Water pollution source identification method and system based on data fusion

Abstract

The invention provides a water body pollution source recognition method and system based on data fusion, and relates to the technical field of water body pollution source recognizing.The method includes the steps that a plurality of monitoring points are arranged from a water body inlet to a water body outlet, historical water body pollution sources, environmental data and water quality parameters of the monitoring points are collected, and the data are preprocessed; comprising abnormal value elimination and missing value filling. The water quality parameter of each monitoring point is obtained through a sensor and then combined with environmental data to generate an environmental index, and different pollution sources are coded. And taking a feature vector formed by splicing the water quality parameters and the environmental indexes as input, constructing a pollution source identification model by using a support vector machine algorithm, and evaluating the performance of the model through training and testing. And finally, based on the output of the model, screening out monitoring points with pollution at the current moment, and continuously updating model parameters according to an incremental learning method so as to realize dynamic water body pollution source identification and monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of water pollution source identification, and in particular to a water pollution source identification method and system based on data fusion. Background Art

[0002] With the intensification of industrialization, urbanization and agricultural activities, the sources of water pollution are becoming increasingly diverse, including industrial wastewater, loss of agricultural fertilizers and pesticides, direct discharge of domestic sewage, etc. These pollutants not only change the physical and chemical properties of water bodies, but also threaten the survival and reproduction of aquatic organisms, and thus affect people's lives.

[0003] Especially in the context of accelerated industrialization and urbanization, rapid identification and real-time monitoring of water pollution sources are particularly important. Traditional water quality monitoring methods mostly rely on regular sampling and laboratory analysis, which is not only time-consuming, but also unable to reflect water quality changes in a timely manner, resulting in pollution incidents often being discovered a long time after they occur. In addition, existing technologies rely on a single data source for data processing and analysis, and lack a comprehensive assessment of multi-dimensional environmental factors, resulting in insufficient accuracy and real-time identification of pollution sources. At the same time, the traceability and security of data are not effectively guaranteed, which poses challenges to environmental management and emergency response.

[0004] The prior art has the following deficiencies:

[0005] First, traditional water quality monitoring methods usually rely on manual sampling and laboratory analysis. This method is not only time-consuming and labor-intensive, but also has spatial and temporal limitations, making it impossible to achieve timely monitoring and response. Secondly, existing technologies are also insufficient in the accuracy of pollution source identification. Traditional methods can often only identify certain specific types of pollution sources, and it is difficult to deal with complex pollution situations and multi-source pollution problems. In addition, existing technologies are relatively limited in data processing and analysis capabilities. Traditional methods can usually only process simple, low-dimensional data, and it is difficult to deal with the complex data sets involved in modern environmental monitoring. Finally, existing monitoring systems often lack adaptive capabilities and are difficult to dynamically adjust and update according to changes in the environment and pollution patterns. This static approach may make the system lose its effectiveness in long-term use.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for identifying water pollution sources based on data fusion to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for identifying water pollution sources based on data fusion, the specific steps include:

[0010] Step 1: From the entrance to the exit of the water body, mark each 500-meter interval as a monitoring point, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data include temperature, humidity and rainfall one hour before the collection time.

[0011] Step 2: Encode the water pollution sources at the monitoring points to generate pollution codes, and generate environmental indexes reflecting the environmental status at the monitoring points based on environmental data;

[0012] Step 3: Concatenate the water quality parameters and environmental indexes at the monitoring point to form a feature vector, construct and train the feature vector whose input is the target monitoring point and its adjacent monitoring points, and the output is the pollution source identification model of the pollution code at the target monitoring point, and the target monitoring point is any monitoring point;

[0013] Step 4: Obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, filter out the monitoring points with pollution and determine the water pollution source of each monitoring point at the current moment.

[0014] Furthermore, the specific logic for collecting water quality parameters and environmental data is as follows:

[0015] pH sensors, dissolved oxygen concentration sensors, chemical oxygen demand sensors, conductivity sensors, turbidity sensors, and ammonia nitrogen sensors are arranged at each monitoring point to collect water quality parameters. Environmental data are obtained through online meteorological services. The preprocessing of water quality parameters includes removing outliers, filling missing values, and normalizing data. The Z-score method is used for outlier detection. If the |Z| of the data is greater than 3, it is regarded as an outlier and is removed. After the outlier detection is completed, the corresponding missing value is filled with the mean of the same type of data from all monitoring points. The min-max normalization method is used to standardize the data to the [0,1] interval.

[0016] Furthermore, the water pollution sources include four types: industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources. The pollution code is a four-digit code, and the four codes from left to right correspond to industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources respectively. When a certain type of pollution source does not exist at the monitoring point, the code at that position in the pollution code is 0, otherwise it is 1.

[0017] Furthermore, the specific logic for calculating the environmental index is as follows:

[0018] And nonlinear transformation is performed on air temperature, temperature and rainfall. The transformation of air temperature is expressed as:

[0019] T r =ln(T+c T )

[0020] Among them, T r is the temperature change, T is the collected temperature data, c t is a constant term;

[0021] The transformation of humidity is expressed as:

[0022] H r =H 2

[0023] Among them, H r is the transformation of humidity, H is the collected humidity data;

[0024] The transformation of rainfall is expressed as:

[0025]

[0026] Among them, R r is the transformation of rainfall, R is the collected rainfall data, c R is a constant term;

[0027] After nonlinear transformation of environmental data, the min-max standardization method is performed. The standardized temperature variables, humidity variables and rainfall variables are expressed as T n , H n , R n ; Introducing the interaction terms between environmental variables, the interaction between temperature and humidity is expressed as:

[0028] I TH =T n ·H n

[0029] Among them, I TH represents the interaction term between air temperature and humidity;

[0030] Similarly, the interaction terms of temperature and rainfall and humidity and rainfall can be obtained, which are I TR ,I HR ;

[0031] The formula for generating the environmental index that reflects the environmental status at the monitoring point is:

[0032] EI=ω T ·T n +ω H ·H n +ω R ·R n +ω TH I TH +ω TR I TR +ω HR I HR

[0033] Among them, EI is the environmental index, ω t ,ω H ,ω R ,ω TH ,ω TR ,ω HR are the weights of each variable and interaction term, 0<ω R <ω H <ω T <1,0<ω HR <ω TR <ω TH <1.

[0034] Furthermore, the specific logic for building the pollution source identification model is as follows:

[0035] The water quality parameters and environmental indexes at the monitoring points are concatenated to form a feature vector, expressed as:

[0036]

[0037] Among them, F represents the characteristic vector, pH is the pH value, DO is the dissolved oxygen concentration, COD is the chemical oxygen demand, EC is the electrical conductivity, TB is the turbidity, and NH3 is the ammonia nitrogen concentration;

[0038] The formed feature vectors are integrated into a sample data set, and the sample data set is divided into a training set and a test set with a division ratio of 7:3. The feature vectors of the target monitoring point and its adjacent monitoring points are used as input, and the pollution code at the target monitoring point is used as the label. A pollution source identification model is constructed based on the support vector machine algorithm, and a hyperplane is set to separate different types of pollution sources. The optimization goal of the model is to minimize the loss function. Accuracy, recall rate and precision rate are selected as the performance indicators of the model. When the accuracy is ≥85%, the recall rate is ≥80%, and the precision rate is ≥80%, it means that the model training is completed. The feature vector is formed according to the water quality parameters and environmental data in the test set, and input into the model for testing.

[0039] Furthermore, the specific logic for determining the water pollution source of each monitoring point at the current moment is:

[0040] Based on the output of the pollution source identification model, the monitoring points with pollution codes containing 1 are screened out, and the pollution source types of the corresponding monitoring points are determined by comparing the pollution codes.

[0041] The present invention further provides a water pollution source identification system based on data fusion, wherein the water pollution source identification system based on data fusion is used to implement the above-mentioned water pollution source identification method based on data fusion, and comprises:

[0042] The historical data collection module is used to mark a monitoring point every 500 meters from the entrance to the exit of the water body, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data includes temperature, humidity and rainfall one hour before the collection time.

[0043] An environmental status assessment module, for encoding water pollution sources at monitoring points to generate pollution codes, and for generating environmental indexes reflecting environmental status at monitoring points based on environmental data;

[0044] A pollution source identification model building module is used to concatenate water quality parameters and environmental indexes at monitoring points to form feature vectors, build and train feature vectors whose input is the target monitoring point and its adjacent monitoring points, and whose output is a pollution source identification model with pollution codes at the target monitoring point, and the target monitoring point is any monitoring point;

[0045] The pollution source identification and model updating module is used to obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, the monitoring points with pollution are screened out and the water pollution sources of each monitoring point at the current moment are determined.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0047] The combination of data fusion and machine learning algorithms not only improves the accuracy of pollution source identification, but also effectively processes complex data sets. Through multi-angle data fusion of water quality parameters and environmental data, feature extraction and model training, the system can identify various types of pollution sources, including industrial, agricultural, domestic and natural pollution. This multi-dimensional analysis capability enables the method to adapt to different water environments and pollution conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0049] Figure 2 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0051] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] Example:

[0053] See also Figure 1 , the present invention provides a technical solution:

[0054] A method for identifying water pollution sources based on data fusion, the specific steps include:

[0055] Step 1: From the entrance to the exit of the water body, mark each 500-meter interval as a monitoring point, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data include temperature, humidity and rainfall one hour before the collection time.

[0056] In this embodiment, the specific logic for collecting water quality parameters and environmental data is as follows:

[0057] pH sensors, dissolved oxygen concentration sensors, chemical oxygen demand sensors, conductivity sensors, turbidity sensors, and ammonia nitrogen sensors are arranged at each monitoring point to collect water quality parameters. Environmental data are obtained through online meteorological services. The preprocessing of water quality parameters includes removing outliers, filling missing values, and normalizing data. The Z-score method is used for outlier detection. If the |Z| of the data is greater than 3, it is regarded as an outlier and is removed. After the outlier detection is completed, the corresponding missing value is filled with the mean of the same type of data from all monitoring points. The min-max normalization method is used to standardize the data to the [0,1] interval.

[0058] By setting up monitoring points every 500 meters from the entrance to the exit of the water body, comprehensive and continuous monitoring of the water pollution status can be achieved, ensuring the spatial coverage and representativeness of data collection, which is crucial for identifying pollution sources. Secondly, the use of multiple water quality parameters (such as pH, dissolved oxygen, chemical oxygen demand, etc.) and environmental data (such as temperature, humidity, rainfall) for data collection can fully reflect the chemical and physical state of the water body, making the identification of pollution sources more accurate and scientifically based. In terms of data preprocessing, the Z-score method is used for outlier detection, which effectively improves the data quality and ensures the credibility of the analysis results. The operations of eliminating outliers and filling missing values ​​ensure the integrity of the data and reduce potential misjudgments caused by inaccurate data. In addition, the min-max normalization method is used to normalize the data to the [0,1] interval, eliminating the impact of different dimensions and making the features more comparable, which is crucial for subsequent data fusion and model training.

[0059] Step 2: Encode the water pollution sources at the monitoring points to generate pollution codes, and generate environmental indexes reflecting the environmental status at the monitoring points based on environmental data;

[0060] In this embodiment, the water pollution sources include four types: industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources. The pollution code is a four-digit code, and the four codes from left to right correspond to the industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources respectively. When a certain type of pollution source does not exist at the monitoring point, the code at that position in the pollution code is 0, otherwise it is 1; the industrial pollution source code is 1000, the agricultural pollution source code is 0100, the domestic pollution source code is 0010, and the natural pollution source code is 0001. For example, when there are industrial and agricultural pollution sources at the same time, and there are no domestic and natural pollution sources, the pollution code is 1100.

[0061] In this embodiment, the specific logic for calculating the environment index is:

[0062] And nonlinear transformation is performed on air temperature, temperature and rainfall. The transformation of air temperature is expressed as:

[0063] T r =ln(T+c T )

[0064] Among them, T r is the temperature change, T is the collected temperature data, c T is a constant term, set to 1 to prevent the input of the logarithmic function from being zero or negative;

[0065] The transformation of humidity is expressed as:

[0066] H r =T 2

[0067] Among them, H r is the transformation of humidity, H is the collected humidity data;

[0068] The transformation of rainfall is expressed as:

[0069]

[0070] Among them, R r is the transformation of rainfall, R is the collected rainfall data, c R is a constant term, which is set to 1 to avoid negative values ​​when taking square roots;

[0071] After nonlinear transformation of environmental data, the min-max standardization method is performed. The standardized temperature variables, humidity variables and rainfall variables are expressed as T n , H n , R n ; Introducing the interaction terms between environmental variables, the interaction between temperature and humidity is expressed as:

[0072] ITH =T n ·H n

[0073] Among them, I Th represents the interaction term between air temperature and humidity;

[0074] The interaction between temperature and rainfall is expressed as:

[0075] I TR =T n ·R n

[0076] Among them, I TR represents the interaction term between temperature and rainfall;

[0077] The interaction between humidity and rainfall is expressed as:

[0078] I HR =H n ·R n

[0079] Among them, I HR represents the interaction term between humidity and rainfall;

[0080] The formula for generating the environmental index that reflects the environmental status at the monitoring point is:

[0081] EI=ω T ·T n +ω H ·H n +ω R ·R n +ω TH I TH +ω TR I TR +ω HR I HR

[0082] Among them, EI is the environmental index, ω t ,ω H ,ω R ,ω TH ,ω TR ,ω HR are the weights of each variable and interaction term, 0<ω R <ω H <ω T <1,0<ω HR <ω TR <ω TH <1.

[0083] EI takes into account multiple environmental factors such as temperature, humidity and rainfall, so that the environmental status of the monitoring point can be reflected by a simple index. The larger the EI, the better the environmental status, which may mean suitable climate conditions, higher ecological adaptability, or less pollution impact. Suitable temperature helps to maintain ecological balance. In the range of 10℃-30℃, the higher the temperature, the better the climate conditions, which helps to maintain ecological balance, and the larger the environmental index. Relatively moderate humidity helps to maintain ecological balance. In the range of 50%-80%, the higher the humidity, the better the climate conditions, which helps to maintain ecological balance, and the larger the environmental index. Moderate rainfall contributes to the health of the ecosystem. In the range of 500mm-1500mm per year, the higher the rainfall, the better the climate conditions, which helps to maintain ecological balance, and the higher the environmental index. In actual environments, the relationship between temperature, humidity and rainfall is usually nonlinear. For example, an increase in temperature may lead to accelerated evaporation, which in turn affects humidity. The interaction between humidity and rainfall may affect water evaporation and loss, thereby affecting water quality. This nonlinear relationship cannot be captured by simple linear weighting. The relationship between temperature and water quality or ecological health is usually nonlinear. Within a moderate temperature range, the growth and metabolic rate of aquatic organisms may increase with increasing temperature. Therefore, using a logarithmic transformation can better capture this relationship. There are boundary values ​​in the range of temperature changes, which will have a greater impact on the analysis results. Logarithmic transformation can reduce the impact of these boundary values. When humidity increases within a certain range, the square of the humidity value (H 2 ) can effectively enhance the positive impact of humidity on ecology or water quality. In the range of 50%-80%, an increase in humidity usually promotes plant growth and the ecological health of water bodies. Moderate humidity is essential for plant photosynthesis and growth. The square transformation can reflect the nonlinear response of organisms to changes in humidity. Within the humidity range, the effect of increasing humidity will be more significant than a simple linear relationship, and the square transformation can emphasize the positive feedback effect under suitable temperature conditions. Precipitation data often presents a right-skewed distribution, especially in arid areas. The square root transformation can reduce the impact of boundary value rainfall on the model, making the data distribution more symmetrical, thereby improving the stability of the model. The impact of rainfall on the ecosystem is usually nonlinear. Appropriate rainfall helps plant growth, but excessive rainfall may cause floods and soil erosion. In the range of 500mm-1500mm, an increase in rainfall will also increase the environmental index. The square root transformation can help the model better capture this marginal effect, so that the model can reflect the importance of moderate rainfall. By introducing interaction terms (such as I TH ,I TR ,I HR), can better capture these complex interactions and improve the effectiveness of environmental indices. Taking multiple variables into consideration, a single environmental index is formed by weighting, making it easier for managers and decision makers to use. This index integrates multiple environmental factors into an indicator that is easy to understand and apply, making it easier to identify water pollution sources.

[0084] In many aquatic ecosystems, temperature often has a significant effect on biological activity, dissolved oxygen content, and chemical reaction rates, so temperature is usually given a higher weight. Humidity has an important impact on water evaporation, precipitation, and ecosystem function, so it is second only to temperature. Although precipitation has an important impact on water quality, especially in controlling runoff and water dilution, its impact may be relatively small in some cases (such as arid areas). Therefore, the weight of precipitation is usually lower than that of temperature and humidity. The interaction between temperature and humidity is very important in many ecological and water quality models, especially in climate change and extreme weather conditions. Therefore, the interaction terms are usually given higher weights. The interaction between temperature and rainfall is also important, especially in areas with alternating dry and humid climates, so its weight is slightly lower than the interaction term between temperature and humidity. Although the interaction between humidity and rainfall is also important, its impact may be relatively small in many cases, so it is usually given the lowest weight.

[0085] Step 3: Concatenate the water quality parameters and environmental indexes at the monitoring point to form a feature vector, construct and train the feature vector whose input is the target monitoring point and its adjacent monitoring points, and the output is the pollution source identification model of the pollution code at the target monitoring point, and the target monitoring point is any monitoring point;

[0086] In this embodiment, the specific logic for constructing the pollution source identification model is:

[0087] The water quality parameters and environmental indexes at the monitoring points are concatenated to form a feature vector, expressed as:

[0088]

[0089] Among them, F represents the characteristic vector, pH is the pH value, DO is the dissolved oxygen concentration, COD is the chemical oxygen demand, EC is the electrical conductivity, TB is the turbidity, and NH3 is the ammonia nitrogen concentration;

[0090] The formed feature vectors are integrated into a sample data set, and the sample data set is divided into a training set and a test set with a division ratio of 7:3. The feature vectors of the target monitoring point and its adjacent monitoring points are used as input, and the pollution code at the target monitoring point is used as the label. A pollution source identification model is constructed based on the support vector machine algorithm, and a hyperplane is set to separate different types of pollution sources. The optimization goal of the model is to minimize the loss function. Accuracy, recall rate and precision rate are selected as the performance indicators of the model. When the accuracy is ≥85%, the recall rate is ≥80%, and the precision rate is ≥80%, it means that the model training is completed. The feature vector is formed according to the water quality parameters and environmental data in the test set, and input into the model for testing.

[0091] The process of extracting water quality parameters and environmental data into feature vectors can simplify complex data relationships into easy-to-handle numerical forms. The design of feature vectors comprehensively considers multiple water quality parameters and environmental factors, so that subsequent models can more comprehensively capture the characteristics of pollution sources. Extracting features can not only reduce model complexity, but also improve the interpretability of data and the convenience of analysis. Feature vectors cover multi-dimensional information, which helps to capture the full picture of pollution sources. By combining the data of target monitoring points and adjacent monitoring points, richer feature information can be obtained. This diverse data can more comprehensively reflect environmental changes and capture the potential impact of pollution sources. Single monitoring point data may be affected by local factors, such as instantaneous pollution events or measurement errors. Introducing data from adjacent monitoring points can smooth these local fluctuations and reduce the impact of chance on the model. Water quality changes often have spatial correlation. Data from adjacent monitoring points can provide the overall water quality status of the water body, help the model understand the interconnection between different monitoring points, and better identify the diffusion pattern of pollutants. In water quality monitoring, time factors usually lead to dynamic changes in data. The characteristic vectors of adjacent monitoring points can also help identify temporal trends and patterns, thereby better predicting changes in pollution sources. The model can more quickly identify and respond to temporary pollution incidents or sudden environmental changes, thereby improving the efficiency of environmental governance.

[0092] The pollution source identification model is constructed based on the support vector machine. With its powerful classification ability and adaptability, it can effectively process high-dimensional data. By setting a hyperplane to separate different types of pollution sources, the model can accurately identify potential pollution source types and improve the accuracy of identification. The loss function adopted fully considers the classification performance of the model and the penalty for misclassification. Such a design can continuously optimize the model during the training process and reduce the classification error rate. By adjusting the hyperparameters, the complexity of the model can be controlled and the recognition accuracy can be further improved. Accuracy, recall rate and precision are selected as evaluation indicators of model performance. These indicators can comprehensively measure the recognition ability of the model and ensure the effectiveness of the model in practical applications. When the model reaches the set performance threshold, it indicates the success of model training and enhances the credibility of subsequent applications.

[0093] Step 4: Obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, filter out the monitoring points with pollution and determine the water pollution source of each monitoring point at the current moment.

[0094] In this embodiment, the specific logic for determining the water pollution source of each monitoring point at the current moment is:

[0095] Based on the output of the pollution source identification model, the monitoring points with pollution codes containing 1 are screened out, and the pollution source types of the corresponding monitoring points are determined by comparing the pollution codes.

[0096] By screening out polluted monitoring points based on pollution codes, the pollution sources in the current water body can be identified quickly and accurately, which is the key to achieving real-time monitoring and emergency response. The use of pollution codes provides a systematic method to identify and classify different types of pollution sources, making the pollution status of the monitoring points clear at a glance and facilitating targeted analysis and governance. In addition, the incremental learning method is used to update the pollution source identification model parameters, ensuring that the model can adapt to the changing environment and data characteristics. This real-time update capability enables the model to quickly adjust and maintain efficient recognition accuracy when faced with new pollution events and data. This process not only improves the learning efficiency of the model, but also reduces the prediction inaccuracy caused by data aging, ensuring the long-term stability and reliability of the monitoring system.

[0097] See also Figure 2 The present invention further provides a water pollution source identification system based on data fusion, the water pollution source identification system based on data fusion is used to implement the above-mentioned water pollution source identification method based on data fusion, including:

[0098] The historical data collection module is used to mark a monitoring point every 500 meters from the entrance to the exit of the water body, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data includes temperature, humidity and rainfall one hour before the collection time.

[0099] An environmental status assessment module, for encoding water pollution sources at monitoring points to generate pollution codes, and for generating environmental indexes reflecting environmental status at monitoring points based on environmental data;

[0100] A pollution source identification model building module is used to concatenate water quality parameters and environmental indexes at monitoring points to form feature vectors, build and train feature vectors whose input is the target monitoring point and its adjacent monitoring points, and whose output is a pollution source identification model with pollution codes at the target monitoring point, and the target monitoring point is any monitoring point;

[0101] The pollution source identification and model updating module is used to obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, the monitoring points with pollution are screened out and the water pollution sources of each monitoring point at the current moment are determined.

[0102] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0103] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for identifying water pollution sources based on data fusion, characterized in that: The specific steps include: Step 1: From the entrance to the exit of the water body, mark each 500-meter interval as a monitoring point, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data include temperature, humidity and rainfall one hour before the collection time. Step 2: Encode the water pollution sources at the monitoring points to generate pollution codes, and generate environmental indexes reflecting the environmental status at the monitoring points based on environmental data; Step 3: Concatenate the water quality parameters and environmental indexes at the monitoring point to form a feature vector, construct and train the feature vector whose input is the target monitoring point and its adjacent monitoring points, and the output is the pollution source identification model of the pollution code at the target monitoring point, and the target monitoring point is any monitoring point; Step 4: Obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, filter out the monitoring points with pollution and determine the water pollution source of each monitoring point at the current moment.

2. The water pollution source identification method based on data fusion according to claim 1 is characterized in that: The specific logic behind the collection of water quality parameters and environmental data is: pH sensors, dissolved oxygen concentration sensors, chemical oxygen demand sensors, conductivity sensors, turbidity sensors, and ammonia nitrogen sensors are arranged at each monitoring point to collect water quality parameters. Environmental data are obtained through online meteorological services. The preprocessing of water quality parameters includes removing outliers, filling missing values, and normalizing data. The Z-score method is used for outlier detection. If the |Z| of the data is greater than 3, it is regarded as an outlier and is removed. After the outlier detection is completed, the corresponding missing value is filled with the mean of the same type of data from all monitoring points. The min-max normalization method is used to standardize the data to the [0,1] interval.

3. The water pollution source identification method based on data fusion according to claim 2 is characterized in that: The water pollution sources include four types: industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources. The pollution code is a four-digit code, and the four codes from left to right correspond to industrial pollution sources, agricultural pollution sources, domestic pollution sources and natural pollution sources respectively. When a certain type of pollution source does not exist at the monitoring point, the code at that position in the pollution code is 0, otherwise it is 1.

4. The water pollution source identification method based on data fusion according to claim 2 is characterized in that: The specific logic for calculating the environmental index is: And nonlinear transformation is performed on air temperature, temperature and rainfall. The transformation of air temperature is expressed as: T r =ln(T+c T ) Among them, T r is the temperature change, T is the collected temperature data, c T is a constant term; The transformation of humidity is expressed as: H r =H 2 Among them, H r is the transformation of humidity, H is the collected humidity data; The transformation of rainfall is expressed as: Among them, R r is the transformation of rainfall, R is the collected rainfall data, c R is a constant term; After nonlinear transformation of environmental data, the min-max standardization method is performed. The standardized temperature variables, humidity variables and rainfall variables are expressed as T n , H n , R n ; Introducing the interaction terms between environmental variables, the interaction between temperature and humidity is expressed as: I TH =T n ·H n Among them, I TH represents the interaction term between air temperature and humidity; Similarly, the interaction terms of temperature and rainfall and humidity and rainfall can be obtained, which are I TR ,I HR ; The formula for generating the environmental index that reflects the environmental status at the monitoring point is: EI=ω T ·T n +ω H ·H n +ω R ·R n +ω TH ·I TH +ω TR ·I TR +ω HR ·I HR Among them, HI is the environmental index, ω T ,ω H ,ω R ,ω TH ,ω TR ,ω HR are the weights of each variable and interaction term, 0<ω R <ω H <ω T <1,0<ω HR <ω TR <ω TH <1.

5. The method for identifying water pollution sources based on data fusion according to claim 4 is characterized in that: The specific logic for building the pollution source identification model is: The water quality parameters and environmental indexes at the monitoring points are concatenated to form a feature vector, expressed as: Among them, F represents the characteristic vector, pH is the pH value, DO is the dissolved oxygen concentration, COD is the chemical oxygen demand, EC is the electrical conductivity, TB is the turbidity, and NH3 is the ammonia nitrogen concentration; The formed feature vectors are integrated into a sample data set, and the sample data set is divided into a training set and a test set with a division ratio of 7:

3. The feature vectors of the target monitoring point and its adjacent monitoring points are used as input, and the pollution code at the target monitoring point is used as the label. A pollution source identification model is constructed based on the support vector machine algorithm, and a hyperplane is set to separate different types of pollution sources. The optimization goal of the model is to minimize the loss function. Accuracy, recall rate and precision rate are selected as the performance indicators of the model. When the accuracy is ≥85%, the recall rate is ≥80%, and the precision rate is ≥80%, it means that the model training is completed. The feature vector is formed according to the water quality parameters and environmental data in the test set and input into the model for testing.

6. The method for identifying water pollution sources based on data fusion according to claim 5 is characterized in that: The specific logic for determining the water pollution sources at each monitoring point at the current moment is: Based on the output of the pollution source identification model, the monitoring points with pollution codes containing 1 are screened out, and the pollution source types of the corresponding monitoring points are determined by comparing the pollution codes.

7. A water pollution source identification system based on data fusion, characterized in that: The water pollution source identification system based on data fusion is used to implement the water pollution source identification method based on data fusion according to any one of claims 1 to 6, comprising: The historical data collection module is used to mark a monitoring point every 500 meters from the entrance to the exit of the water body, obtain the historical water pollution sources, historical environmental data and historical water quality parameters of the monitoring point at the historical collection time, and pre-process the water quality parameters and environmental data. The water quality parameters include pH value, dissolved oxygen concentration, chemical oxygen demand, conductivity, turbidity and ammonia nitrogen concentration. The environmental data includes temperature, humidity and rainfall one hour before the collection time. An environmental status assessment module, for encoding water pollution sources at monitoring points to generate pollution codes, and for generating environmental indexes reflecting environmental status at monitoring points based on environmental data; A pollution source identification model building module is used to concatenate water quality parameters and environmental indexes at monitoring points to form feature vectors, build and train feature vectors whose input is the target monitoring point and its adjacent monitoring points, and whose output is a pollution source identification model with pollution codes at the target monitoring point, and the target monitoring point is any monitoring point; The pollution source identification and model updating module is used to obtain the environmental data and water quality data of the monitoring point at the current moment, convert them into feature vectors and input them into the trained pollution source identification model to obtain the pollution code of each monitoring point at the current moment. Based on the pollution code, the monitoring points with pollution are screened out and the water pollution sources of each monitoring point at the current moment are determined.

Images