A rapid pollution source localization system and method based on mobile monitoring data and source tracing algorithms

Abstract

This invention discloses a rapid pollution source location system and method based on mobile monitoring data and a source tracing algorithm. The system includes mobile water quality monitoring data collection, data preprocessing, source tracing algorithm model construction, model optimization, pollution source location, and result display. During water quality source tracing using mobile monitoring equipment, a dynamic tracking mechanism is employed. A source tracing algorithm is constructed to enable the monitoring equipment to move autonomously, adapting to changes in the pollution source's location and enabling rapid source location. Considering the energy consumption-accuracy trade-off during the mobile monitoring equipment's navigation, an energy consumption penalty is introduced into the reward function to avoid ineffective movement and improve the source tracing efficiency. Combining historical monitoring data sequences reflects the trend of water quality changes over time, improving the accuracy of model training. This allows for accurate and rapid water quality source tracing, achieving both high efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of mobile monitoring technology, and more specifically, to a rapid pollution source location system and method based on mobile monitoring data and source tracing algorithms. Background Technology

[0002] With the acceleration of industrialization and the improvement of urbanization, environmental pollution problems are becoming increasingly serious. Traditional pollution source monitoring methods often suffer from problems such as long monitoring cycles, data lag, and inaccurate location, making it difficult to meet the urgent needs of current environmental protection work. In particular, although fixed monitoring stations can provide continuous pollution data, their coverage is limited, making it difficult to cope with the need for rapid response to sudden pollution events. In addition, fixed monitoring stations are costly to build and lack flexibility, failing to meet the monitoring needs of large areas and high spatiotemporal resolution.

[0003] To compensate for the shortcomings of fixed monitoring stations, mobile monitoring technology has emerged. Mobile monitoring uses mobile platforms (such as vehicles, drones, and ships) equipped with high-precision monitoring equipment, enabling it to cover large areas in a short time and acquire real-time data on pollutant concentration distribution. However, when used alone, it faces difficulties in accurately locating pollution sources in complex environments, requiring combination with effective source tracing algorithms. Existing source tracing algorithms, such as the Lagrange particle diffusion model, rely on large amounts of meteorological data and are computationally complex; they also require sufficient sample data, limiting their application in pollution source location.

[0004] The accuracy of source tracing algorithms is highly dependent on model assumptions and parameter selection. The current technical challenge of using mobile monitoring data for rapid pollution source location is to combine real-time monitoring data to build models and optimize parameters, thereby achieving rapid and accurate location of pollution sources and providing strong technical support for emergency response to sudden pollution incidents and environmental pollution control. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid pollution source location system and method based on mobile monitoring data and source tracing algorithms, in order to solve the above-mentioned problems existing in the prior art.

[0006] The application is as follows:

[0007] A rapid pollution source location system based on mobile monitoring data and source tracing algorithm is provided, characterized by including a water quality mobile monitoring data collection module, a data preprocessing module, a source tracing algorithm model construction module, a source tracing algorithm model optimization module, a pollution source location module, and a result display module;

[0008] The water quality mobile monitoring data collection module: starts the mobile monitoring equipment, collects monitoring information of water quality environmental parameters in real time through the dynamic position adjustment of the mobile monitoring equipment, obtains mobile monitoring data, and transmits the mobile monitoring data to the data preprocessing module;

[0009] The data preprocessing module preprocesses the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmits the monitoring feature data to the source tracing algorithm model construction module.

[0010] The source tracing algorithm model construction module: describes the state of different monitoring feature data of the preprocessed mobile monitoring data, designs a reward function in combination with the source tracing target, controls the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and constructs a source tracing algorithm model for dynamic pollution source tracing;

[0011] The source tracing algorithm model optimization module: combines historical monitoring data sequences for model training, adjusts the magnitude of the pruning strategy update, performs backpropagation based on the loss calculation of the learning strategy and value function, and optimizes parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer.

[0012] The pollution source location module: uses a source tracing algorithm model to process the pre-treated water quality mobile monitoring data, determines the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, controls the mobile monitoring equipment to cruise according to the target navigation trajectory, and determines the location of the pollution source;

[0013] The results display module displays the data from the pollution source location module, marks the location of the mobile monitoring equipment on a map, visualizes the movement path of the mobile monitoring equipment, marks the location information of the pollution source, determines whether the source tracing task is successful, and calculates and outputs the success rate.

[0014] The process of describing the state of the preprocessed mobile monitoring data based on different monitoring feature data includes:

[0015] The location coordinates and pollutant concentrations of the mobile monitoring equipment are concatenated to form a status data set for the mobile monitoring equipment, which includes the location of the mobile monitoring equipment and the pollutant concentration. This data set is (Δx, Δy, C...). p ); where (Δx, Δy) represent the position coordinates of the mobile monitoring equipment, C p This indicates the pollutant concentration at the current location, calculated using the water quality diffusion equation.

[0016] The method of combining the source tracing target design reward function to control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment includes:

[0017] Positive rewards or negative penalties are applied based on the reward function of the mobile monitoring equipment's distance from the pollution source, energy consumption cost, time cost, and information value assessment to incentivize the learning of correct source tracing strategies. The navigation direction of the mobile monitoring equipment is determined, and the movement range and monitoring frequency of the mobile monitoring equipment are limited.

[0018] The reward function consists of two parts: positive reward and negative penalty, expressed as:

[0019] R=ω1*(-d)+ω2*(-(P*T))+ω3*(-T)+ω4*(C p *A)

[0020] Where R is the reward function; ω1, ω2, ω3, and ω4 represent weights used to balance the importance of each reward item; d represents the Euclidean distance between the mobile monitoring equipment and the pollution source; P represents the power of the mobile monitoring equipment; T represents the travel time of the mobile monitoring equipment; C p The value represents the pollutant concentration, and A represents the accuracy coefficient of the mobile monitoring data.

[0021] The method of combining the source tracing target design reward function to control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment also includes:

[0022] Set the termination conditions for the mobile monitoring equipment. The first termination condition is the distance condition. Set the first threshold condition for the Euclidean distance between the mobile monitoring equipment and the pollution source.

[0023] The second termination condition is a time condition, setting a second threshold condition for the movement time of the mobile monitoring equipment;

[0024] When the first threshold condition or the second threshold condition is met, a data set {St, Re, Do, In} is returned and stored, where St represents the status data set of the mobile monitoring equipment, St = (Δx, Δy, C). p Re indicates the reward obtained after the mobile monitoring equipment performs an action; Do indicates whether the termination condition has been met; In indicates additional information.

[0025] The construction of the source tracing algorithm model for dynamic pollution source tracing also includes:

[0026] The feature extraction network layer is configured to extract feature information of the state description. The feature extraction network layer consists of two fully connected layers and a ReLU activation function. The policy network layer and the value network layer share a feature extraction network layer.

[0027] The mobile monitoring strategy network layer extracts the feature information to obtain the mean value of the mobile monitoring device's navigation actions. Gaussian noise is added to the mean value of the navigation actions, and the probability distribution of the navigation actions is output to guide the mobile monitoring device in selecting the navigation actions to be performed.

[0028] The mobile monitoring value network layer extracts the feature information, outputs a value estimate of the mobile monitoring equipment status based on the value function, performs a value assessment of the current status, and helps the mobile monitoring strategy network layer update the strategy.

[0029] The mobile monitoring strategy network layer extracts the feature information to obtain the mean value of the mobile monitoring device's navigation actions. Gaussian noise is added to the mean value of the navigation actions, and the probability distribution of the navigation actions is output to guide the mobile monitoring device in selecting navigation actions, including:

[0030] The average navigation action of the mobile monitoring equipment is:

[0031] μ = ReLU(W*St + b)

[0032] Where μ represents the mean, W represents the weight vector, St represents the set of state data of the mobile monitoring equipment, b represents the bias vector, and ReLU() represents the activation function;

[0033] Adding Gaussian noise to the mean navigation motion result in:

[0034] a=μ+σ*ε

[0035] Where 'a' represents the subsequent navigation actions of the mobile monitoring equipment, 'σ' represents the standard deviation, controlling for the intensity of the noise, and 'ε' represents the noise sampled from the standard normal distribution.

[0036] The probability distribution of navigation actions is as follows:

[0037] The mobile monitoring value network layer extracts the feature information, outputs a value estimate of the mobile monitoring equipment status based on the value function, performs a value assessment of the current status, and helps the mobile monitoring strategy network layer update the strategy, including:

[0038] The value function is expressed as:

[0039]

[0040] Where V(St) represents the state value, which is the expected value of the cumulative reward for the mobile monitoring equipment starting from state St, and R... k It is the immediate reward at time step k, γ represents the discount factor, 0≤γ<1, used to balance the importance of current rewards and future rewards, and St0 represents the initial state of the mobile monitoring equipment;

[0041] For each state St of the mobile monitoring equipment, calculate the target state value:

[0042] V t (St)=R+γ(1-Do)V(St')

[0043] Among them, V t (St) represents the target state value of the mobile monitoring equipment, R represents the reward of the current state, V(St') represents the future state value, and Do represents whether the termination condition has been met. When Do meets the termination condition, the future reward is ignored, that is, V(St') = 0.

[0044] Calculate the value loss: Ls = (V(St) - V t (St)) 2 .

[0045] The magnitude of the pruning strategy update is calculated through backpropagation based on the loss of the learning strategy and the value function, and parameter optimization is performed by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer, including:

[0046] The magnitude of the pruning strategy update is achieved through the pruning objective function, which is expressed as:

[0047]

[0048] Where, r t (θ) represents the proportion of policy updates, At is the dominance function, representing the performance of the current mobile monitoring equipment relative to the average performance, and ∈ is the clipping range used to limit the magnitude of policy updates. This represents the expected value at time step k.

[0049] The pollution source location module also includes:

[0050] By using pre-processed water quality mobile monitoring data, the navigation position coordinates and pollutant concentrations of the mobile monitoring equipment are obtained, and the parameter indicators of the source tracing algorithm model are determined. The parameter indicators include the location information of the pollution source and the characteristic information that can reflect the change trend of water quality over time. The target navigation trajectory is determined based on the navigation coordinates of the mobile monitoring equipment, and the mobile monitoring equipment is controlled to cruise according to the target navigation trajectory to determine the final location of the pollution source.

[0051] This application also provides a method for rapid pollution source localization based on mobile monitoring data and source tracing algorithms, applied to the aforementioned rapid pollution source localization system based on mobile monitoring data and source tracing algorithms, specifically including the following steps:

[0052] Step 1: Start the mobile monitoring equipment and collect real-time monitoring information on water quality environmental parameters by dynamically adjusting the position of the mobile monitoring equipment to obtain mobile monitoring data;

[0053] Step 2: Preprocess the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmit the monitoring feature data to the source tracing algorithm model construction module;

[0054] Step 3: Describe the state of different monitoring characteristics of the preprocessed mobile monitoring data, design a reward function in combination with the source tracing target, control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and build a source tracing algorithm model for dynamic pollution source tracing;

[0055] Step 4: Train the model using historical monitoring data sequences, prune the update magnitude of the pruning strategy, perform backpropagation based on the loss calculation of the learning strategy and value function, and optimize the parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer.

[0056] Step 5: Process the pre-treated water quality mobile monitoring data using the source tracing algorithm model, determine the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, control the mobile monitoring equipment to cruise along the target navigation trajectory, and determine the location of the pollution source;

[0057] Step 6: Mark the location of the mobile monitoring equipment on the map, visualize the movement path of the mobile monitoring equipment, mark the location information of the pollution source, determine whether the source tracing task is successful, and calculate and output the success rate.

[0058] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0059] 1. When using mobile monitoring equipment for water quality source tracing, a dynamic tracking mechanism is adopted. By constructing a source tracing algorithm, the monitoring equipment can move autonomously to adapt to changes in the location of the pollution source and quickly locate the pollution source.

[0060] 2. Considering the energy consumption-accuracy trade-off of mobile monitoring equipment during navigation, an energy consumption penalty is introduced into the set reward function to avoid invalid movement and improve the traceability efficiency of the monitoring equipment;

[0061] 3. By combining historical monitoring data sequences, the model reflects the trend of water quality changes over time, improves the accuracy of model training, and enables accurate and rapid water quality source tracing, thus achieving high efficiency and accuracy in water quality source tracing. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of a pollution source rapid location system based on mobile monitoring data and source tracing algorithm provided in an embodiment of the present invention;

[0063] Figure 2 This is a flowchart illustrating a method for rapid pollution source location based on mobile monitoring data and source tracing algorithms provided in an embodiment of the present invention. Detailed Implementation

[0064] The present invention will now be described in detail with reference to the accompanying drawings.

[0065] Example 1

[0066] This invention provides a rapid pollution source location system based on mobile monitoring data and source tracing algorithms, such as... Figure 1 As shown, it includes a water quality mobile monitoring data collection module, a data preprocessing module, a source tracing algorithm model construction module, a source tracing algorithm model optimization module, a pollution source location module, and a results display module;

[0067] The water quality mobile monitoring data collection module: starts the mobile monitoring equipment, collects monitoring information of water quality environmental parameters in real time through the dynamic position adjustment of the mobile monitoring equipment, obtains mobile monitoring data, and transmits the mobile monitoring data to the data preprocessing module;

[0068] Specifically, the mobile monitoring equipment refers to mobile monitoring equipment such as unmanned boats and drones that include automatic data collection devices, without any specific limitations. The mobile monitoring equipment is deployed in the area to be monitored to ensure that the equipment can collect mobile monitoring data in real time. The mobile monitoring data includes: chemical oxygen demand (COD), ammonia nitrogen concentration, pH, coordinates of the mobile monitoring equipment, water flow velocity, wind direction, and temperature.

[0069] The data preprocessing module preprocesses the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmits the monitoring feature data to the source tracing algorithm model construction module.

[0070] Specifically, the preprocessing includes cleaning, denoising, and standardizing the mobile monitoring data; data cleaning can remove incomplete, inaccurate, duplicate, or irrelevant data; data denoising can remove noise interference from the data through technical means; and data standardization can unify the processing of data from different sources and in different formats.

[0071] The source tracing algorithm model construction module: describes the state of different monitoring feature data of the preprocessed mobile monitoring data, designs a reward function in combination with the source tracing target, controls the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and constructs a source tracing algorithm model for dynamic pollution source tracing;

[0072] Specifically, the pre-processed mobile monitoring data includes different monitoring characteristics such as: chemical oxygen demand (COD), ammonia nitrogen concentration, and pH, which reflect the characteristics of the collected water quality; the location coordinates of the mobile monitoring equipment, which reflect the characteristics of the collection location; and the water flow velocity, wind direction, and temperature, which reflect the characteristics of the collection environment. At the same time, historical monitoring data sequences are added to reflect the trend of water quality changes over time.

[0073] Based on the monitoring characteristic data, a set of characteristic data is extracted. The position coordinates of the mobile monitoring equipment are defined as (Δx, Δy). The diffusion coefficient D of pollutants and the velocity vector U are calculated by combining water quality characteristic data and environmental characteristic data, where the velocity component is U. x U y , representing the flow velocity of the fluid in the x and y directions; since different pollutants contribute differently to the pollutant concentration, the relative molar mass of water M0 is obtained based on different water quality characteristic data and environmental characteristic data. Based on the water quality characteristic data of each monitoring point in the monitoring data, as well as the water flow velocity, wind direction, and temperature in the environmental characteristic data, a diffusion model is established, and the pollutant concentration is calculated using the water quality diffusion equation.

[0074] The state description of the preprocessed mobile monitoring data based on different monitoring characteristic data includes:

[0075] The location coordinates and pollutant concentrations of the mobile monitoring equipment are concatenated to form a status data set for the mobile monitoring equipment, which includes the location of the mobile monitoring equipment and the pollutant concentration. This data set is (Δx, Δy, C...). p ); where (Δx, Δy) represent the position coordinates of the mobile monitoring equipment, C p This indicates the pollutant concentration at the current location, calculated using the water quality diffusion equation.

[0076]

[0077] Where (x0, y0) represents the location of the pollution source, U x U y Let M0 be the velocity component, M0 represent the total mass of pollutants released from the pollution source, i.e., the initial total amount of pollutants released, D represent the diffusion coefficient of the pollutants, and t represent time.

[0078] The design of a reward function based on the source tracing target, and the control of the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, includes: giving positive rewards or negative penalties based on the reward function of the mobile monitoring equipment's distance from the pollution source, energy consumption cost, time cost, and information value assessment, to incentivize the learning of the correct source tracing strategy, determining the navigation direction of the mobile monitoring equipment, and controlling the mobile monitoring equipment's movement range and monitoring frequency.

[0079] Based on the real-time location information of the mobile monitoring equipment, the reward is calculated, and the Euclidean distance between the mobile monitoring equipment and the pollution source is calculated:

[0080]

[0081] The location coordinates of the mobile monitoring equipment are (Δx, Δy), the location coordinates of the pollution source are (x0, y0), and d represents the straight-line distance between the two points.

[0082] The goal of source tracing is to comprehensively consider the distance, energy cost, time cost, and information value of the pollution source, guiding equipment to complete monitoring tasks efficiently and at low cost, while maximizing information value. The reward function consists of two parts: positive reward and negative penalty. The weights of the reward function are determined by simulation experiments or optimization algorithms to ensure its effectiveness, expressed as:

[0083] R=ω1*(-d)+ω2*(-(P*T))+ω3*(-T)+ω4*(C p *A)

[0084] Where R is the reward function; ω1, ω2, ω3, and ω4 represent weights used to balance the importance of each reward item; d represents the Euclidean distance between the mobile monitoring equipment and the pollution source; P represents the power of the mobile monitoring equipment; T represents the travel time of the mobile monitoring equipment; C p A represents the pollutant concentration, and A represents the accuracy coefficient of the mobile monitoring data.

[0085] Specifically, the navigation coordinates of the mobile monitoring equipment are (Δx, Δy), with a range limited to a movement of no more than 10 meters per trip, and selectable monitoring frequency adjustment is provided. If the mobile monitoring equipment's actions bring it closer to the pollution source, i.e., if the sampling data reveals a trend in water quality indicators pointing to a specific area, a positive reward is given. Conversely, if the actions cause it to move away from the pollution source or make an incorrect decision, such as choosing the wrong navigation direction, resulting in no more valuable information being obtained for a period of time, a negative reward is given. If the mobile monitoring equipment accurately identifies and locates the pollution source, a large positive reward is given to incentivize it to learn the correct source tracing strategy. The reward design encourages the monitoring equipment to approach the pollution source as quickly as possible while avoiding unnecessary movements.

[0086] By using real-time data collected from mobile monitoring equipment to determine the navigation direction of the equipment, and controlling its movement range and monitoring frequency, the following measures are also taken:

[0087] Set the termination conditions for the mobile monitoring equipment. The first termination condition is the distance condition. Set the first threshold condition for the Euclidean distance between the mobile monitoring equipment and the pollution source.

[0088] The second termination condition is a time condition, setting a second threshold condition for the movement time of the mobile monitoring equipment;

[0089] When the first threshold condition or the second threshold condition is met, a data set {St, Re, Do, In} is returned and stored, where St represents the status data set of the mobile monitoring equipment, St = (Δx, Δy, C). pRe indicates the reward obtained after the mobile monitoring equipment performs an action; Do indicates whether the termination condition has been met; In indicates additional information.

[0090] Acquiring status and action monitoring data from mobile monitoring equipment, describing the preprocessed mobile monitoring data with different monitoring feature information, designing a reward function based on the source tracing target, controlling the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and constructing a source tracing algorithm model for dynamic pollution source tracing also includes:

[0091] The feature extraction network layer is configured to extract feature information of the state description. The feature extraction network layer consists of two fully connected layers and a ReLU activation function. The policy network layer and the value network layer share a feature extraction network layer.

[0092] The mobile monitoring strategy network layer extracts the feature information to obtain the mean value of the mobile monitoring device's navigation actions. Gaussian noise is added to the mean value of the navigation actions, and the probability distribution of the navigation actions is output to guide the mobile monitoring device in selecting the navigation actions to be performed.

[0093] The mobile monitoring value network layer extracts the feature information, outputs the value estimate of the mobile monitoring equipment status according to the value function, performs the value assessment of the current status, and helps the mobile monitoring strategy network layer update the strategy.

[0094] Specifically, the feature extraction network layer extracts feature information describing the state. It consists of two fully connected layers and a ReLU activation function. The network structure is as follows: First layer: a 64-dimensional fully connected layer, with input dimension: the state dimension of the mobile monitoring device, and output dimension: 64; the ReLU activation function introduces non-linearity to alleviate the gradient vanishing problem; Second layer: a 64-dimensional fully connected layer, with input dimension: 64, and output dimension: 64; the ReLU activation function again introduces non-linearity; The third layer branches into the mobile monitoring strategy network layer and the mobile monitoring value network layer. The input of the mobile monitoring strategy network layer is the output dimension of the feature extraction network layer (64), which outputs the probability distribution of actions; the input of the mobile monitoring value network layer is the output dimension of the feature extraction network layer (64), which outputs a 1-dimensional scalar representing the value estimate of the state.

[0095] The mobile monitoring strategy network layer extracts the feature information to obtain the mean value of the mobile monitoring device's navigation actions. Gaussian noise is added to the mean value of the navigation actions, and the probability distribution of the navigation actions is output to guide the mobile monitoring device in selecting navigation actions, including:

[0096] The average navigation action of the mobile monitoring equipment is:

[0097] μ = ReLU(W*St + b)

[0098] Where μ represents the mean, W represents the weight vector, St represents the set of state data of the mobile monitoring equipment, b represents the bias vector, and ReLU() represents the activation function;

[0099] Adding Gaussian noise to the mean navigation motion result in:

[0100] a=μ+σ*ε

[0101] Where 'a' represents the subsequent navigation actions of the mobile monitoring equipment, 'σ' represents the standard deviation, controlling for the intensity of the noise, and 'ε' represents the noise sampled from the standard normal distribution.

[0102] The probability distribution of navigation actions is as follows:

[0103]

[0104] The mobile monitoring value network layer extracts the feature information, outputs a value estimate of the mobile monitoring equipment status based on the value function, performs a value assessment of the current status, and helps the mobile monitoring strategy network layer update the strategy, including:

[0105] The value function is expressed as:

[0106]

[0107] Where V(St) represents the state value, which is the expected value of the cumulative reward for the mobile monitoring equipment starting from state St, and R... k It is the immediate reward at time step k, γ represents the discount factor, 0≤γ<1, used to balance the importance of current rewards and future rewards, and St0 represents the initial state of the mobile monitoring equipment;

[0108] For each state St of the mobile monitoring equipment, calculate the target state value:

[0109] V t (St)=R+γ(1-Do)V(St')

[0110] Among them, V t (St) represents the target state value of the mobile monitoring equipment, R represents the reward of the current state, V(St') represents the future state value, and Do represents whether the termination condition has been met. When Do meets the termination condition, the future reward is ignored, that is, V(St') = 0.

[0111] Calculate the loss of value:

[0112] Ls=(V(St)-V t (St)) 2

[0113] The mobile monitoring strategy network layer and the mobile monitoring value network layer output action probabilities and state values, respectively, to solve reinforcement learning problems in continuous or discrete action spaces. In specific implementations, the state space of the mobile monitoring device is set to have 3 dimensions, and the state is represented as St=(Δx, Δy, C p The motion space has a dimension of 2, which is represented by the moving coordinates (Δx, Δy) of the mobile monitoring device. Hyperparameters including learning rate, discount factor, and clipping range are set.

[0114] The source tracing algorithm model optimization module: combines historical monitoring data sequences for model training, adjusts the magnitude of the pruning strategy update, performs backpropagation based on the loss calculation of the learning strategy and value function, and optimizes parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer.

[0115] The magnitude of the pruning strategy update is achieved through the pruning objective function, which is expressed as:

[0116]

[0117] Where, r t (θ) represents the proportion of policy updates, At is the dominance function, representing the performance of the current mobile monitoring equipment relative to the average performance, and ∈ is the clipping range used to limit the magnitude of policy updates. This represents the expected value at time step k;

[0118] Specifically, the strategy update ratio r t (θ) is the ratio between the current policy and the old policy. Its function is to measure the degree of change in the action selection of the current policy relative to the old policy in a given state. If r t A value greater than 1 for (θ) indicates that the current policy is more inclined to choose this action; a value less than 1 indicates that the current policy has a lower probability of choosing this action. The change in policy is evaluated by the ratio to ensure that each update is not too large.

[0119] The gradient of the loss function with respect to the model parameters is calculated using the chain rule, and the parameters of the mobile monitoring strategy network layer and the value network layer are updated using the gradient descent method. The source tracing algorithm is optimized by adjusting the hyperparameters of the learning rate, discount factor, and pruning range.

[0120] Example 2

[0121] As a specific implementation, during model training, the environment is initialized: the location of the pollution source is set to (50, 50). The flow velocity vector is set to (0.1, 0.0), indicating that the fluid flows at a speed of 0.1 in the x-direction, and the mobile monitoring strategy network layer and value network layer are trained.

[0122] Training rounds: Set the total number of training rounds to 500. Initialization: At the start of each training round, reset the environment and obtain the initial state of the mobile monitoring device, initialize an empty list to store trajectory data, and set the status list, action list, reward list, next status list, and termination flag list.

[0123] Select actions for the mobile monitoring equipment, limiting the action range to [-1, 1]. Zoom this range to [-10, 10] meters and execute the actions of the mobile monitoring equipment. Obtain the next state, reward, and termination flag. Store trajectory data: Store the current state, action, reward, next state, and termination flag into the corresponding lists. Update state: Assign the next state to the current state and continue interacting with the environment.

[0124] The current round of trajectory data is used to update the mobile monitoring strategy network layer and the value network layer. The specific update logic includes calculating the advantage function, pruning the policy update magnitude, calculating the loss function, and performing backpropagation. The training progress is output every 50 rounds. The number of training rounds is adjusted according to the task complexity to ensure that the model converges fully.

[0125] The pollution source location module: uses a source tracing algorithm model to process the pre-treated water quality mobile monitoring data, determines the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, controls the mobile monitoring equipment to cruise according to the target navigation trajectory, and determines the location of the pollution source;

[0126] By using the pre-processed water quality mobile monitoring data, the navigation position coordinates and pollutant concentrations of the mobile monitoring equipment are obtained, and the parameter indicators of the source tracing algorithm model are determined. The parameter indicators include the location information of the pollution source and the characteristic information that can reflect the change trend of water quality over time. The target navigation trajectory is determined based on the navigation coordinates of the mobile monitoring equipment, and the mobile monitoring equipment is controlled to cruise according to the target navigation trajectory to determine the final pollution source location information.

[0127] The results display module displays the data from the pollution source location module, marks the location of the mobile monitoring equipment on a map, visualizes the movement path of the mobile monitoring equipment, marks the pollution source location information, determines whether the source tracing task is successful, and calculates and outputs the success rate.

[0128] Specifically, the mobile monitoring equipment's movement path is recorded, and the success of each source tracing task is determined, i.e., whether the mobile monitoring equipment is close to the pollution source. The movement path of the monitoring equipment is visualized, and the location of the pollution source is marked. The location of the mobile monitoring equipment is recorded, and the success of the source tracing task is determined. If the distance between the monitoring equipment and the pollution source is less than 5, the task is considered successful. The success rate is calculated and output. The movement path of the mobile monitoring equipment is represented by a line connecting blue dots, and the location of the pollution source is represented by a red star mark.

[0129] Example 3

[0130] like Figure 2 As shown, the present invention also provides a method for rapid pollution source localization based on mobile monitoring data and source tracing algorithms, applied to the aforementioned rapid pollution source localization system based on mobile monitoring data and source tracing algorithms, specifically including the following steps:

[0131] Step 1: Start the mobile monitoring equipment and collect real-time monitoring information on water quality environmental parameters by dynamically adjusting the position of the mobile monitoring equipment to obtain mobile monitoring data;

[0132] Step 2: Preprocess the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmit the monitoring feature data to the source tracing algorithm model construction module;

[0133] Step 3: Describe the state of different monitoring characteristics of the preprocessed mobile monitoring data, design a reward function in combination with the source tracing target, control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and build a source tracing algorithm model for dynamic pollution source tracing;

[0134] Step 4: Train the model using historical monitoring data sequences, prune the update magnitude of the pruning strategy, perform backpropagation based on the loss calculation of the learning strategy and value function, and optimize the parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer.

[0135] Step 5: Process the pre-treated water quality mobile monitoring data using the source tracing algorithm model, determine the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, control the mobile monitoring equipment to cruise along the target navigation trajectory, and determine the location of the pollution source;

[0136] Step 6: Mark the location of the mobile monitoring equipment on the map, visualize the movement path of the mobile monitoring equipment, mark the location information of the pollution source, determine whether the source tracing task is successful, and calculate and output the success rate.

[0137] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0138] Similarly, it should be understood that, in order to simplify this disclosure and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0139] Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not other features, combinations of features from different embodiments are intended to be within the scope of the present invention and to form different embodiments. For example, in the claims below, any of the claimed embodiments may be used in any combination.

Claims

1. A rapid pollution source location system based on mobile monitoring data and source tracing algorithms, characterized in that, It includes a water quality mobile monitoring data collection module, a data preprocessing module, a source tracing algorithm model construction module, a source tracing algorithm model optimization module, a pollution source location module, and a results display module; The water quality mobile monitoring data collection module: starts the mobile monitoring equipment, collects monitoring information of water quality environmental parameters in real time through the dynamic position adjustment of the mobile monitoring equipment, obtains mobile monitoring data, and transmits the mobile monitoring data to the data preprocessing module; The data preprocessing module preprocesses the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmits the monitoring feature data to the source tracing algorithm model construction module. The source tracing algorithm model construction module: describes the state of different monitoring feature data of the preprocessed mobile monitoring data, designs a reward function in combination with the source tracing target, controls the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and constructs a source tracing algorithm model for dynamic pollution source tracing; The step of describing the state of the preprocessed mobile monitoring data with different monitoring characteristics includes: concatenating the location coordinates of the mobile monitoring device and the pollutant concentration into a state data set of the mobile monitoring device that includes the location of the mobile monitoring device and the pollutant concentration, wherein the data set is (Δx, Δy, C). p ); where (Δx, Δy) represent the position coordinates of the mobile monitoring equipment, C p The concentration of pollutants at the current location is calculated using the water quality diffusion equation; The design of a reward function based on the source tracing target, and the control of the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, includes: giving positive rewards or negative penalties based on the reward function of the mobile monitoring equipment's distance from the pollution source, energy consumption cost, time cost, and information value assessment, to incentivize the learning of the correct source tracing strategy, determining the navigation direction of the mobile monitoring equipment, and limiting the mobile monitoring equipment's movement range and monitoring frequency. The reward function consists of two parts: positive reward and negative penalty, expressed as: R=ω1*(-d)+ω2*(-(P*T))+ω3*(-T)+ω4*(C p *A) Where R is the reward function; ω1, ω2, ω3, and ω4 represent weights used to balance the importance of each reward item; d represents the Euclidean distance between the mobile monitoring equipment and the pollution source; P represents the power of the mobile monitoring equipment; T represents the travel time of the mobile monitoring equipment; C p A represents the pollutant concentration, and A represents the accuracy coefficient of the mobile monitoring data. The construction of the source tracing algorithm model for dynamic pollution source tracing also includes: setting a feature extraction network layer to extract feature information describing the state, wherein the feature extraction network layer consists of two fully connected layers and a ReLU activation function, and the policy network layer and the value network layer share a feature extraction network layer; the mobile monitoring policy network layer extracts the feature information to obtain the mean value of the mobile monitoring equipment's navigation actions, adds Gaussian noise to the mean value of the navigation actions, and outputs the probability distribution of the navigation actions to guide the mobile monitoring equipment in selecting navigation actions; the mobile monitoring value network layer extracts the feature information, outputs a value estimate of the mobile monitoring equipment's state according to the value function, performs a value assessment of the current state, and helps the mobile monitoring policy network layer update the policy; The source tracing algorithm model optimization module: combines historical monitoring data sequences for model training, adjusts the magnitude of the pruning strategy update, performs backpropagation based on the loss calculation of the learning strategy and value function, and optimizes parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer. The magnitude of the pruning strategy update is calculated through backpropagation based on the loss of the learning strategy and the value function, and parameter optimization is performed by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer, including: The magnitude of the pruning strategy update is achieved through the pruning objective function, which is expressed as: Where, r t (θ) represents the proportion of policy updates, At is the dominance function, representing the performance of the current mobile monitoring equipment relative to the average performance, and ∈ is the clipping range used to limit the magnitude of policy updates. This represents the expected value at time step k, and clip() represents the clipping function. The pollution source location module: uses a source tracing algorithm model to process the pre-treated water quality-mobile monitoring data, determines the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, controls the mobile monitoring equipment to cruise along the target navigation trajectory, and determines the location of the pollution source; The results display module displays the data from the pollution source location module, marks the location of the mobile monitoring equipment on a map, visualizes the movement path of the mobile monitoring equipment, marks the location information of the pollution source, determines whether the source tracing task is successful, and calculates and outputs the success rate.

2. The pollution source rapid location system based on mobile monitoring data and source tracing algorithm according to claim 1, characterized in that, The method of combining the source tracing target design reward function to control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment also includes: Set the termination conditions for the mobile monitoring equipment. The first termination condition is the distance condition. Set the first threshold condition for the Euclidean distance between the mobile monitoring equipment and the pollution source. The second termination condition is a time condition, setting a second threshold condition for the movement time of the mobile monitoring equipment; When the first threshold condition or the second threshold condition is met, a data set {St, Re, Do, In} is returned and stored, where St represents the status data set of the mobile monitoring equipment, St = (Δx, Δy, C). p Re indicates the reward obtained after the mobile monitoring equipment performs an action; Do indicates whether the termination condition has been met; In indicates additional information.

3. The pollution source rapid location system based on mobile monitoring data and source tracing algorithm according to claim 1, characterized in that, The mobile monitoring strategy network layer extracts the feature information to obtain the mean value of the mobile monitoring device's navigation actions. Gaussian noise is added to the mean value of the navigation actions, and the probability distribution of the navigation actions is output to guide the mobile monitoring device in selecting navigation actions, including: The average navigation action of the mobile monitoring equipment is: μ = ReLU(W*St + b) Where μ represents the mean, W represents the weight vector, St represents the set of state data of the mobile monitoring equipment, b represents the bias vector, and ReLU() represents the activation function; Adding Gaussian noise to the mean navigation motion result in: a=μ+σ*ε Where 'a' represents the subsequent navigation actions of the mobile monitoring equipment, 'σ' represents the standard deviation, controlling for the intensity of the noise, and 'ε' represents the noise sampled from the standard normal distribution. The probability distribution of navigation actions is as follows:

4. The pollution source rapid location system based on mobile monitoring data and source tracing algorithm according to claim 1, characterized in that, The mobile monitoring value network layer extracts the feature information, outputs a value estimate of the mobile monitoring equipment status based on the value function, performs a value assessment of the current status, and helps the mobile monitoring strategy network layer update the strategy, including: The value function is expressed as: Where V(St) represents the state value, which is the expected value of the cumulative reward for the mobile monitoring equipment starting from state St, and R... k It is the immediate reward at time step k, γ represents the discount factor, 0≤γ<1, used to balance the importance of current rewards and future rewards, and St0 represents the initial state of the mobile monitoring equipment; For each state St of the mobile monitoring equipment, calculate the target state value: V t (St)=R+γ(1-Do)V(St') Among them, V t (St) represents the target state value of the mobile monitoring equipment, R represents the reward of the current state, V(St') represents the future state value, and Do represents whether the termination condition has been met. When Do meets the termination condition, the future reward is ignored, that is, V(St') = 0. Calculate the value loss: Ls = (V(St) - V t (St)) 2 .

5. A rapid pollution source location system based on mobile monitoring data and source tracing algorithm according to claim 1, characterized in that, The pollution source location module also includes: By using pre-processed water quality mobile monitoring data, the navigation position coordinates and pollutant concentrations of the mobile monitoring equipment are obtained, and the parameter indicators of the source tracing algorithm model are determined. The parameter indicators include the location information of the pollution source and the characteristic information that can reflect the change trend of water quality over time. The target navigation trajectory is determined based on the navigation coordinates of the mobile monitoring equipment, and the mobile monitoring equipment is controlled to cruise according to the target navigation trajectory to determine the final location of the pollution source.

6. A method for rapid pollution source location based on mobile monitoring data and source tracing algorithms, characterized in that, The pollution source rapid location system based on mobile monitoring data and source tracing algorithm as described in any one of claims 1 to 5 specifically includes the following steps: Step 1: Start the mobile monitoring equipment and collect real-time monitoring information on water quality environmental parameters by dynamically adjusting the position of the mobile monitoring equipment to obtain mobile monitoring data; Step 2: Preprocess the collected mobile monitoring data to obtain the monitoring feature data of the mobile monitoring data, and transmit the monitoring feature data to the source tracing algorithm model construction module; Step 3: Describe the state of different monitoring characteristics of the preprocessed mobile monitoring data, design a reward function in combination with the source tracing target, control the subsequent navigation direction and monitoring frequency of the mobile monitoring equipment, and build a source tracing algorithm model for dynamic pollution source tracing; Step 4: Train the model using historical monitoring data sequences, prune the update magnitude of the pruning strategy, perform backpropagation based on the loss calculation of the learning strategy and value function, and optimize the parameters by combining the mobile monitoring strategy network layer and the mobile monitoring value network layer. Step 5: Process the pre-treated water quality mobile monitoring data using the source tracing algorithm model, determine the target navigation trajectory based on the navigation coordinates of the mobile monitoring equipment, control the mobile monitoring equipment to cruise along the target navigation trajectory, and determine the location of the pollution source; Step 6: Mark the location of the mobile monitoring equipment on the map, visualize the movement path of the mobile monitoring equipment, mark the location information of the pollution source, determine whether the source tracing task is successful, and calculate and output the success rate.

Images