution increment sequence of each upstream basin and the historical pollution increment sequence of the basin to be monitored;

[0061] The ratio of the historical runoff of each upstream basin to the historical runoff of the basin to be monitored is taken as the flow impact ratio;

[0062] The correlation degree of each upstream basin is taken as an element to construct a correlation degree sequence;

[0063] The flow impact ratio of each upstream basin is used as an element to construct an impact ratio sequence.

[0064] In the present invention, one upstream basin corresponds to one upstream basin historical pollution increment sequence.

[0065] The present invention measures the degree of correlation between pollution changes in the upstream basin and the basin to be monitored by calculating the correlation between the historical pollution increment sequences of the upstream basin and the basin to be monitored, and can obtain the degree of correlation between the pollution increments of the upstream basin and the basin to be monitored. The flow impact ratio reflects the composition structure of the water volume in the basin to be monitored. The water flows of various upstream basins will converge to the basin to be monitored. The runoff proportions of different upstream basins are different. The contribution of each upstream area to the downstream water volume can be clarified through the ratio. The present invention strengthens the corresponding features of real-time pollution increments through correlation sequences and impact ratio sequences. Through two dimensions, the corresponding features of real-time pollution increments are strengthened, and the attention of different features is improved, so that the pollution monitoring model has the ability to pay attention to features with high correlation and high flow impact ratio.

[0066] In this embodiment, the formula for calculating the correlation is: , where r is the correlation, z up,i is the i-th historical pollution increment in the historical pollution increment sequence of the upstream basin, is the mean of the historical pollution increment sequence of the upstream basin, z s,i is the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, is the mean of the historical pollution increment sequence of the basin to be monitored, I is the sequence length, and i is a positive integer.

[0067] In this embodiment, the specific process of calculating the pollution attenuation ratio of each upstream river basin includes:

[0068] Add up the historical pollution increments of each upstream basin at the same time to obtain the total historical pollution increment of each upstream basin;

[0069] The total pollution attenuation coefficient is calculated based on the difference between the historical total pollution increment of the upstream and the historical pollution increment of the basin to be monitored:

[0070] At the same time, the ratio of the historical pollution increment in the upstream basin to the historical total pollution increment is taken as the incremental contribution value;

[0071] Multiply each incremental contribution value by the total pollution attenuation coefficient to obtain the pollution attenuation ratio;

[0072] The pollution attenuation ratio of each upstream basin at each moment is taken as an element to construct the pollution attenuation sequence of the upstream basin.

[0073] In this embodiment, the historical pollution increment sequence of the upstream basin is: up,k,t}, the historical pollution increment sequence of the basin to be monitored is: {z s,t}, where z up,k,t is the historical pollution increment at the tth moment in the historical pollution increment sequence of the kth upstream basin, z s,t is the historical pollution increment at the tth moment in the historical pollution increment sequence of the monitored basin, and k is a positive integer. It can be seen that a historical pollution increment sequence of an upstream basin contains historical pollution increments at multiple moments. The historical pollution increments of all upstream basins at the same moment are added together to obtain the historical total pollution increment of the upstream basin at a moment.

[0074] In this embodiment, ,h k,t is the incremental contribution value of the kth upstream basin at time t, , μ k,t is the pollution attenuation ratio of the kth upstream basin at the tth moment, K is the number of upstream basins, θ t is the total pollution attenuation coefficient at time t.

[0075] The present invention calculates the total pollution attenuation coefficient at each moment according to the difference between the historical total pollution increment of the upstream and the historical pollution increment of the monitored basin, and then distributes the total pollution attenuation coefficient at each moment to each upstream basin in combination with the ratio of the historical pollution increment of each upstream basin to the historical total pollution increment, and obtains the pollution attenuation sequence of each upstream basin, which reflects the attenuation situation of the upstream basin at each moment.

[0076] In this embodiment, the formula for calculating the total pollution attenuation coefficient is: , where θ i is the ith total pollution attenuation coefficient, G up,i is the total historical pollution increment of the i-th upstream, z s,i It is the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, and i is a positive integer.

[0077] In this embodiment, the value of i is equal to t.

[0078] In this embodiment, the specific process of obtaining the reference increment sequence of the river basin to be monitored includes: multiplying the real-time pollution increment of the upstream river basin by the characteristic value in the pollution attenuation sequence of the corresponding river basin to obtain the attenuation amount; adding the attenuation amounts of all upstream river basins for the same characteristic value to obtain the total attenuation amount; adding the real-time pollution increments of each upstream river basin to obtain the real-time total pollution increment, and subtracting the total attenuation amount from the real-time total pollution increment to obtain the reference increment of the river basin to be monitored; using each reference increment of the river basin to be monitored as an element to construct a reference increment sequence, wherein the characteristic value is the maximum value, minimum value or mean value in the pollution attenuation sequence.

[0079] Multiply the real-time pollution increment of the upstream basin with the maximum, minimum or mean value in the pollution attenuation sequence of the corresponding basin to obtain three attenuation amounts respectively; add the attenuation amounts of all upstream basins belonging to the minimum value to obtain the minimum total attenuation; add the attenuation amounts of all upstream basins belonging to the maximum value to obtain the maximum total attenuation; add the attenuation amounts of all upstream basins belonging to the mean value to obtain the medium total attenuation; construct the minimum total attenuation, maximum total attenuation and medium total attenuation into a reference increment sequence.

[0080] The present invention uses historical data to calculate the pollution attenuation ratio, thereby estimating the real-time pollution increment based on the pollution attenuation ratio, obtaining a reference increment range based on previous attenuation ratio conditions, and providing more reference data for the pollution monitoring model.

[0081] The formula for the total attenuation is: , where R t is the total attenuation at time t, z real,k is the real-time pollution increment of the kth upstream basin.

[0082] In the present invention, since there are multiple attenuation total amounts, a reference increment is obtained by subtracting one attenuation total amount from the real-time total pollution increment each time. Therefore, there are multiple reference increments.

[0083] like Figure 2 As shown, the pollution monitoring model includes: a first convolutional layer, a second convolutional layer, a third convolutional layer, a first multiplier C1, a second multiplier C2, a first LSTM layer, a second LSTM layer, a first Concat layer, a second Concat layer, a first fully connected layer, a second fully connected layer and a weighted layer.

[0084] In this embodiment, the input end of the first convolution layer is used to input the correlation sequence; the input end of the second convolution layer is used to input the real-time pollution increment sequence, wherein the elements in the real-time pollution increment sequence are the real-time pollution increments of each upstream basin; the input end of the third convolution layer is used to input the impact ratio sequence;

[0085] The first input end of the first multiplier C1 is connected to the output end of the first convolutional layer, the second input end thereof is connected to the output end of the second convolutional layer, and the output end thereof is connected to the input end of the first LSTM layer; the first input end of the second multiplier C2 is connected to the output end of the second convolutional layer, the second input end thereof is connected to the output end of the third convolutional layer, and the output end thereof is connected to the input end of the second LSTM layer;

[0086] The first input end of the first Concat layer is connected to the output end of the first LSTM layer, its second input end is used to input the reference incremental sequence, and its output end is connected to the input end of the first fully connected layer; the first input end of the second Concat layer is connected to the output end of the second LSTM layer, its second input end is used to input the reference incremental sequence, and its output end is connected to the input end of the second fully connected layer; the input end of the weighted layer is respectively connected to the output end of the first fully connected layer and the output end of the second fully connected layer, and its output end serves as the output end of the pollution monitoring model.

[0087] In this embodiment, the first convolution layer is used to perform a convolution operation on the correlation sequence to obtain a correlation feature; the second convolution layer is used to perform a convolution operation on the real-time pollution increment sequence to obtain a real-time pollution increment feature; the third convolution layer is used to perform a convolution operation on the influence ratio sequence to obtain an influence ratio feature;

[0088] The first multiplier C1 is used to multiply the correlation feature and the real-time pollution increment feature element by element to obtain the first pollution association feature; the second multiplier C2 is used to multiply the impact ratio feature and the real-time pollution increment feature element by element to obtain the second pollution association feature;

[0089] The first LSTM layer is used to extract the first deep pollution association feature from the first pollution association feature; the second LSTM layer is used to extract the second deep pollution association feature from the second pollution association feature;

[0090] The first Concat layer is used to concatenate the first deep pollution correlation feature and the reference incremental sequence to obtain a first concatenation vector; the second Concat layer is used to concatenate the second deep pollution correlation feature and the reference incremental sequence to obtain a second concatenation vector;

[0091] The first fully connected layer is used to map the first splicing vector to obtain a first predicted increment; the second fully connected layer is used to map the second splicing vector to obtain a second predicted increment; the weighted layer is used to weight the first predicted increment and the second predicted increment to obtain the predicted pollution increment of the watershed to be monitored.

[0092] In the present invention, the first convolution layer, the second convolution layer, and the third convolution layer are all one-dimensional convolutions.

[0093] The present invention performs convolution operations on the correlation sequence, the real-time pollution increment sequence and the impact ratio sequence through three convolution layers respectively, realizes feature extraction of the three sequences, and then uses the first multiplier C1 to multiply the correlation feature and the real-time pollution increment feature. The correlation feature reflects the degree of correlation between the upstream basin and the basin to be monitored in terms of pollution changes, and the real-time pollution increment feature reflects the actual situation of the pollution increment at the current moment. Multiplying the two can highlight the impact of the correlation between the pollution increment of the upstream basin and the basin to be monitored on the current pollution increment, emphasize the interaction between them, and enable the model to more accurately capture the contribution of this correlation to the final pollution increment. The second multiplier C2 is then used to multiply the impact ratio feature and the real-time pollution increment feature element by element. The flow impact ratio reflects the composition structure of the water volume in the basin to be monitored. By multiplying the flow impact ratio feature reflecting the water volume composition structure with the real-time pollution increment feature, the model can take into account the different contributions of water from different sources to the pollution increment and more accurately evaluate the pollution status.

[0094] The first LSTM layer and the second LSTM layer are used to further extract deep features. The first and second Concat layers respectively concatenate the first and second deep pollution-related features with the reference increment sequence, so that the model considers each reference increment and improves the prediction accuracy. The present invention obtains the first predicted increment through the first fully connected layer, obtains the second predicted increment through the second fully connected layer, and uses the weighted layer to weight the first predicted increment and the second predicted increment to achieve the predicted pollution increment. In this embodiment, the two predicted increments can be averaged to obtain the predicted pollution increment, so that the predicted increments of the upper and lower channels are considered half each.

[0095] The present invention obtains the correlation and flow impact ratio of the upstream basin, reflecting the follow-up characteristics of the upstream basin and the basin to be monitored in terms of pollution increment, as well as the proportion of runoff in the upstream basin to the runoff in the basin to be monitored. Then, according to the difference between the historical total pollution increment of the upstream basin and the historical pollution increment of the basin to be monitored, the pollution attenuation ratio of each upstream basin is calculated to reflect the attenuation of the pollution value, thereby calculating the reference increment of the basin to be monitored, which is convenient for the pollution monitoring model to refer to and improves the accuracy of predicting pollution increment. When the pollution monitoring model processes the real-time pollution increment of each upstream basin, it combines the correlation sequence and the impact ratio sequence for feature enhancement, which can fully integrate multi-dimensional information, overcome the simulation deviation problem caused by the difficulty in determining the parameters of the water quality model and the complexity of the upstream and downstream transport and diffusion paths, and improve the prediction accuracy of the pollution increment of the basin to be monitored.

[0096] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A watershed pollution monitoring method based on a watershed non-point source pollution monitoring model, characterized in that: include: Collect historical pollution values ​​of upstream basins and the basin to be monitored, and construct historical pollution sequences of each basin; Based on the historical pollution sequence of each river basin, the historical pollution increment sequence of each river basin is constructed; According to the historical pollution increment sequence and historical runoff of each basin, the correlation and flow impact ratio of the upstream basin are obtained, and the correlation sequence and impact ratio sequence are constructed; Calculate the pollution attenuation ratio of each upstream basin based on the difference between the historical total pollution increment of the upstream basin and the historical pollution increment of the basin to be monitored; Collect the real-time pollution increment of each upstream basin, and obtain the reference increment sequence of the basin to be monitored based on the pollution attenuation ratio of each upstream basin; The pollution monitoring model is used to process the real-time pollution increment of each upstream basin, and the characteristics are enhanced through the correlation sequence and the impact ratio sequence. Combined with the reference increment sequence, the predicted pollution increment of the basin to be monitored is obtained; The specific process of calculating the pollution attenuation ratio for each upstream basin includes: Add up the historical pollution increments of each upstream basin at the same time to obtain the total historical pollution increment of each upstream basin; The total pollution attenuation coefficient is calculated based on the difference between the historical total pollution increment of the upstream and the historical pollution increment of the basin to be monitored: At the same time, the ratio of the historical pollution increment in the upstream basin to the historical total pollution increment is taken as the incremental contribution value; Multiply each incremental contribution value by the total pollution attenuation coefficient to obtain the pollution attenuation ratio; The pollution attenuation ratio of each upstream basin at each moment is used as an element to construct the pollution attenuation sequence of the upstream basin; The formula for calculating the total pollution attenuation coefficient is: , where θ i is the i-th total pollution attenuation coefficient, G up,i is the total historical pollution increment of the i-th upstream, z s,i is the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, i is a positive integer; The specific process of obtaining the reference increment sequence of the monitored basin includes: multiplying the real-time pollution increment of the upstream basin by the characteristic value in the pollution attenuation sequence of the corresponding basin to obtain the attenuation amount; adding the attenuation amounts of all upstream basins for the same characteristic value to obtain the total attenuation amount; adding the real-time pollution increments of each upstream basin to obtain the real-time total pollution increment; subtracting the total attenuation amount from the real-time total pollution increment to obtain the reference increment of the monitored basin; using each reference increment of the monitored basin as an element to construct a reference increment sequence, wherein the characteristic value is the maximum value, minimum value or mean value in the pollution attenuation sequence.

2. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 1 is characterized in that: The process of constructing the historical pollution increment sequence specifically includes: Subtract the historical pollution values ​​at adjacent moments in the historical pollution sequence of the upstream basin to obtain the historical pollution increment, and construct each historical pollution increment into the historical pollution increment sequence of the upstream basin; Subtract the pollution values ​​at adjacent moments in the historical pollution sequence of the monitored basin to obtain the historical pollution increment, and construct each historical pollution increment into a historical pollution increment sequence of the monitored basin.

3. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 1 is characterized in that: The specific process of obtaining the relevance of the upstream basin includes: Calculate the correlation between the historical pollution increment sequence of each upstream basin and the historical pollution increment sequence of the basin to be monitored; The ratio of the historical runoff of each upstream basin to the historical runoff of the basin to be monitored is taken as the flow impact ratio; The correlation degree of each upstream basin is taken as an element to construct a correlation degree sequence; The flow impact ratio of each upstream basin is used as an element to construct an impact ratio sequence.

4. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 1 is characterized in that: The formula for calculating the correlation is: , where r is the correlation, z up,i is the i-th historical pollution increment in the historical pollution increment sequence of the upstream basin, is the mean of the historical pollution increment sequence of the upstream basin, z s,i is the i-th historical pollution increment in the historical pollution increment sequence of the basin to be monitored, is the mean of the historical pollution increment sequence of the basin to be monitored, I is the sequence length, and i is a positive integer.

5. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 1 is characterized in that: The pollution monitoring model includes: a first convolution layer, a second convolution layer, a third convolution layer, a first multiplier C1, a second multiplier C2, a first LSTM layer, a second LSTM layer, a first Concat layer, a second Concat layer, a first fully connected layer, a second fully connected layer and a weighted layer.

6. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 5 is characterized in that: The input end of the first convolutional layer is used to input the correlation sequence; The input end of the second convolutional layer is used to input the real-time pollution increment sequence, where the elements in the real-time pollution increment sequence are the real-time pollution increments of each upstream basin; The input end of the third convolutional layer is used to input the influence ratio sequence; A first input terminal of the first multiplier C1 is connected to the output terminal of the first convolutional layer, a second input terminal thereof is connected to the output terminal of the second convolutional layer, and an output terminal thereof is connected to the input terminal of the first LSTM layer; A first input terminal of the second multiplier C2 is connected to the output terminal of the second convolutional layer, a second input terminal thereof is connected to the output terminal of the third convolutional layer, and an output terminal thereof is connected to the input terminal of the second LSTM layer; The first input end of the first Concat layer is connected to the output end of the first LSTM layer, the second input end of the first Concat layer is used to input the reference increment sequence, and the output end of the first Concat layer is connected to the input end of the first fully connected layer; The first input end of the second Concat layer is connected to the output end of the second LSTM layer, the second input end is used to input the reference increment sequence, and the output end is connected to the input end of the second fully connected layer; The input end of the weighted layer is connected to the output end of the first fully connected layer and the output end of the second fully connected layer respectively, and the output end thereof serves as the output end of the pollution monitoring model.

7. The watershed pollution monitoring method based on the watershed non-point source pollution monitoring model according to claim 5 is characterized in that: The first convolution layer is used to perform convolution operation on the correlation sequence to obtain the correlation feature; the second convolution layer is used to perform convolution operation on the real-time pollution increment sequence to obtain the real-time pollution increment feature; the third convolution layer is used to perform convolution operation on the influence ratio sequence to obtain the influence ratio feature; The first multiplier C1 is used to multiply the correlation feature and the real-time pollution increment feature element by element to obtain the first pollution association feature; the second multiplier C2 is used to multiply the impact ratio feature and the real-time pollution increment feature element by element to obtain the second pollution association feature; The first LSTM layer is used to extract the first deep pollution association feature from the first pollution association feature; The second LSTM layer is used to extract the second deep pollution association feature from the second pollution association feature; The first Concat layer is used to concatenate the first deep pollution correlation feature and the reference incremental sequence to obtain a first concatenation vector; the second Concat layer is used to concatenate the second deep pollution correlation feature and the reference incremental sequence to obtain a second concatenation vector; The first fully connected layer is used to map the first splicing vector to obtain a first prediction increment; the second fully connected layer is used to map the second splicing vector to obtain a second prediction increment; The weighted layer is used to perform weighted processing on the first predicted increment and the second predicted increment to obtain the predicted pollution increment of the watershed to be monitored.

Images