Method, device and computer-readable storage medium for determining pollution source

Abstract

A method (100), apparatus (600, 700), and computer-readable storage medium for determining pollution sources relate to the field of pollution monitoring technology. The method (100) includes: determining a causal relationship (101) between a sensor and a potential pollution source in a target wind direction based on a positional relationship between the sensor and the potential pollution source; establishing a Bayesian network (102) corresponding to the target wind direction and containing the causal relationship; training the Bayesian network (103) based on historical sensor data when the wind direction is the target wind direction; and determining a pollution source from potential pollution sources based on current sensor data when the wind direction is the target wind direction and the trained Bayesian network (104). Since the influence of wind direction on sensor monitoring is taken into account, the pollution source determined is highly accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of pollution monitoring, and in particular to a method, device and computer-readable storage medium for determining a pollution source. Background Art

[0002] An industrial park is a designated area of ​​land planned for the location of industrial facilities. These areas typically have strict regulations on the discharge of pollutants (liquids or gases). A growing number of organizations are recognizing the importance of eco-friendly industrial parks (EIPs) for sustainable development.

[0003] Currently, monitoring points are usually set up at fixed locations in industrial parks to monitor pollutant emissions, and pollution sources are identified based on the monitoring results.

[0004] However, the wind direction within the industrial park may cause deviations in the monitoring results of the monitoring points, resulting in inaccurate pollution source investigation results. Summary of the Invention

[0005] Embodiments of the present invention provide a method, an apparatus, and a computer-readable storage medium for determining a pollution source.

[0006] First, a method for determining pollution sources is provided, including:

[0007] Based on the positional relationship between the sensor and the potential pollution source, determine the causal relationship between the sensor and the potential pollution source in the target wind direction;

[0008] Establishing a Bayesian network corresponding to the target wind direction and including the causal relationship;

[0009] Training the Bayesian network based on historical sensor data when the wind direction is the target wind direction;

[0010] Based on the current data of the sensor when the wind direction is the target wind direction and the trained Bayesian network, a pollution source is determined from the potential pollution sources.

[0011] In a second aspect, a device for determining a pollution source is provided, comprising:

[0012] a relationship determination module for determining a causal relationship between the sensor and the potential pollution source in the target wind direction based on the positional relationship between the sensor and the potential pollution source;

[0013] a network establishment module, configured to establish a Bayesian network corresponding to the target wind direction and including the causal relationship;

[0014] A training module, configured to train the Bayesian network based on historical sensor data when the wind direction is the target wind direction;

[0015] A determination module is used to determine a pollution source from the potential pollution sources based on current sensor data when the wind direction is the target wind direction and the trained Bayesian network.

[0016] In a third aspect, a device for determining a pollution source is provided, comprising a processor and a memory;

[0017] The memory stores an application program executable by the processor, which is used to enable the processor to execute the method for determining a pollution source as described in any one of the above items.

[0018] In a fourth aspect, a computer-readable storage medium is provided, in which computer-readable instructions are stored, and the computer-readable instructions are used to execute the method for determining the pollution source as described in any one of the above items.

[0019] It can be seen that the embodiment of the present invention establishes respective Bayesian networks corresponding to different target wind directions, and then uses the Bayesian network corresponding to the current target wind direction to accurately determine the pollution source. Therefore, the embodiment of the present invention takes into account the influence of wind direction on the monitoring effect of the sensor, and the accuracy of the determined pollution source is high.

[0020] For any of the above aspects, preferably, it also includes: determining the prior probability distribution of the Bayesian network; and calculating the posterior probability distribution of the Bayesian network using a belief propagation algorithm based on the prior probability distribution of the Bayesian network and the sensor historical data when the wind direction is the target wind direction.

[0021] Therefore, the embodiments of the present invention calculate the posterior probability distribution of the Bayesian network through the belief propagation algorithm, so that the Bayesian network can be trained quickly.

[0022] For any of the above aspects, preferably, the calculation of the posterior probability distribution of the Bayesian network using the belief propagation algorithm includes: in each iterative calculation step of the belief propagation algorithm, the wind speed when the wind direction is the target wind direction is used as a parameter affecting the sensor monitoring value, wherein the greater the wind speed, the smaller the sensor monitoring value.

[0023] It can be seen that the embodiment of the present invention also takes into account the influence of wind speed on the monitoring effect of the sensor, and uses wind speed as a parameter in the training process of the Bayesian network, so that the trained Bayesian network is more accurate.

[0024] For any of the above aspects, preferably, determining the pollution source from the potential pollution sources based on the current sensor data when the wind direction is the target wind direction and the trained Bayesian network includes:

[0025] Inputting the current data of the sensor when the wind direction is the target wind direction into the trained Bayesian network;

[0026] Calculating an emission probability distribution of each potential pollution source based on the posterior probability of the trained Bayesian network and current sensor data when the wind direction is the target wind direction;

[0027] Sort the emission values ​​corresponding to the highest probability point of the emission probability distribution of each potential pollution source;

[0028] The potential pollution source corresponding to the maximum emission value is determined as the pollution source.

[0029] Therefore, the embodiment of the present invention calculates the emission probability distribution of each potential pollution source and determines the potential pollution source corresponding to the maximum emission value as the pollution source, thereby achieving accurate identification of the pollution source.

[0030] For any of the above aspects, preferably, the training of the Bayesian network based on historical sensor data when the wind direction is the target wind direction includes:

[0031] Training the Bayesian network based on actual historical data of the sensor when the wind direction is the target wind direction; or

[0032] The Bayesian network is trained based on sensor simulation historical data when the wind direction is the target wind direction.

[0033] It can be seen that the embodiments of the present invention can use both actual historical data of sensors to train the Bayesian network and simulated historical data of sensors to train the Bayesian network, thereby enriching the training data. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is an exemplary flow chart of a method for determining pollution sources according to an embodiment of the present invention.

[0035] Figure 2 This is an exemplary schematic diagram of an industrial park topology map according to an embodiment of the present invention.

[0036] Figure 3 FIG. 4 is an exemplary schematic diagram of a Bayesian network corresponding to wind direction according to an embodiment of the present invention.

[0037] Figure 4 An exemplary schematic diagram of determining pollution sources according to an embodiment of the present invention.

[0038] Figure 5 This is an exemplary flow chart of a method for determining pollution sources in an industrial park according to an embodiment of the present invention.

[0039] Figure 6 This is an exemplary structural diagram of an apparatus for determining pollution sources according to an embodiment of the present invention.

[0040] Figure 7This is an exemplary structural diagram of an apparatus for determining pollution sources according to an embodiment of the present invention.

[0041] The accompanying drawings are numerals as follows:

[0042] Label meaning 100 Methods for determining pollution sources 101～104 step 10 Industrial Park 11、12、13 building 31 Bayesian network corresponding to the northeast wind 32 Bayesian network corresponding to the northwest wind 33 Bayesian network corresponding to southwest wind 34 Bayesian network corresponding to the southeast wind 41 Emission probability distribution curve of potential pollution source A 42 Emission probability distribution curve of potential pollution source B 43 Emission probability distribution curve of potential pollution source C 500 Methods for Identifying Pollution Sources in Industrial Parks 501～505 step 600 Device for identifying pollution sources 601 Relationship determination module

[0043] 602 Network determination module 603 Training Module 604 Determine the module 700 Device for identifying pollution sources 701 processor 702 Memory DETAILED DESCRIPTION

[0044] In order to make the technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to illustrate the present invention and are not configured to limit the scope of protection of the present invention.

[0045] For the sake of brevity and intuitiveness in description, the solution of the present invention is explained below by describing several representative embodiments. A large number of details in the embodiments are only configured to help understand the solution of the present invention. However, it is obvious that the technical solution of the present invention may not be limited to these details when implemented. In order to avoid unnecessarily obscuring the solution of the present invention, some embodiments are not described in detail, but only a framework is given. Hereinafter, "including" means "including but not limited to", and "according to..." means "at least according to..., but not limited to only according to...". Due to the language habits of Chinese, when the number of a component is not specifically specified below, it means that the component can be one or more, or can be understood as at least one.

[0046] The applicant discovered that the existing technology ignores the impact of wind direction on the monitoring effect of the sensor, resulting in low accuracy in determining the pollution source.

[0047] To solve this technical problem, the embodiment of the present invention takes into account the influence of wind direction on the monitoring effect of the sensor, pre-establishes a Bayesian network corresponding to different wind directions for determining the pollution source, and uses the training data under each wind direction to train the Bayesian network of each wind direction, and then uses the current data of the sensor at the current wind direction and the trained Bayesian network corresponding to the current wind direction to accurately determine the pollution source.

[0048] Furthermore, the embodiment of the present invention also takes into account the impact of wind speed on the monitoring effect of the sensor (for example, when the wind speed is strong, the monitored pollutant concentration is lower), and uses wind speed as a parameter in the training process of the Bayesian network, so that the trained Bayesian network is more accurate.

[0049] Figure 1 This is an exemplary flow chart of a method for determining pollution sources according to an embodiment of the present invention.

[0050] like Figure 1As shown, the method 100 includes:

[0051] Step 101: Based on the positional relationship between the sensor and the potential pollution source, determine the causal relationship between the sensor and the potential pollution source in the target wind direction.

[0052] Here, the number of sensors can be one or more, and the number of potential pollution sources can also be one or more. The sensors are adapted to monitor the physical quantity (e.g., concentration) of pollutants emitted by the potential pollution sources. Furthermore, the deployment locations of the sensors are fixed, and the deployment locations of the potential pollution sources are also fixed. When the potential pollution sources emit multiple types of pollutants, the sensors are multiple sensors adapted to detect the respective pollutant types.

[0053] Applicants have discovered that when the wind direction aligns with the sensor's detection direction, the sensor is more likely to detect pollutants. However, when the wind direction is opposite to the sensor's detection direction, the sensor has a harder time detecting pollutants. Therefore, a method can first determine several predetermined target wind directions. Then, based on the positional relationship between the sensor and the potential pollution source, the causal relationship between the sensor and the potential pollution source can be determined for each target wind direction.

[0054] Let’s take an industrial park as an example. Figure 2 This is an exemplary schematic diagram of an industrial park topology map according to an embodiment of the present invention.

[0055] exist Figure 2 In the figure, industrial park 10 contains buildings 11, 12, and 13. Building 11 contains potential pollution source B, building 12 contains potential pollution source D, and building 13 contains potential pollution sources A and C. Potential pollution sources A, B, C, and D are all capable of emitting the same type of pollutant, such as sulfur dioxide. Arrow N indicates the direction north.

[0056] In addition, sensor X is arranged at the western edge of industrial park 10, sensor Y is arranged at the southern edge of industrial park 10, and sensor Z is arranged at the eastern edge of industrial park 10. Sensor X, sensor Y, and sensor Z are respectively adapted to monitor pollutants emitted by potential pollution source A, potential pollution source B, potential pollution source C, and potential pollution source D. Preferably, sensor X, sensor Y, and sensor Z are each set with a concentration threshold value. When the monitored pollutant exceeds the concentration threshold value, the pollutant is determined to be detected and the concentration threshold value is recorded.

[0057] based on Figure 2The industrial park shown can determine the causal relationship between sensors and potential pollution sources in the target wind direction. The number of target wind directions can be multiple, based on different wind direction angles. For example, when the wind direction angle is 90 degrees, four target wind directions can be identified: northwest, northeast, southwest, and southeast. The smaller the wind direction angle, the greater the number of target wind directions.

[0058] The following example uses four target wind directions: northwest, northeast, southwest, and southeast. The causal relationships between sensors X, Y, and Z and potential pollution sources A, B, C, and D are as follows:

[0059] (1) When the target wind direction is northwest wind (i.e., the wind blows from the northwest), the pollutants emitted by potential pollution sources A, potential pollution source B, and potential pollution source C will be monitored by sensor Y.

[0060] (2) When the target wind direction is northeast (i.e., the wind blows from the northeast), the pollutants emitted by potential pollution source A will be monitored by sensor X; the pollutants emitted by potential pollution source B will be monitored by sensors X and Y; and the pollutants emitted by potential pollution source D will be monitored by sensor Y.

[0061] (3) When the target wind direction is southwest (i.e., the wind blows from the southwest), the pollutants emitted by potential pollution source A will be monitored by sensor Z; the pollutants emitted by potential pollution source B will be monitored by sensor Z; the pollutants emitted by potential pollution source C will be monitored by sensor Z; and the pollutants emitted by potential pollution source D will be monitored by sensor Z.

[0062] (4) When the target wind direction is southeast wind (i.e., the wind blows from the southeast), the pollutants emitted by potential pollution source A will be monitored by sensor X; the pollutants emitted by potential pollution source C will be monitored by sensor X; and the pollutants emitted by potential pollution source D will be monitored by sensors X and Z.

[0063] The above uses four target wind directions as an example to illustrate the causal relationship between sensors and potential pollution sources. Those skilled in the art will appreciate that this description is merely illustrative and does not limit the scope of protection of the embodiments of the present invention. In practice, the causal relationship between sensors and potential pollution sources can be implemented in a variety of ways, based on the different wind direction classifications and the locational relationships between sensors and potential pollution sources.

[0064] Step 102: Establish a Bayesian network corresponding to the target wind direction and including causal relationships.

[0065] Here, when the number of target wind directions is multiple, there are also multiple Bayesian networks corresponding to the target wind directions.

[0066] Specifically, each Bayesian network corresponding to its own target wind direction is a directed acyclic graph (DAG), which is composed of nodes and directed edges connecting these nodes. The nodes represent random variables, and the directed edges between the nodes represent the mutual relationships between the nodes (from the parent node to its child node), where the relationship strength is expressed by conditional probability, and those without parent nodes are expressed by prior probability. The random variables of the Bayesian network for each target wind direction include sensors and potential pollution sources. Each arrow indicates that the pollutants emitted by the potential pollution source can be monitored by the sensor when the target wind direction occurs, which is a causal relationship modeling.

[0067] against Figure 2 The industrial park shown, Figure 3 FIG. 1 is an exemplary schematic diagram of a Bayesian network corresponding to a target wind direction according to an embodiment of the present invention. Figure 3 In the figure, the N arrow points to the north; the E arrow points to the east.

[0068] exist Figure 3 middle:

[0069] The Bayesian network 31 corresponding to the northeast wind includes nodes A, B, D, X, and Y. Node A represents potential pollution source A, B represents potential pollution source B, C represents potential pollution source C, and D represents potential pollution source D. Node X represents sensor X, and node Y represents sensor Y. Nodes A and B point to sensor X, respectively, and nodes B and D point to sensor Y, respectively.

[0070] The Bayesian network 32 corresponding to the northwest wind includes nodes A, B, C, and Y. Node A represents potential pollution source A, B represents potential pollution source B, C represents potential pollution source C, and Y represents sensor Y. Nodes A, B, and C each point to sensor Y.

[0071] The Bayesian network 33 corresponding to the southwest wind includes nodes A, B, C, D, and Z. Node A represents potential pollution source A, node B represents potential pollution source B, node C represents potential pollution source C, node D represents potential pollution source D, and node Z represents sensor Z. Nodes A, B, C, and D each point to sensor Z.

[0072] The Bayesian network 34 corresponding to the southeast wind includes nodes A, C, D, X, and Z. Node A represents potential pollution source A, C represents potential pollution source C, D represents potential pollution source D, X represents sensor X, and Z represents sensor Z. Nodes A, C, and D each point to sensor X, and D also points to sensor Z.

[0073] The above describes the Bayesian network of each target wind direction by taking four target wind directions as an example. Those skilled in the art will appreciate that this description is merely exemplary and is not intended to limit the scope of protection of the embodiments of the present invention.

[0074] Step 103: Training the Bayesian network based on historical sensor data when the wind direction is the target wind direction.

[0075] Here, the sensor's historical data when the wind direction is the target wind direction can be used to train the Bayesian network corresponding to the target wind direction. The sensor whose historical data is used to train the Bayesian network corresponding to the target wind direction belongs to the Bayesian network corresponding to the target wind direction.

[0076] For example, historical sensor data from northeasterly winds is used to train Bayesian network 31 for northeasterly winds. Bayesian network 31 for northeasterly winds includes sensor X and sensor Y. Therefore, Bayesian network 31 for northeasterly winds is trained using historical data from sensor X when the wind direction is northeasterly and historical data from sensor Y when the wind direction is northeasterly.

[0077] The Bayesian network 32 corresponding to the northwest wind is trained using the sensor historical data when the wind direction is northwest. The Bayesian network 32 corresponding to the northwest wind includes sensor Y. Therefore, the Bayesian network 32 corresponding to the northwest wind is trained using the historical data of sensor Y when the wind direction is northwest.

[0078] The Bayesian network 33 corresponding to the southwest wind is trained using the sensor historical data when the wind direction is southwest. The Bayesian network 33 corresponding to the southwest wind includes the sensor Z. Therefore, the Bayesian network 33 corresponding to the southwest wind is trained using the historical data of the sensor Z when the wind direction is southwest.

[0079] The Bayesian network 34 for the southeast wind is trained using historical sensor data when the wind direction is southeast. The Bayesian network 34 for the southeast wind includes sensor X and sensor Z. Therefore, the Bayesian network 34 for the southeast wind is trained using historical data from sensor X when the wind direction is southeast and historical data from sensor Z when the wind direction is southeast.

[0080] In one embodiment, a Bayesian network corresponding to the target wind direction can be trained based on the actual historical sensor data when the wind direction is the target wind direction. The actual historical sensor data (monitoring values) for the target wind direction can be obtained from the sensor's historical database, and the Bayesian network corresponding to the target wind direction can be trained using the actual historical sensor data for the target wind direction. For example, the Bayesian network 31 corresponding to the northeast wind can be trained using the actual historical data of sensor X and sensor Y included in the Bayesian network 31 corresponding to the northeast wind.

[0081] In one embodiment, a Bayesian network corresponding to the target wind direction can be trained based on historical sensor simulation data for the target wind direction. Here, a sensor simulation model is pre-established, and the sensor simulation model is used to obtain historical sensor simulation data for the target wind direction. The Bayesian network corresponding to the target wind direction is then trained using the historical sensor simulation data for the target wind direction. For example, the Bayesian network 31 corresponding to northeast wind is trained using the historical simulation data for sensor X and sensor Y included in the Bayesian network 31 corresponding to northeast wind.

[0082] In one embodiment, training a Bayesian network based on historical sensor data when the wind direction is the target wind direction includes: separately determining the prior probability distribution of each Bayesian network corresponding to the target wind direction; and calculating the posterior probability distribution of the Bayesian network corresponding to the target wind direction using a belief propagation algorithm based on the prior probability distribution of the Bayesian network corresponding to the target wind direction and the historical sensor data when the wind direction is the target wind direction.

[0083] The belief propagation algorithm, also known as the sum-product message passing algorithm, is a message passing algorithm for inference on graphical models that can be used in Bayesian networks and Markov random fields.

[0084] Below is Figure 2 Taking the topological structure shown in the figure as an example, the process of calculating the prior probability distribution of the Bayesian network is described.

[0085] (1) Assume that potential pollution sources A, B, C, and D obey the uniform distribution U(0, m), where m can be set to the maximum value detected by sensor X, sensor Y, and sensor Z in the wind direction in history.

[0086] Taking potential pollution source A as an example, when the wind is southeast:

[0087] m=max{y1,y2,y3,...y T}; where y1y2, y3, ...y T is the detection value of sensor Y at each time point;

[0088] The potential pollution source A obeys a uniform distribution, so the probability density function f(a) of the potential pollution source A is:

[0089]

[0090] f(a)=0, a≥m or a≤0;

[0091] (2) For sensor X, sensor Y and sensor Z, assume that they obey discrete normal distribution. Take sensor X as an example, satisfying X~N(μ x ,σ x ),μ x is the historical mean of the wind direction; σ x The variance of the wind direction monitored historically is: Where f(x) is the probability density function of sensor X.

[0092] After determining the prior probability distribution of the Bayesian network based on the above description, the posterior probability distribution of the Bayesian network can be calculated using the belief propagation algorithm using historical sensor data corresponding to wind direction. In one embodiment, calculating the posterior probability distribution of the Bayesian network using the belief propagation algorithm includes: during each iterative calculation step of the belief propagation algorithm, using the wind speed when the wind direction is the target wind direction as a parameter that affects the monitoring value of the sensor in the Bayesian network corresponding to the target wind direction. The greater the wind speed, the smaller the sensor monitoring value. Correspondingly, the smaller the wind speed, the larger the sensor monitoring value.

[0093] For example, when the wind is southeast, the belief propagation algorithm is as follows:

[0094] Input: random variables A, B, C, Y and their respective prior probabilities f(a), f(b), f(c), and f(y), convergence threshold ε, wind speed s;

[0095] Definition: Node N∈{A, B, C, Y};

[0096] initialization:

[0097] (1) When t = 0, the initial information is equal to the prior probability: m(n) = f(n);

[0098] (2) Define the initial information queue M = {m(a), m(b), m(c), m(y)};

[0099] (3) Select an initial node N1 from N to start transmitting information.

[0100] Start the iterative process (where t = {2, 3, 4, ...}). The iterative process includes: (1) Node N1 transmits a message to node N2 (there needs to be a direct arrow link between N1 and N2, such as A->Y, Y->B, Y->C); (2) Calculate the joint probability distribution p(n1, n2) of N1 and N2, where p(n1, n2) = m(n1) * m(n2); (3) Update the message If convergence: Belief (t) (n2)–Belief (t-1) (n2) <= ε, then remove node N2 from the information queue M. When the information queue becomes an empty set When , the posterior probability distribution Pr(N) is returned, and its probability density function is p(n) = Belief (t) (n).

[0101] The above describes an implementation example of the belief propagation algorithm using the southeast wind as an example. Those skilled in the art will appreciate that this description is merely exemplary and is not intended to limit the scope of protection of the embodiments of the present invention.

[0102] Step 104: Based on the current sensor data when the wind direction is the target wind direction and the trained Bayesian network, a pollution source is determined from potential pollution sources.

[0103] Here, the current data of the sensor when the wind direction is the target wind direction is input into a trained Bayesian network corresponding to the target wind direction, so that the Bayesian network determines the pollution source from the potential pollution sources.

[0104] In one embodiment, step 104 determines the pollution source from the potential pollution sources based on the current data of the sensor when the wind direction is the target wind direction and the trained Bayesian network, specifically including: inputting the current data of the sensor when the wind direction is the target wind direction into the trained Bayesian network; calculating the emission probability distribution of each potential pollution source based on the posterior probability of the trained Bayesian network and the current data of the sensor when the wind direction is the target wind direction; sorting the emission values ​​corresponding to the highest probability point of the emission probability distribution of each potential pollution source; and determining the potential pollution source corresponding to the maximum emission value as the pollution source.

[0105] For example, when it is determined that the current wind direction is northwest wind, the current data of the sensor at northwest wind (for example, Figure 3The method comprises the following steps: inputting the current data of sensor Y in the northwest wind into a trained Bayesian network corresponding to the northwest wind; calculating the emission probability distribution of each potential pollution source based on the posterior probability of the trained Bayesian network corresponding to the northwest wind and the current data of the sensor when the northwest wind is blowing; sorting the emission values ​​corresponding to the highest probability point of each emission probability distribution; and determining the potential pollution source corresponding to the maximum emission value as the pollution source.

[0106] For example, when the current wind direction is determined to be northeast wind, the current data of the sensor when the wind is northeast wind (for example, when the wind is northeast wind, Figure 3 The method comprises the following steps: inputting the current data of sensor X and the current data of sensor Y in the northeast wind into a trained Bayesian network corresponding to the northeast wind; calculating the emission probability distribution of each potential pollution source based on the posterior probability of the trained Bayesian network corresponding to the northeast wind and the current sensor data during the northeast wind; sorting the emission values ​​corresponding to the highest probability point in each emission probability distribution; and determining the potential pollution source corresponding to the maximum emission value as the pollution source.

[0107] Figure 4 An exemplary schematic diagram of determining pollution sources according to an embodiment of the present invention.

[0108] exist Figure 4 In the graph, the horizontal axis represents emission values, the vertical axis represents probability, and the current wind direction is northwest. Curve 41 is the emission probability distribution curve for potential pollution source A; curve 42 is the emission probability distribution curve for potential pollution source B; and curve 43 is the emission probability distribution curve for potential pollution source C.

[0109] The emission value corresponding to the highest probability point on curve 41 (i.e., the emission value of potential pollution source A) is MA, the emission value corresponding to the highest probability point on curve 42 (i.e., the emission value of potential pollution source B) is MB, and the emission value corresponding to the highest probability point on curve 43 (i.e., the emission value of potential pollution source C) is MC. As can be seen, MC is greater than MB, and MB is greater than MA, so the pollution source is determined to be potential pollution source C.

[0110] exist Figure 4 In the embodiment of the present invention, a specific example of determining the pollution source is described by taking the current wind direction as northwest wind as an example. Those skilled in the art will appreciate that a similar method can be used to determine the pollution source for other wind directions, and the embodiments of the present invention will not be described in detail.

[0111] against Figure 2 The industrial park topology shown in the figure is Figure 5 This is an exemplary flow chart of a method for determining pollution sources in an industrial park according to an embodiment of the present invention.

[0112] like Figure 5 As shown, the method includes:

[0113] Step 501: Based on the positional relationship between the sensor and the potential pollution source in the industrial park, determine the causal relationship between the sensor and the potential pollution source in a predetermined target wind direction.

[0114] Assuming that the predetermined target wind directions are northeast, northwest, southeast, and southwest, the determined causal relationships include: (1) When the wind direction is northwest, the pollutants emitted by potential pollution source A, potential pollution source B, and potential pollution source C will be monitored by sensor Y. (2) When the wind direction is northeast, the pollutants emitted by potential pollution source A will be monitored by sensor X; the pollutants emitted by potential pollution source B will be monitored by sensors X and Y; and the pollutants emitted by potential pollution source D will be monitored by sensor Y. (3) When the wind direction is southwest, the pollutants emitted by potential pollution source A will be monitored by sensor Z; the pollutants emitted by potential pollution source B will be monitored by sensor Z; the pollutants emitted by potential pollution source C will be monitored by sensor Z; and the pollutants emitted by potential pollution source D will be monitored by sensor Z. (4) When the wind direction is southeast, the pollutants emitted by potential pollution source A will be monitored by sensor X; the pollutants emitted by potential pollution source C will be monitored by sensor X; and the pollutants emitted by potential pollution source D will be monitored by sensors X and Z.

[0115] Step 502: Establish a Bayesian network corresponding to the target wind direction and including the causal relationship.

[0116] Specifically, a Bayesian network 31 corresponding to northeast wind, a Bayesian network 32 corresponding to northwest wind, a Bayesian network 33 corresponding to southwest wind, and a Bayesian network 34 corresponding to southeast wind are established.

[0117] The Bayesian network 32 corresponding to the northwest wind includes nodes A, B, C, and Y. Node A represents potential pollution source A, node B represents potential pollution source B, node C represents potential pollution source C, and node Y represents sensor Y. Nodes A, B, and C each point to sensor Y.

[0118] The Bayesian network 31 corresponding to the northeast wind includes nodes A, B, D, X, and Y. Node A represents potential pollution source A, node B represents potential pollution source B, node C represents potential pollution source C, node D represents potential pollution source D, node X represents sensor X, and node Y represents sensor Y. Nodes A and B point to sensor X, respectively, and nodes B and D point to sensor Y, respectively.

[0119] The Bayesian network 33 corresponding to the southwest wind includes nodes A, B, C, D, and Z. Node A represents potential pollution source A, node B represents potential pollution source B, node C represents potential pollution source C, node D represents potential pollution source D, and node Z represents sensor Z. Nodes A, B, C, and D each point to sensor Z.

[0120] The Bayesian network 34 corresponding to the southeast wind includes nodes A, C, D, X, and Z. Node A represents potential pollution source A, node C represents potential pollution source C, node D represents potential pollution source D, node X represents sensor X, and node Z represents sensor Z. Nodes A, C, and D each point to sensor X, and node D also points to sensor Z.

[0121] Step 503: Acquire actual historical data of the sensors at respective target wind directions.

[0122] Here, actual historical data of the sensor during northeast wind, actual historical data of the sensor during northwest wind, actual historical data of the sensor during southeast wind, and actual historical data of the sensor during southwest wind are obtained respectively.

[0123] Step 504: Using the actual historical data of the sensors at the respective wind directions, train the Bayesian network corresponding to the respective wind directions.

[0124] Here, the Bayesian network 31 corresponding to northeast wind is trained using actual historical sensor data during northeast wind conditions. Preferably, the prior probability distribution of Bayesian network 31 corresponding to northeast wind conditions is first determined. Then, based on the prior probability distribution of Bayesian network 31 corresponding to northeast wind conditions and the actual historical sensor data during northeast wind conditions, the posterior probability distribution of Bayesian network 31 corresponding to northeast wind conditions is calculated using a belief propagation algorithm. More preferably, during each iteration of the belief propagation algorithm, the wind speed of the northeast wind is used as a parameter influencing the monitoring values ​​of sensor X and sensor Y in Bayesian network 31. The greater the wind speed, the smaller the monitoring values ​​of sensor X and sensor Y; and the smaller the wind speed, the larger the monitoring values ​​of sensor X and sensor Y.

[0125] The Bayesian network 32 corresponding to the northwest wind is trained using actual historical sensor data during northwest wind conditions. Preferably, the prior probability distribution of the Bayesian network 32 corresponding to the northwest wind is first determined. Then, based on the prior probability distribution of the Bayesian network 32 corresponding to the northwest wind and the actual historical sensor data during northwest wind conditions, the posterior probability distribution of the Bayesian network 32 corresponding to the northwest wind is calculated using a belief propagation algorithm. More preferably, during each iteration of the belief propagation algorithm, the northwest wind speed is used as a parameter influencing the monitoring value of sensor Y in the Bayesian network 32. Specifically, the greater the wind speed, the smaller the monitoring value of sensor Y; and the smaller the wind speed, the larger the monitoring value of sensor Y.

[0126] The Bayesian network 33 for southwest wind is trained using actual historical sensor data during southwest wind conditions. Preferably, the prior probability distribution of the Bayesian network 33 for southwest wind is first determined. Then, based on the prior probability distribution of the Bayesian network 33 for southwest wind and the actual historical sensor data during southwest wind conditions, the posterior probability distribution of the Bayesian network 33 for southwest wind is calculated using a belief propagation algorithm. More preferably, during each iteration of the belief propagation algorithm, the southwest wind speed is used as a parameter influencing the monitoring value of sensor Z in the Bayesian network 33. Specifically, the greater the wind speed, the smaller the monitoring value of sensor Z; and the smaller the wind speed, the larger the monitoring value of sensor Z.

[0127] The Bayesian network 34 corresponding to the southeast wind is trained using actual historical sensor data during southeast winds. Preferably, the prior probability distribution of the Bayesian network 34 corresponding to the southeast wind is first determined. Then, based on the prior probability distribution of the Bayesian network 34 corresponding to the southeast wind and the actual historical sensor data during southeast winds, the posterior probability distribution of the Bayesian network 34 corresponding to the southeast wind is calculated using a belief propagation algorithm. More preferably, during each iteration of the belief propagation algorithm, the southeast wind speed is used as a parameter influencing the monitoring values ​​of sensors X and Z in the Bayesian network 34. Specifically, the greater the wind speed, the smaller the monitoring values ​​of sensors X and Z; and the smaller the wind speed, the larger the monitoring values ​​of sensors X and Z.

[0128] Step 505: Determine the pollution source from potential pollution sources based on the current data of the sensor at the current wind direction and the trained Bayesian network corresponding to the current wind direction.

[0129] For example, assuming the current wind direction is northwest, select the Bayesian network 32 corresponding to the northwest wind. Then, input the current sensor data for the current wind direction (i.e., northwest wind) into the Bayesian network 32 corresponding to the northwest wind (preferably, the current northwest wind speed is also input into the Bayesian network 32 corresponding to the northwest wind). The Bayesian network 32 corresponding to the northwest wind outputs the emission probability distributions for potential pollution sources A, B, and C. Next, sort the emission values ​​corresponding to the highest probability points for potential pollution sources A, B, and C, and determine the potential pollution source corresponding to the maximum emission value as the pollution source.

[0130] For another example, assume that the current wind direction is northeast. Then, the Bayesian network 31 corresponding to the northeast wind is selected, and the current sensor data at the current wind direction (i.e., northeast wind) is input into the Bayesian network 31 corresponding to the northeast wind (preferably, the current northeast wind speed is also input into the Bayesian network 31 corresponding to the northeast wind). The Bayesian network 31 corresponding to the northeast wind outputs the emission probability distribution of potential pollution source A, potential pollution source B, and potential pollution source D. Next, the emission values ​​corresponding to the highest probability point of each emission probability distribution are sorted; the potential pollution source corresponding to the maximum emission value is determined as the pollution source.

[0131] Based on the above description, an embodiment of the present invention further proposes a device for determining a pollution source.

[0132] Figure 6 This is an exemplary structural diagram of an apparatus for determining pollution sources according to an embodiment of the present invention.

[0133] like Figure 6 As shown, the device 600 for determining a pollution source includes:

[0134] A relationship determination module 601 is configured to determine a causal relationship between the sensor and the potential pollution source in the target wind direction based on the positional relationship between the sensor and the potential pollution source;

[0135] A network establishing module 602 is configured to establish a Bayesian network corresponding to the target wind direction and including the causal relationship;

[0136] A training module 603 is configured to train the Bayesian network based on historical sensor data when the wind direction is the target wind direction;

[0137] The determination module 604 is configured to determine a pollution source from the potential pollution sources based on the current sensor data when the wind direction is the target wind direction and the trained Bayesian network.

[0138] In one embodiment, the training module 603 is used to determine the prior probability distribution of the Bayesian network; based on the prior probability distribution of the Bayesian network and the sensor historical data when the wind direction is the target wind direction, the posterior probability distribution of the Bayesian network is calculated using a belief propagation algorithm.

[0139] In one embodiment, the training module 603 is configured to use the wind speed when the wind direction is the target wind direction as a parameter affecting the sensor monitoring value during each iterative calculation of the belief propagation algorithm, wherein the greater the wind speed, the smaller the sensor monitoring value.

[0140] In one embodiment, the determination module 604 is used to input the current data of the sensor when the wind direction is the target wind direction into the trained Bayesian network; based on the posterior probability of the Bayesian network and the current data of the sensor when the wind direction is the target wind direction, calculate the emission probability distribution of each potential pollution source; sort the emission values ​​corresponding to the highest probability point of the emission probability distribution of each potential pollution source; and determine the potential pollution source corresponding to the maximum emission value as the pollution source.

[0141] In one embodiment, the training module 603 trains the Bayesian network based on actual historical data of the sensor when the wind direction is the target wind direction; or trains the Bayesian network based on simulated historical data of the sensor when the wind direction is the target wind direction.

[0142] Figure 7 This is an exemplary structural diagram of an apparatus for determining pollution sources according to an embodiment of the present invention.

[0143] exist Figure 7 In the embodiment, the device 700 for determining the pollution source includes a memory 702 and a processor 701; the memory 702 stores an application program that can be executed by the processor 701, which is used to enable the processor 701 to execute the method for determining the pollution source as described in any of the above items.

[0144] Specifically, the memory 702 may be implemented as various storage media, such as an electrically erasable programmable read-only memory (EEPROM), a flash memory, or a programmable read-only memory (PROM). The processor 701 may be implemented as including one or more central processing units (CPUs) or one or more field programmable gate arrays (FPGAs), wherein the FPGAs integrate one or more CPU cores. Specifically, the CPU or CPU core may be implemented as a CPU, an MCU, a DSP, or the like.

[0145] It should be noted that not all steps and modules in the above processes and structure diagrams are required, and certain steps or modules can be omitted based on actual needs. The execution order of the steps is not fixed and can be adjusted as needed. The division of the modules is merely for the convenience of describing the functional division adopted. In actual implementation, a module can be implemented by multiple modules, and the functions of multiple modules can be implemented by the same module. These modules can be located in the same device or in different devices.

[0146] The hardware modules in each embodiment can be implemented mechanically or electronically. For example, a hardware module may include a specially designed permanent circuit or logic device (such as a dedicated processor, such as an FPGA or ASIC) for performing a specific operation. The hardware module may also include a programmable logic device or circuit (such as a general-purpose processor or other programmable processor) temporarily configured by software to perform a specific operation. As for whether to implement the hardware module mechanically, or using a dedicated permanent circuit, or using a temporarily configured circuit (such as configured by software), it can be decided based on cost and time considerations.

[0147] The present invention also provides a machine-readable storage medium, storing instructions for causing a machine to perform a method as described herein. Specifically, a system or device equipped with a storage medium can be provided, on which a software program code for realizing the function of any one of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or device is made to read out and execute the program code stored in the storage medium. In addition, the operating system etc. operated on the computer can also be made to complete part or all of the actual operations by instructions based on the program code. The program code read out from the storage medium can also be written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion unit connected to the computer, and then the CPU etc. installed on the expansion board or expansion unit are made to perform part and all of the actual operations based on the instructions of the program code, thereby realizing the function of any one of the embodiments described above.

[0148] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-Rs, CD-RWs, DVD-ROMs, DVD-RAMs, DVD-RWs, DVD+RWs), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer or a cloud via a communication network.

[0149] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A method (100) for determining a pollution source, characterized in that: include: Based on the positional relationship between the sensor and the potential pollution source, determine the causal relationship between the sensor and the potential pollution source in the target wind direction (101); The causal relationship refers to the fact that pollutants emitted by the potential pollution source in the target wind direction are detected by the sensor; Establishing (102) a Bayesian network corresponding to the target wind direction and including the causal relationship; Training (103) the Bayesian network based on historical sensor data when the wind direction is the target wind direction; Determine (104) a pollution source from the potential pollution sources based on current sensor data when the wind direction is the target wind direction and the trained Bayesian network; The Bayesian network training (103) based on the sensor historical data when the wind direction is the target wind direction includes: determining a prior probability distribution of the Bayesian network; Calculating the posterior probability distribution of the Bayesian network using a belief propagation algorithm based on the prior probability distribution of the Bayesian network and the historical data of the sensor when the wind direction is the target wind direction; Wherein, the determining (104) of the pollution source from the potential pollution sources based on the current sensor data when the wind direction is the target wind direction and the trained Bayesian network includes: Inputting the current data of the sensor when the wind direction is the target wind direction into the trained Bayesian network; Calculating an emission probability distribution of each potential pollution source based on the posterior probability of the trained Bayesian network and current sensor data when the wind direction is the target wind direction; Sort the emission values ​​corresponding to the highest probability point of the emission probability distribution of each potential pollution source; The potential pollution source corresponding to the maximum emission value is determined as the pollution source.

2. The method (100) for determining a pollution source according to claim 1, characterized in that: The calculation of the posterior probability distribution of the Bayesian network using the belief propagation algorithm includes: in each iterative calculation of the belief propagation algorithm, the wind speed when the wind direction is the target wind direction is used as a parameter affecting the sensor monitoring value, wherein the greater the wind speed, the smaller the sensor monitoring value.

3. The method (100) for determining a pollution source according to any one of claims 1-2, characterized in that: The Bayesian network training (103) based on the sensor historical data when the wind direction is the target wind direction includes: Training the Bayesian network based on actual historical data of the sensor when the wind direction is the target wind direction; or The Bayesian network is trained based on sensor simulation historical data when the wind direction is the target wind direction.

4. A device (600) for determining a pollution source, characterized in that: include: A relationship determination module (601) is configured to determine a causal relationship between the sensor and the potential pollution source in a target wind direction based on a positional relationship between the sensor and the potential pollution source; wherein the causal relationship refers to the pollutants emitted by the potential pollution source in the target wind direction being detected by the sensor; A network establishment module (602) is used to establish a Bayesian network corresponding to the target wind direction and including the causal relationship; A training module (603) is used to train the Bayesian network based on historical sensor data when the wind direction is the target wind direction; The training module (603) is used to determine the prior probability distribution of the Bayesian network; based on the prior probability distribution of the Bayesian network and the historical data of the sensor when the wind direction is the target wind direction, the posterior probability distribution of the Bayesian network is calculated using a belief propagation algorithm; A determination module (604) is configured to determine a pollution source from the potential pollution sources based on current sensor data when the wind direction is the target wind direction and the trained Bayesian network; The determination module (604) is used to input the current data of the sensor when the wind direction is the target wind direction into the trained Bayesian network; calculate the emission probability distribution of each potential pollution source based on the posterior probability of the Bayesian network and the current data of the sensor when the wind direction is the target wind direction; sort the emission values ​​corresponding to the highest probability point of the emission probability distribution of each potential pollution source; and determine the potential pollution source corresponding to the maximum emission value as the pollution source.

5. The device (600) for determining a pollution source according to claim 4, characterized in that: The training module (603) is used to use the wind speed when the wind direction is the target wind direction as a parameter affecting the sensor monitoring value during each iterative calculation of the belief propagation algorithm, wherein the greater the wind speed, the smaller the sensor monitoring value.

6. The device (600) for determining a pollution source according to any one of claims 4-5, characterized in that: The training module (603) is used to train the Bayesian network based on actual historical data of the sensor when the wind direction is the target wind direction; or to train the Bayesian network based on simulated historical data of the sensor when the wind direction is the target wind direction.

7. A device (700) for determining a pollution source, characterized in that: comprising a processor (701) and a memory (702); The memory (702) stores an application program that can be executed by the processor (701), and is used to enable the processor (701) to execute the method (100) for determining a pollution source according to any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that Computer-readable instructions are stored therein, and the computer-readable instructions are used to execute the method (100) for determining a pollution source according to any one of claims 1 to 3.

Images