River Water Quality Prediction Methods

Abstract

The present application relates to the technical field of commercial information technology systems, and in particular to a method for predicting river water quality. The method for predicting river water quality includes the following steps: setting an automatic water quality station at the starting position of the river section to be predicted to obtain the starting water flow data of the river section to be predicted; dividing the river section to be predicted into multiple sub-prediction sections according to the multiple automatic water quality stations set in the river section to be predicted; using a preset dynamic grouping model, based on the starting water flow data of the river section to be predicted and the water flow data of the current multiple sub-prediction sections, predicting and grouping the multiple sub-prediction sections to obtain a prediction grouping scheme consisting of multiple prediction groups; using a preset river water quality prediction model, based on the starting water flow data of the river section to be predicted and the current water flow data of the prediction group, cascadingly predicting the water quality data of each prediction group of the prediction grouping scheme. The present application can predict river water quality more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of commercial information technology systems, and in particular to a method for predicting river water quality. Background Art

[0002] In the prior art, the prediction of river water quality is generally performed directly based on the water flow data of the upstream river section to be predicted, and the water quality data of the predicted river section is predicted as a whole.

[0003] However, there are some problems with this approach. For example, the structural characteristics of rivers formed during their natural form and artificial transformation (such as bends, narrowing or widening, etc.) can affect the speed and direction of water flow. These structural differences lead to different interception and accumulation of different types of pollutants in different sections of the river.

[0004] Therefore, the prior art still has some problems and cannot predict the water quality of the river section more accurately. Summary of the invention

[0005] In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides a river water quality prediction method, which can more accurately predict the river water quality.

[0006] The present application provides a method for predicting river water quality, which comprises the following steps:

[0007] An automatic water quality station is set up at the starting position of the river section to be predicted to obtain the starting water flow data of the river section to be predicted;

[0008] According to multiple water quality automatic stations set up in the river section to be predicted, the river section to be predicted is divided into multiple sub-prediction sections;

[0009] Through a preset dynamic grouping model, according to the initial water flow data of the river section to be predicted and the water flow data of the current multiple sub-prediction sections, multiple sub-prediction sections are predicted and grouped to obtain a prediction grouping scheme composed of multiple prediction groups;

[0010] The preset river water quality prediction model is used to predict the water quality data of each prediction group of the prediction group scheme in a cascade manner according to the initial water flow data of the river section to be predicted and the current water flow data of the prediction group.

[0011] Optionally, the preset dynamic grouping model is trained by the following steps:

[0012] The initial water flow data includes water flow velocity at the entrance of the river section to be predicted and water quality data at the entrance of the river section to be predicted;

[0013] The water flow data of the current sub-prediction section includes the water flow velocity of the current sub-prediction section and the water quality data of the current sub-prediction section;

[0014] Constructing a first CNN-LSTM neural network, wherein the first CNN-LSTM includes a first input layer, a first CNN layer, a first LSTM layer, a first fully connected layer, and a first output layer;

[0015] The first input layer is used to receive the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window;

[0016] The first CNN layer is connected to the first input layer to receive input from the first input layer, and outputs the data after data processing;

[0017] The first LSTM layer is connected to the first CNN layer to receive the output of the first CNN layer, and outputs the data after data processing;

[0018] The first fully connected layer is connected to the first LSTM layer to obtain the output of the first LSTM layer to transmit the output externally;

[0019] An input end of the first fully connected layer is connected to the first fully connected layer to obtain all features of the first fully connected layer, and an output end of the first fully connected layer is connected to the first output layer;

[0020] The first output layer is a softmax classifier, the softmax classifier has a plurality of grouping scheme identification output units, each of which corresponds to a grouping scheme, and the grouping scheme includes a plurality of grouping schemes formed by grouping sub-prediction segments and sorting them according to river flow directions;

[0021] Collecting in advance a plurality of initial water flow data and water flow data of a plurality of current sub-prediction segments of a first sliding time window and corresponding grouping schemes as training data for a dynamic grouping model;

[0022] The first CNN-LSTM neural network is used as the network framework, the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window are used as input, and the corresponding grouping schemes are used as labels to train and obtain a dynamic grouping model.

[0023] Optionally, the river section to be predicted is divided into a plurality of sub-prediction sections according to the automatic water quality stations set in the river section to be predicted, comprising the following steps:

[0024] The river section between two automatic water quality stations is divided into a sub-prediction section, and the water flow data of the sub-prediction section is obtained by averaging the water flow data at the beginning of the sub-prediction section and the water flow data at the end of the sub-prediction section.

[0025] Optionally, the river water quality prediction model is trained by the following steps:

[0026] Constructing a second CNN-LSTM neural network, wherein the second CNN-LSTM includes a second input layer, a second CNN layer, a second merging layer, a second LSTM layer, a second fully connected layer, and a second output layer, wherein each model layer sequentially connects and transfers data features until output at the second output layer;

[0027] Obtain water quality data of multiple different group schemes as training data for the second CNN-LSTM neural network;

[0028] The group scheme identifier is used as one of the inputs of the second input layer;

[0029] The initial water flow data of the group scheme is used as the second input of the second input layer, and the current water quality data of the first group is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme;

[0030] Iterate the following steps until the future water quality data for all groups in the group scheme are predicted:

[0031] The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group;

[0032] According to the water quality data of the actual grouping scheme and the water quality data of the predicted grouping scheme, the model weights of the second CNN-LSTM neural network are adjusted through a preset loss function to obtain a river water quality prediction model.

[0033] Optionally, the water quality data of each prediction group of the prediction group scheme is predicted in a cascade manner according to the initial water flow data of the river section to be predicted and the current water flow data of the prediction group through a preset river water quality prediction model, including the following steps:

[0034] The prediction group scheme identifier is used as one of the inputs of the second input layer;

[0035] The initial water flow data of the prediction group scheme is used as the second input of the second input layer, and the current water quality data of the first group of the prediction group scheme is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme;

[0036] Iterate the following steps until the future water quality data of all groups in the prediction group scheme are predicted:

[0037] The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group.

[0038] Compared with the prior art, the technical solution provided by this application has the following advantages:

[0039] One of its beneficial effects is that due to the differences in river structure and water flow velocity, different types of pollutants are intercepted and accumulated differently in different sections of the river, that is, different areas in the predicted river section may form uneven water quality data due to the interception and accumulation of pollution. This application can more accurately predict the accumulation of different pollutants in the river section by performing cascade water quality prediction after subdividing the predicted river section, thereby obtaining more accurate water quality data.

[0040] The second beneficial effect is that the present application adopts a technical solution of predicting the water quality of each group after grouping, instead of directly subdividing the river section to be predicted into multiple sub-prediction sections and directly predicting the water quality data of multiple sub-prediction sections. This is because the present application utilizes an important feature, that is, for a specific water flow velocity and pollutant content, within a fixed river shape, although the distribution breadth of its pollution will change due to the water flow velocity and the content of pollutants, due to the interception effect of the river shape, the content and type of pollutants will always show an upward or downward trend in a certain area. That is, for a specific water flow velocity and pollutant content, its change curve in the entire river section to be predicted will always form a similar fixed trend. If each subdivided segment is predicted directly, the differences between different sub-prediction sections are large, and the model needs to process the differences of these different sub-prediction sections at the same time, and cannot utilize this similar trend, which will lead to instability and decreased accuracy of the prediction.

[0041] By grouping, river sections with similar water quality characteristics and environmental conditions can be combined into one group. The benefit of this is that the water quality changes in each group of the forecast group scheme can better reflect this trend. Therefore, the cascade forecasting model can further benefit from this grouping. In the cascade process, the model uses the output from the previous group forecast as input to predict the water quality of the next group. Grouping ensures that the data trend on which each step of the forecast depends can be consistent with the generally formed fixed trend, making the overall forecast more reliable and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of the flow chart of the river water quality prediction method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical solution in this application will be described below in conjunction with the accompanying drawings.

[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application can also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, not all of the embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0045] The present application provides a method for predicting river water quality, which comprises the following steps:

[0046] S101: Setting up an automatic water quality station at the starting position of the river section to be predicted to obtain starting water flow data of the river section to be predicted.

[0047] Specifically, the initial water flow data includes the water flow velocity at the beginning of the river section to be predicted and the water quality data at the beginning of the river section to be predicted, and the water quality data is dissolved oxygen content, turbidity, algae density or permanganate index.

[0048] S102: Divide the river section to be predicted into multiple sub-prediction sections according to multiple automatic water quality stations set in the river section to be predicted.

[0049] Multiple automatic water quality stations are set up in the river section to be predicted, and the river section between two automatic water quality stations is divided into a sub-prediction section.

[0050] The water flow data of the sub-prediction segment is obtained by averaging the water flow data at the beginning of the sub-prediction segment and the water flow data at the end of the sub-prediction segment.

[0051] The water flow data of the sub-prediction segment includes the water flow velocity and the water quality data of the sub-prediction segment, and the water quality data is dissolved oxygen content, turbidity, algae density or permanganate index.

[0052] S103: by using a preset dynamic grouping model, according to the initial water flow data of the river section to be predicted and the water flow data of the current multiple sub-prediction sections, multiple sub-prediction sections are predicted and grouped to obtain a prediction grouping scheme consisting of multiple prediction groups.

[0053] Specifically, after the initial water flow data of the current river section to be predicted and the water flow data of each sub-prediction section are input into the dynamic grouping model, a prediction grouping scheme can be obtained through the dynamic grouping model prediction.

[0054] Specifically, the preset dynamic grouping model is trained through the following steps:

[0055] Constructing a first CNN-LSTM neural network, wherein the first CNN-LSTM includes a first input layer, a first CNN layer, a first LSTM layer, a first fully connected layer, and a first output layer;

[0056] The first input layer is used to receive the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window;

[0057] For example, in the embodiment of the present application, there are 3 sub-prediction segments, namely sub-prediction segment a, sub-prediction segment b, and sub-prediction segment c. The size of the first sliding time window is 4. Da0, Da1, Da2, and Da3 respectively represent the water quality data of sub-prediction segment a collected at preset time intervals and sorted from the past to the present. Sa0, Sa1, Sa2, and Sa3 respectively represent the water flow rate of sub-prediction segment a collected at preset time intervals and sorted from the past to the present.

[0058] Then the water flow data of multiple sub-prediction segments of the first sliding time window are [(Da0, Da1, Da2, Da3, Sa0, Sa1, Sa2, Sa3), (Db0, Db1, Db2, Db3, Sb0, Sb1, Sb2, Sb3), (Dc0, Dc1, Dc2, Dc3, Sc0, Sc1, Sc2, Sc3)].

[0059] In the embodiment of the present application, the water quality data of each individual sub-prediction segment of the first sliding time window is preprocessed by the ICEEMDAN decomposition method and then input into the first CNN layer.

[0060] The ICEEMDAN decomposition method is an existing technology, the full name of which is Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, and is a signal processing method based on Empirical Mode Decomposition (EMD). It helps to obtain more accurate water quality data from the original water quality data while retaining all the information of the original signal.

[0061] The first CNN layer is connected to the first input layer to receive input from the first input layer, and outputs the data after data processing;

[0062] The first LSTM layer is connected to the first CNN layer to receive the output of the first CNN layer, and outputs the data after data processing;

[0063] The first fully connected layer is connected to the first LSTM layer to obtain the output of the first LSTM layer to transmit the output externally;

[0064] An input end of the first fully connected layer is connected to the first fully connected layer to obtain all features of the first fully connected layer, and an output end of the first fully connected layer is connected to the first output layer;

[0065] The first output layer is a softmax classifier, the softmax classifier has a plurality of grouping scheme identification output units, each of which corresponds to a grouping scheme, and the grouping scheme includes a plurality of grouping schemes formed by grouping sub-prediction segments and sorting them according to river flow directions;

[0066] Collecting in advance a plurality of initial water flow data and water flow data of a plurality of current sub-prediction segments of a first sliding time window and corresponding grouping schemes as training data for a dynamic grouping model;

[0067] Specifically, the grouping scheme used for training here is a grouping scheme formed by artificially pre-grouping the sub-prediction segments according to the concentration of water quality changes in the sub-prediction segments. For example, when the subsequent changes of the water quality data of two sub-prediction segments tend to be consistent under the conditions of the water flow data of the current sub-prediction segment and the starting water flow data, they are divided into the same combination. Thus, different grouping schemes are formed according to different water flow data.

[0068] The first CNN-LSTM neural network is used as the network framework, the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window are used as input, and the identifier of the corresponding grouping scheme is used as a label to train and obtain a dynamic grouping model.

[0069] S104: using a preset river water quality prediction model, based on the initial water flow data of the river section to be predicted and the current water flow data of the prediction group, the water quality data of each prediction group of the prediction group scheme is predicted in a cascade manner.

[0070] Specifically, the following steps are included:

[0071] The identification of the prediction group scheme is used as one of the inputs of the second input layer;

[0072] The initial water flow data of the prediction group scheme is used as the second input of the second input layer, and the current water quality data of the first group of the prediction group scheme is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme;

[0073] Iterate the following steps until the future water quality data of all groups in the prediction group scheme are predicted:

[0074] The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group.

[0075] Specifically, the current water quality data of each group is composed of the water quality data of the sub-prediction segment of the group. For example, the current group is composed of sub-prediction segment a and sub-prediction segment b. The water quality data of sub-prediction segment a is Da, and the water quality data of sub-prediction segment b is Db. Then the current water quality data of the current group is [Da, Db].

[0076] The water quality data of the next group to be predicted is Dc, and the predicted value Dc is also used as the predicted water quality data of each sub-prediction segment in the next group to be predicted.

[0077] The river water quality prediction model is trained by the following steps:

[0078] Constructing a second CNN-LSTM neural network, wherein the second CNN-LSTM includes a second input layer, a second CNN layer, a second merging layer, a second LSTM layer, a second fully connected layer, and a second output layer, wherein each model layer sequentially connects and transfers data features until output at the second output layer;

[0079] Obtain water quality data of multiple different group schemes as training data for the second CNN-LSTM neural network;

[0080] The group scheme identifier is used as one of the inputs of the second input layer;

[0081] The initial water flow data of the group scheme is used as the second input of the second input layer, and the current water quality data of the first group is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme;

[0082] Iterate the following steps until the future water quality data for all groups in the group scheme are predicted:

[0083] The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group;

[0084] According to the water quality data of the actual grouping scheme and the water quality data of the predicted grouping scheme, the model weights of the second CNN-LSTM neural network are adjusted through a preset loss function to obtain a river water quality prediction model.

[0085] Specifically, the preset loss function is the MSE loss function.

[0086] In summary, the river water quality prediction method provided by the embodiment of the present application has the following beneficial effects:

[0087] One of its beneficial effects is that due to the differences in river structure and water flow velocity, different types of pollutants are intercepted and accumulated differently in different sections of the river, that is, different areas in the predicted river section may form uneven water quality data due to the interception and accumulation of pollution. This application can more accurately predict the accumulation of different pollutants in the river section by performing cascade water quality prediction after subdividing the predicted river section, thereby obtaining more accurate water quality data.

[0088] The second beneficial effect is that the present application adopts a technical solution of predicting the water quality of the groups after grouping, instead of directly subdividing the river section to be predicted into multiple sub-prediction sections and directly predicting the water quality data of multiple sub-prediction sections. This is because the present application utilizes an important feature, that is, for a specific water flow velocity and pollutant content, within a fixed river shape, although the distribution breadth of its pollution will change due to the water flow velocity and the content of pollutants, due to the interception effect of the river shape, the content and type of pollutants will always show an upward or downward trend in a certain area. That is, for a specific water flow velocity and pollutant content, its change curve in the entire river section to be predicted will always form a similar fixed trend. If each subdivided segment is predicted directly, the differences between different sub-prediction sections are large, and the model needs to process the differences of these different sub-prediction sections at the same time, and cannot utilize this similar trend, which will lead to instability and decreased accuracy of the prediction.

[0089] By grouping, river sections with similar water quality characteristics and environmental conditions can be combined into one group. The benefit of this is that the water quality changes in each group of the forecast group scheme can better reflect this trend. Therefore, the cascade forecasting model can further benefit from this grouping. In the cascade process, the model uses the output from the previous group forecast as input to predict the water quality of the next group. Grouping ensures that the data trend on which each step of the forecast depends can be consistent with the generally formed trend, making the overall forecast more reliable and accurate.

[0090] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. In addition, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements. Moreover, in the description of the embodiments of the present application, unless otherwise specified, " / " means or, for example, A / B can represent A or B; "and / or" in this article is only a kind of association relationship describing the associated objects, indicating that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. Furthermore, in the description of the embodiments of the present application, “plurality” refers to two or more than two.

[0091] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting river water quality, characterized in that: The river water quality prediction method comprises the following steps: An automatic water quality station is set up at the starting position of the river section to be predicted to obtain the starting water flow data of the river section to be predicted; According to multiple water quality automatic stations set up in the river section to be predicted, the river section to be predicted is divided into multiple sub-prediction sections; Through a preset dynamic grouping model, according to the initial water flow data of the river section to be predicted and the water flow data of multiple current sub-prediction sections, multiple sub-prediction sections are predicted and grouped to obtain a prediction grouping scheme consisting of multiple prediction groups, whose data trend can be consistent with the fixed trend; The preset river water quality prediction model is used to predict the water quality data of each prediction group of the prediction group scheme in a cascade manner according to the initial water flow data of the river section to be predicted and the current water flow data of the prediction group; The preset dynamic grouping model is trained by the following steps: The initial water flow data includes water flow velocity at the entrance of the river section to be predicted and water quality data at the entrance of the river section to be predicted; The water flow data of the current sub-prediction segment includes the water flow velocity of the current sub-prediction segment and the water quality data of the current sub-prediction segment; Constructing a first CNN-LSTM neural network, wherein the first CNN-LSTM includes a first input layer, a first CNN layer, a first LSTM layer, a first fully connected layer, and a first output layer; The first input layer is used to receive the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window; The first CNN layer is connected to the first input layer to receive input from the first input layer, and outputs the data after data processing; The first LSTM layer is connected to the first CNN layer to receive the output of the first CNN layer, and outputs the data after data processing; The first fully connected layer is connected to the first LSTM layer to obtain the output of the first LSTM layer to transmit the output externally; An input end of the first fully connected layer is connected to the first fully connected layer to obtain all features of the first fully connected layer, and an output end of the first fully connected layer is connected to the first output layer; The first output layer is a softmax classifier, the softmax classifier has a plurality of grouping scheme identification output units, each of which corresponds to a grouping scheme, and the grouping scheme includes a plurality of grouping schemes formed by grouping sub-prediction segments and sorting them according to river flow directions; Collecting in advance a plurality of initial water flow data and water flow data of a plurality of current sub-prediction segments of a first sliding time window and corresponding grouping schemes as training data for a dynamic grouping model; The first CNN-LSTM neural network is used as the network framework, the initial water flow data and the water flow data of the current multiple sub-prediction segments of the first sliding time window are used as input, and the corresponding grouping schemes are used as labels to train and obtain a dynamic grouping model.

2. The river water quality prediction method according to claim 1, characterized in that: According to the automatic water quality stations set up in the river section to be predicted, the river section to be predicted is divided into multiple sub-prediction sections, including the following steps: The river section between two automatic water quality stations is divided into a sub-prediction section, and the water flow data of the sub-prediction section is obtained by averaging the water flow data at the beginning of the sub-prediction section and the water flow data at the end of the sub-prediction section.

3. The river water quality prediction method according to claim 1, characterized in that: The river water quality prediction model is trained by the following steps: Constructing a second CNN-LSTM neural network, wherein the second CNN-LSTM includes a second input layer, a second CNN layer, a second merging layer, a second LSTM layer, a second fully connected layer, and a second output layer, wherein each model layer sequentially connects and transfers data features until output at the second output layer; Obtain water quality data of multiple different group schemes as training data for the second CNN-LSTM neural network; The group scheme identifier is used as one of the inputs of the second input layer; The initial water flow data of the group scheme is used as the second input of the second input layer, and the current water quality data of the first group is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme; Iterate the following steps until the future water quality data for all groups in the group scheme are predicted: The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group; According to the water quality data of the actual grouping scheme and the water quality data of the predicted grouping scheme, the model weights of the second CNN-LSTM neural network are adjusted through a preset loss function to obtain a river water quality prediction model.

4. The method for predicting river water quality according to claim 3, characterized in that: The preset river water quality prediction model is used to predict the water quality data of each prediction group of the prediction group scheme in a cascade manner according to the initial water flow data of the river section to be predicted and the current water flow data of the prediction group, including the following steps: The prediction group scheme identifier is used as one of the inputs of the second input layer; The initial water flow data of the prediction group scheme is used as the second input of the second input layer, and the current water quality data of the first group of the prediction group scheme is used as the third input of the second input layer to predict the future water quality data of the first group in the group scheme; Iterate the following steps until the future water quality data of all groups in the prediction group scheme are predicted: The predicted water quality data of the group is added to the original input data as new input, and the future water quality data of the next group is predicted based on the current water quality data of the next group.

Images