Multi-target deployment method of pollution monitoring sensors in water supply network based on pollution probability

Through a multi-objective optimization method that considers the probability changes of pollution events, the sensor layout of the water supply pipeline network is optimized, and the problems of low detection rate and redundancy in the existing technology are solved, and priority detection of high-impact pollution events and impact of pipeline networks are realized. reduce.

CN113887000BInactive Publication Date: 2024-11-08CHANGJIANG RIVER SCI RES INST CHANGJIANG WATER RESOURCES COMMISSION +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111073010.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-09-14
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing water supply pipeline sensor optimization layout method fails to effectively consider the changes in node pollution probability, resulting in low detection rate of pollution events with high pollution consequences, redundant information of the sensor network, and serious pollution events are easily ignored when the number of sensors is limited.

Method used

Using a multi-objective optimization method based on the probability changes of pollution events, the pollution probability of each node is calculated by simulating the pollution impact matrix and detection matrix, and the multi-criteria decision-making cluster analysis method is used to optimize the sensor layout to ensure priority detection of high-impact pollution events. .

Benefits of technology

The real detection rate of pollution incidents in the sensor layout scheme is improved, the impact of pollutant intrusion on the pipeline network is reduced, and serious pollution incidents can be effectively detected when the number of sensors is limited, reducing the redundant information of the sensor network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113887000B_ABST
    Figure CN113887000B_ABST
Patent Text Reader

Abstract

The present invention provides a multi-objective arrangement method of pollution monitoring sensors for a water supply network based on pollution probability, comprising: step 1, simulating the situation of pollutant invasion at each node of the water supply network to obtain the influence of various pollution events on the water supply network; step 2, obtaining the pollution probability of each node according to the influence of pollution at each node on the water supply network; step 3, obtaining the Pareto front solution under each pollution probability based on a multi-objective optimization method; step 4, clustering and sorting the Pareto front solutions under various pollution probabilities using a multi-criteria decision clustering analysis method to obtain the optimal solutions of each cluster under different situations, and comparing and analyzing various optimal solutions using the influence of pollutant invasion on the network. The present invention takes into account the change of node pollution probability when optimizing the arrangement of sensors, and can obtain the sensor arrangement solution with the least influence on the network when pollutants invade.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to an optimized arrangement of sensors for a city water supply network, and in particular to a multi-target arrangement method of pollution monitoring sensors for a water supply network based on pollution probability. Background Art

[0002] The water supply network is an important urban infrastructure that provides safe drinking water for residents. Due to its decentralized spatial structure, each node of the water supply network is prone to the risk of pollutant intrusion. In order to ensure the safety of water supply, it is necessary to establish a safe and effective monitoring system for the water supply network to effectively identify pollutant intrusion. The accuracy of the monitoring system is closely related to the number and location of sensor layout. Therefore, optimizing the layout of sensors is a prerequisite for effectively identifying pollutant intrusion.

[0003] The optimal arrangement of sensors involves the selection of the number and location of sensors. Increasing the number of sensors can shorten the detection time of pollutant intrusion and improve the positioning accuracy. Although increasing the number of sensors can significantly improve the performance of the monitoring system, in engineering applications, considering the installation and maintenance costs of sensors, it is usually necessary to make a trade-off between the number of sensors and the performance of the monitoring system. To address this problem, most methods use multi-objective optimization methods to arrange sensors, such as making the detection time shorter, detecting more pollution events, using fewer sensors, and reducing the impact of pollution events.

[0004] Due to the different population densities and user attributes in different areas of the water supply network, the probability of pollutant invasion at each node is significantly different. Due to the limited number of sensors, it is usually difficult to complete the detection of all pollution events. Considering the changes in node pollution probability when optimizing the sensor layout helps to establish an effective monitoring system. However, current sensor optimization layout methods rarely consider the changes in node pollution probability. There are a few methods that consider the changes in node pollution probability when optimizing the sensor layout, such as obtaining pollution probability based on node demand and based on the location of special users, but these methods do not take into account that the risk level of pollution at each node in the water supply network is different. If the impact of pollutant invasion at each node is not considered when optimizing the layout of sensors (which will result in the same pollution probability at each node), the detection of some pollution events with high pollution consequences will be ignored, which is obviously contrary to the original intention of arranging sensors.

[0005] In addition, the sensor optimization layout schemes obtained by these methods usually place sensors at the network nodes that are most sensitive to water quality changes, which is called "sensor network sensitivity maximization". However, some highly sensitive points in the water supply network are usually scattered in a relatively concentrated area. This clustering effect leads to information redundancy in the sensor network. Summary of the invention

[0006] In view of the above problems existing in the prior art, the present invention proposes a multi-objective optimization layout method for sensors that takes into account the change in the probability of pollution events. When optimizing the layout of sensors, the change in the probability of node pollution is taken into account, and a sensor layout plan with the least impact on the pipeline network when pollutants invade can be obtained.

[0007] To achieve the above objectives, the present invention adopts the following technical solutions:

[0008] A multi-objective optimization arrangement method for sensors considering the probability change of pollution events comprises the following steps:

[0009] Step 1: simulate the situation of pollutant invasion at each node of the water supply network to obtain the impact of various pollution events on the water supply network;

[0010] Step 2: According to the impact of pollution at each node on the water supply network, the pollution probability of each node is obtained;

[0011] Step 3: Based on the multi-objective optimization method, the Pareto frontier solution under each pollution probability is obtained;

[0012] Step 4: Use the multi-criteria decision-making cluster analysis method to cluster and sort the Pareto front solutions under various pollution probabilities, obtain the optimal solution for each cluster under different situations, and compare and analyze various optimal solutions based on the impact of pollutant invasion on the pipeline network.

[0013] Furthermore, the impact on the water supply network in step 1 includes the impact on the network nodes and pipelines.

[0014] Furthermore, the step 1 specifically includes:

[0015] Step 1.1: Simulate the situation of pollutant invasion at each node of the pipeline network to obtain the pollution impact matrix when pollutants invade each node;

[0016] The pollution impact matrix includes the node impact matrix and the pipeline impact matrix, which refers to the impact of pollutant invasion on the pipeline nodes and pipelines at each node of the pipeline network. Assuming that the pollutant invasion occurs at the pipeline node, the number of pipeline nodes is N. n Based on the hydraulic model of the water supply network, the influence matrix I of the network nodes and pipelines after the intrusion of pollutants at each node of the network is obtained. t (i, j) and I t (i,p), as shown in equations (1) and (2) respectively:

[0017]

[0018]

[0019] In the formula, and They represent the impact of pollutant intrusion on node j and pipeline p at time t, respectively. t=Δt, 2Δt, …, AΔt, Δt represents the water quality simulation time interval, AΔt represents the total simulation time. Assuming that the pollutant intrusion occurs at time t=0, the pollution is continuously injected during the entire simulation time period. j=1, 2, …, N n , N n is the total number of nodes in the water supply network, p=1,2,…,N p , N p is the total number of water supply network pipes. If the pollutant concentration at node j is greater than zero at time t, then otherwise, If the pollutant concentration at pipe p is greater than zero at time t, then otherwise,

[0020] Step 1.2: Simulate the situation of pollutant invasion at each node of the pipe network to obtain the pollution detection matrix when pollutant invasion occurs at each node;

[0021] The network hydraulic model is used to simulate the intrusion of pollutants at each node of the water supply network, and the pollution detection matrix D under different pollution situations is obtained. t (i, j), as shown in formula (3):

[0022]

[0023] In the formula, It represents the detection of the sensor at node j after the pollution intrusion at node i at time t. If the pollutant concentration at node j at time t is greater than or equal to C min ,but otherwise, t=Δt′,2Δt′,…,BΔt′,Δt′ is the sampling time interval of the sensor, BΔt′ is the longest detection time of the sensor network for pollutant intrusion on a single node, C min This is the lowest concentration that the sensor can detect.

[0024] Furthermore, five pollution probabilities are obtained in step 2: the pollution probability based on the number of affected nodes, the pollution probability based on the water demand, the pollution probability based on the number of affected pipes, the pollution probability based on the pollutant diffusion distance, and the joint pollution probability that comprehensively considers the above four influences.

[0025] Furthermore, the step 2 specifically includes:

[0026] Step 2.1: Based on the number of impacts on the pipeline network nodes caused by pollutant intrusion at each node, the probability P1 of pollution occurring at each node of the pipeline is obtained:

[0027]

[0028] Where P1(i) represents the probability of pollutant intrusion at node i, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij The impact of pollution in node i on node j;

[0029] Step 2.2: Based on the impact of pollutant intrusion on water demand at each node of the network, the probability P2 of pollution occurring at each node of the network is obtained:

[0030]

[0031] Where P2(i) is the pollution probability of node i based on the affected water demand, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij is the impact of pollution in node i on node j; d jt is the water demand of node j at time t; It represents the total water demand of the water supply network during the simulation period;

[0032] Step 2.3: Based on the number of pipelines affected by the intrusion of pollutants at each node in the pipeline network, the probability P3 of pollution occurring at each node in the pipeline network is obtained:

[0033]

[0034] Where P3(i) is the contamination probability of node i based on the number of affected pipelines, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ip is the impact of pollution on pipeline P after node i occurs;

[0035] Step 2.4: According to the diffusion distance of pollutants after pollution occurs at each node, the probability P4 of pollution occurring at each node in the pipeline network is obtained:

[0036]

[0037] Where P4(i) represents the pollution probability of node i based on the pollution diffusion distance, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ipis the impact of pollution on pipeline P at node i; L p is the length of the pipe P;

[0038] Step 2.5: Combine the four pollution probabilities to obtain the joint probability P5 of pollutant intrusion at each node in the pipeline network:

[0039] P5(i)=w1·P1(i)+w2·P2(i)+w3·P3(i)+w4·P4(i); (8)

[0040] Where P5(i) represents the joint pollution probability based on the four pollution probabilities, w1, w2, w3 and w4 are the weights of each pollution probability, satisfying w1+w2+w3+w4=1.

[0041] Furthermore, in step 3, based on the multi-objective optimization method, four optimization objective functions are used to obtain the Pareto frontier solution under each pollution probability: shortest detection time, maximum number of pollution event detections, minimum number of sensors, and maximum joint information entropy.

[0042] Furthermore, the step 3 specifically includes:

[0043] Step 3.1: Consider optimizing the objective function F1 to minimize the pollutant intrusion detection time;

[0044]

[0045] Where F1(s) represents the detection time of pollutant intrusion at each node of the pipe network by the sensor arrangement scheme s, γ(i,t) represents the shortest detection time of pollutant intrusion at node i by the sensor network, if no sensor alarms the pollutant intrusion at node i, then γ(i,t)=BΔt′, α is an indicator, if the sensor network can detect the pollutant intrusion at node i, then α=1, otherwise α=0, p(i) is the probability of pollution at node i;

[0046] Step 3.2: Consider optimizing the objective function F2 to maximize the number of pollution event detections;

[0047]

[0048] Where p(i) is the pollution probability of node i;

[0049] Step 3.3: Consider optimizing the objective function F3 to minimize the number of sensors:

[0050] F3=min(N s )

[0051] Where N s is the number of sensors placed in the network;

[0052] Step 3.4: Consider optimizing the objective function F4 to maximize the joint information entropy of the sensor network:

[0053]

[0054] Where j = 1,…,N I,i , N I,i is the sensor S at node i i The number of intervals divided for various pollution events, is the probability of the pollution event being distributed in the jth interval, m i is the number of pollution events in the jth interval;

[0055] Step 3.5: Use the NSGA-II multi-objective optimization model to solve and obtain the Pareto frontier solutions under five pollution probabilities.

[0056] Furthermore, step 3.5 is as follows: first, a group of pipe network nodes are randomly selected as sensor deployment nodes, and the pollution detection matrix D t (i,j) Select the corresponding row and add it to the leakage detection matrix of the sensor layout S Then, the objective function value and crowding distance of the sensor layout scheme are calculated; then, a new set of pipe network nodes are selected as to improve the objective function value and evolve to the next generation; finally, the pipe network nodes are continuously selected until no better solution than the existing generation can be obtained, and the sensor optimization layout scheme and its objective function value are output in the form of Pareto front solution.

[0057] Furthermore, the step 4 specifically includes:

[0058] Step 4.1: Cluster and sort the Pareto solutions under each pollution probability based on multi-criteria decision analysis to obtain the optimal sensor solution under various circumstances;

[0059] Step 4.2: Calculate the impact of pollutant intrusion on the water supply network at each node of the network at each time for each scheme, and compare and analyze various schemes; the impact caused by pollutant intrusion is shown in equations (12)-(15):

[0060]

[0061]

[0062]

[0063]

[0064] Where I1(S), I2(S), I3(S) and I4(S) respectively represent the impact of pollutant intrusion on the number of nodes, water demand, number of pipes and pollutant diffusion distance at each node of the water supply network under the sensor layout scheme S, and N S (t) represents the number of pollution events that are not detected by Scheme S at time t, N(t) represents the total number of pollution events, p(i) represents the probability of pollutant intrusion at node i, n(i), d(i), Pipe n (i) and l(i) represent the number of nodes affected, the affected water demand, the number of affected pipes and the pollutant diffusion distance at time t after the pollutant invades node i, respectively. For a scheme, the smaller the various impacts after the pollutant invades, the better the scheme is.

[0065] The present invention obtains the pollution probability of each node of the pipeline network according to the impact of pollutant invasion on the pipeline network when optimizing the arrangement of sensors. When comparing and screening the sensor arrangement schemes, the change of pollution probability of each node of the pipeline network is considered, so as to ensure that the selected sensor arrangement scheme can give priority to the detection of pollution events that cause serious pollution consequences, so that the impact of pollutant invasion on the pipeline network at each node of the pipeline network is smaller. Case tests show that when considering the change of node pollution probability, the true detection rate of pollution events of each sensor arrangement scheme is higher, pollution events with serious consequences are easier to be detected, and the impact on the pipeline network after the invasion of pollutants is smaller. When the existing method arranges the sensors, the change of pollution probability of each pollution event is not considered. When the number of sensors is limited, the sensor arrangement scheme obtained may ignore some pollution events with serious pollution consequences. The method proposed by the present invention distinguishes pollution events according to the consequences of pollution events, and gives priority to the detection of pollution events with serious pollution consequences, avoiding possible omissions, thereby ensuring that the impact of pollutant invasion on the pipeline network at each node of the pipeline network is minimized when the number of sensors is limited. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flow chart of one embodiment of the multi-target arrangement method of pollution monitoring sensors for water supply network based on pollution probability of the present invention;

[0067] Figure 2 Schematic diagram of clustering partitioning for multi-criteria decision analysis;

[0068] Figure 3 This is a schematic diagram of the exemplary pipe network C-town of the present invention;

[0069] Figure 4 The pollution probability distribution of each node of the pipeline network is obtained based on the impact of pollutants invading each node of the pipeline network on the pipeline network in the present invention;

[0070] Figure 5 is the Pareto solution clustering distribution under various pollution probabilities obtained under scenario 1 of the present invention;

[0071] Figure 6 The Pareto solution clustering distribution under various pollution probabilities obtained under scenario 2 of the present invention;

[0072] Figure 7 The Pareto solution clustering distribution under various pollution probabilities obtained under scenario 3 of the present invention;

[0073] Figure 8 Schematic diagram of detection rate, detection time, number of sensors and joint information entropy of various schemes under different situations of the present invention: (a) situation 1; (b) situation 2; (c) situation 3;

[0074] Fig. 9 The impact of each scheme on the water supply network at different pollution probabilities under situation 1 at each time: (a) P1; (b) P2; (c) P3; (d) P4; (e) P5;

[0075] Fig.10 The impact of various schemes on the water supply network at different times under different pollution probabilities under situation 2: (a) P1; (b) P2; (c) P3; (d) P4; (e) P5;

[0076] Fig.11 The impact of each scheme on the water supply network at different times under different pollution probabilities under situation 3: (a) P1; (b) P2; (c) P3; (d) P4; (e) P5. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] See also Figure 1 The embodiment of the present invention provides a multi-objective optimization arrangement method of sensors considering the probability change of pollution events, comprising the following steps:

[0079] Step 1: Simulate the situation of pollutant invasion at each node of the water supply network to obtain the impact of various pollution events on the water supply network. The impact on the water supply network includes the impact on the network nodes (number of nodes and water demand) and pipelines (number of pipelines and pollutant diffusion distance);

[0080] Step 2: According to the impact of pollution at each node on the water supply network, the pollution probability of each node is obtained. Specifically, five types of pollution probabilities are obtained: the pollution probability based on the number of affected nodes, the pollution probability based on the impact on water demand, the pollution probability based on the impact on the number of pipelines, the pollution probability based on the diffusion distance of pollutants, and the combined pollution probability that comprehensively considers the above four impacts;

[0081] Step 3: Based on the multi-objective optimization method, the Pareto frontier solutions under various pollution probabilities are obtained. Specifically, four optimization objective functions are used: shortest detection time, maximum number of pollution event detections, minimum number of sensors, and maximum joint information entropy, to obtain the Pareto frontier solutions of the sensor optimization layout schemes under five pollution probabilities.

[0082] Step 4: Use the multi-criteria decision-making cluster analysis method to cluster and sort the Pareto front solutions under various pollution probabilities, obtain the optimal solution for each cluster under different situations, and compare and analyze various optimal solutions based on the impact of pollutant invasion on the pipeline network.

[0083] Further, the step 1 specifically includes:

[0084] Step 1.1: Simulate the situation of pollutant invasion at each node of the pipeline network to obtain the pollution impact matrix when pollutants invade each node;

[0085] When a pollutant invades a node in the pipeline network, the pollutant spreads to various nodes and pipelines along the pipeline, thus affecting other parts of the pipeline network. At the same time, if the pollutant spreads to a node where a sensor is deployed, as long as the pollutant concentration is greater than the alarm concentration of the sensor, the pollutant invasion event will be detected. Obviously, the more nodes and pipelines a single node invades, the more serious the pollution consequences caused by the pollution event.

[0086] The pollution impact matrix includes the node impact matrix and the pipeline impact matrix, which refers to the impact of pollutant invasion on the pipeline nodes and pipelines at each node of the pipeline network. Assume that the pollutant invasion occurs at the pipeline node, and the number of pipeline nodes is N. n Based on the hydraulic model of the water supply network, the influence matrix I of the network nodes and pipelines after the intrusion of pollutants at each node of the network is obtained. t (i, j) and I t (i,p), as shown in equations (1) and (2) respectively:

[0087]

[0088]

[0089] In the formula, and They represent the impact of pollutant intrusion on node j and pipeline p at time t, respectively. t=Δt, 2Δt, …, AΔt, Δt represents the water quality simulation time interval, AΔt represents the total simulation time. Assuming that the pollutant intrusion occurs at time t=0, the pollution is continuously injected during the entire simulation time period. j=1, 2, …, N n , N n is the total number of nodes in the water supply network, p=1,2,…,N p , N p is the total number of water supply network pipes. If the pollutant concentration at node j is greater than zero at time t, then otherwise, If the pollutant concentration at pipe p is greater than zero at time t, then otherwise, The present invention assumes that during the entire simulation time period, as long as there is one If the pollutant intrusion at node i occurs, it will affect node i (pipeline p);

[0090] Step 1.2: Simulate the situation of pollutant invasion at each node of the pipe network to obtain the pollution detection matrix when pollutant invasion occurs at each node;

[0091] Due to the topological structure of the pipe network itself, the diffusion paths of pollutants are different after the intrusion of pollutants at different nodes of the pipe network. A certain node of the pipe network is usually only affected by the intrusion of pollutants at some nodes, and the time for pollutants to diffuse to the node after the intrusion of pollutants at these nodes is also different. When optimizing the layout of sensors, it is necessary to select the location and number of sensors based on the detection of different pollution events by sensors at different nodes (whether they can be detected and the detection time) to obtain the optimal solution. Therefore, the pipe network hydraulic model is used to simulate the intrusion of pollutants at each node of the water supply pipe network, and the pollution detection matrix D under different pollution situations is obtained. t (i, j), as shown in formula (3):

[0092]

[0093] In the formula, It represents the detection of the sensor at node j after the pollution intrusion at node i at time t. If the pollutant concentration at node j at time t is greater than or equal to C min ,but otherwise, t=Δt′,2Δt′,…,BΔt′,Δt′ is the sampling time interval of the sensor, BΔt′ is the longest detection time of the sensor network for pollutant intrusion on a single node. In order to detect pollutant intrusion in the shortest possible time, the present invention assumes that the pollution event with a detection time exceeding BΔt′ is not detected. min This is the lowest concentration that the sensor can detect.

[0094] Further, the step 2 specifically includes:

[0095] Step 2.1: Based on the number of impacts on the pipeline network nodes caused by the intrusion of pollutants at each node of the pipeline network, the probability P1 of pollution occurring at each node of the pipeline network is obtained;

[0096] When a pollutant invades a node in the pipe network, the pollutant will spread to some nodes and affect the water supply of these nodes. Obviously, the more nodes are affected, the greater the harm caused by the invasion of pollutants to the pipe network. Therefore, based on the number of affected nodes after pollution occurs at each node in the pipe network, the pollution probability P1 is obtained, as shown in formula (4):

[0097]

[0098] Where P1(i) represents the probability of pollutant intrusion at node i, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij is the impact of pollution at node i on node j. The probability of pollution at a node in the pipeline network is proportional to the number of nodes affected by the pollution at the node.

[0099] Step 2.2: Based on the impact of pollutant intrusion on water demand at each node of the pipeline network, the probability P2 of pollution occurring at each node of the pipeline network is obtained;

[0100] In addition to the number of affected nodes, the water demand of the nodes affected by the invasion of pollutants can also reflect the harm caused by the invasion of pollutants. When the number of affected nodes is the same, the higher the water demand of the affected nodes, the greater the harm caused by the invasion of pollutants to the pipeline network. Therefore, based on the affected water demand of each node in the pipeline network after pollution occurs, the pollution probability P2 is obtained, as shown in formula (5):

[0101]

[0102] Where P2(i) is the pollution probability of node i based on the affected water demand, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij is the impact of pollution in node i on node j; d jtis the water demand of node j at time t; It represents the total water demand of the water supply network during the simulation period. The probability of pollution occurring at a node in the network is proportional to the number of affected nodes and the water demand of these nodes.

[0103] Step 2.3: Based on the number of pipelines affected by the intrusion of pollutants at each node of the pipeline network, the probability of pollution at each node of the pipeline network is obtained;

[0104] After the pollution intrusion occurs in the pipeline network, the pollution spreads to each node along the pipeline and affects some pipelines. The affected pipelines need to take measures such as closing valves and flushing pipelines to eliminate the harm of pollution intrusion. Obviously, the number of pipelines affected by pollution intrusion can reflect the harm of pollution intrusion. Therefore, based on the number of pipelines affected by pollution at each node of the pipeline network, the pollution probability P3 is obtained, as shown in formula (6):

[0105]

[0106] Where P3(i) is the contamination probability of node i based on the number of affected pipelines, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ip is the impact of node i on pipeline P after pollution occurs. The pollution probability of a node is proportional to the number of pipelines affected by the pollution of the node;

[0107] Step 2.4: Obtain the probability of pollution occurring at each node in the pipeline network based on the diffusion distance of pollutants after pollution occurs at each node;

[0108] After the pollution intrusion occurs in the pipe network, the pollution will spread in the water supply pipe network. The longer the pollution spreads, the greater the harm caused by the pollution at the node. Obviously, the pollution diffusion distance can reflect the harm of the pollution invasion. Therefore, based on the pollution diffusion distance after the pollution occurs at each node in the pipe network, the probability of pollution P4 is calculated as shown in formula (7):

[0109]

[0110] Where P4(i) represents the pollution probability of node i based on the pollution diffusion distance, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ip is the impact of pollution on pipeline P at node i; L p is the length of pipeline P, and the pollution probability of a node is proportional to the pollution diffusion distance after the node is polluted;

[0111] Step 2.5: Combine the four pollution probabilities to obtain the joint probability P5 of pollutant intrusion at each node in the pipeline network:

[0112] The four pollution probabilities proposed above consider the impact of various pollution events on a certain aspect of the water supply network separately, but do not consider these impacts comprehensively. Therefore, the present invention combines these pollution probabilities and proposes a joint pollution probability P5(i), as shown in formula (8);

[0113] P5(i)=w1·P1(i)+w2·P2(i)+w3·P3(i)+w4·P4(i); (8)

[0114] Wherein, P5(i) represents the joint pollution probability obtained based on the four pollution probabilities, w1, w2, w3 and w4 are the weights of each pollution probability, satisfying w1+w2+w3+w4=1. Considering the change of weights, each pollution influence can be selected. The present invention adopts the same weight, regardless of which weight is adopted;

[0115] Further, the step 3 specifically includes:

[0116] Step 3.1: Consider optimizing the objective function F1 to minimize the pollutant intrusion detection time;

[0117] Only when the intrusion of pollutants is detected can corresponding measures be taken to deal with the intrusion of pollutants so as to minimize the impact of pollution. Therefore, the detection time should be shortened as much as possible to optimize the objective function F1(s) as shown in formula (9):

[0118]

[0119] Where F1(s) represents the detection time of pollutant intrusion at each node of the pipe network by the sensor arrangement scheme s, γ(i,t) represents the shortest detection time of pollutant intrusion at node i by the sensor network, if no sensor alarms the pollutant intrusion at node i, then γ(i,t)=BΔt′, α is an indicator, if the sensor network can detect the pollutant intrusion at node i, then α=1, otherwise α=0, p(i) is the probability of pollution at node i;

[0120] Step 3.2: Consider optimizing the objective function F2 to maximize the number of pollution event detections;

[0121] For a sensor network, the more pollution events detected, the better the sensor network is. The optimization objective function F2(s) is shown in formula (10):

[0122]

[0123] Where p(i) is the contamination probability of node i.

[0124] Step 3.3: Consider optimizing the objective function F3 to minimize the number of sensors;

[0125] The number of sensors is mainly related to the detection cost. The fewer the number of sensors, the lower the detection cost.

[0126] F3=min(N s )

[0127] Where N s is the number of sensors placed in the network;

[0128] Step 3.4: Consider optimizing the objective function F4 to maximize the joint information entropy of the sensor network;

[0129] In a water supply network, a single sensor can usually only respond to pollution events of some intrusion nodes. For a set of pollution events, the combination of different sensor detection conditions (detected and undetected) divides the pollution events into different intervals, thereby identifying these pollution events. Joint information entropy has been applied in the optimal placement of sensors, which can evaluate the recognition ability of the sensor network for various events. Therefore, the joint information entropy of the sensor network should be maximized;

[0130]

[0131] Where j = 1,…,N I,i , N I,i is the sensor S at node i i The number of intervals divided for various pollution events, is the probability of the pollution event being distributed in the jth interval, m i is the number of pollution events in the jth interval.

[0132] Step 3.5: Use the NSGA-II multi-objective optimization model to solve and obtain the Pareto frontier solutions under five pollution probabilities;

[0133] NSGA-II is an improved NSGA algorithm. NSGA-II obtains a set of Pareto front solutions based on the trade-offs of various objective functions. The binary decision variables of the NSGA-II multi-objective optimization model are the selection of sensor locations. The length of the chromosome is the total number of network nodes. If a sensor is deployed at node i, the value at node i is 1, otherwise its value is 0. First, a group of network nodes are randomly selected as sensor deployment nodes. According to these network nodes, the pollution detection matrix D t (i,j) Select the corresponding row and add it to the leakage detection matrix of the sensor layout S Then, the objective function value and crowding distance of the sensor layout scheme are calculated. Then, a new set of pipe network nodes are selected as the next generation to improve the objective function value. Finally, the pipe network nodes are continuously selected until no better solution than the existing generation can be obtained, and the sensor optimization layout scheme and its objective function value are output in the form of Pareto front solution.

[0134] Further, the step 4 specifically includes:

[0135] Step 4.1: Cluster and sort the Pareto solutions under each pollution probability based on multi-criteria decision analysis to obtain the optimal sensor solution under various circumstances;

[0136] As a multi-criteria decision analysis method, PROMETHEE has been widely used in various methods. Since PROMETHEE is easy to use, easy to understand, and strong in handling uncertainty, it can select the optimal solution from multiple non-optimal solutions. Therefore, the present invention uses PROMETHEE to cluster and sort Pareto solutions, providing decision makers with a set of simplified representative solutions and reducing the decision set.

[0137] Step 4.2: Calculate the impact of pollutant intrusion on the water supply network at each node of the network at each time for each scheme, and compare and analyze various schemes:

[0138] After a pollutant intrusion occurs at a certain node in the pipeline network, the pollutant spreads to some nodes along the pipeline, thereby affecting the pipeline network. The present invention mainly considers four types of impacts, including: impact on the number of nodes (I1), impact on water demand (I2), impact on the number of pipelines (I3) and pollutant diffusion distance (I4). If the pollutant intrusion is not detected by the sensor network, the impact caused by the pollutant intrusion will continue to increase over time. After the pollutant intrusion is detected by the sensor network, corresponding measures can be taken to eliminate the harm caused by the pollutant intrusion. Therefore, assuming that once the pollutant intrusion is detected, the impact caused by the pollutant intrusion at the node is zero at the corresponding detection time. The impact caused by the pollutant intrusion is shown in formulas (12)-(15):

[0139]

[0140]

[0141]

[0142]

[0143] Where I1(S), I2(S), I3(S) and I4(S) respectively represent the impact of pollutant intrusion on the number of nodes, water demand, number of pipes and pollutant diffusion distance of each node in the water supply network under the sensor layout scheme S. S (t) represents the number of pollution events that are not detected by Scheme S at time t, N(t) represents the total number of pollution events, p(i) represents the probability of pollutant intrusion at node i, n(i), d(i), Pipe n (i) and l(i) respectively represent the number of nodes affected, the affected water demand, the number of affected pipes and the pollutant diffusion distance at time t after the pollutant intrusion occurs in node i. For a scheme, the smaller the various impacts after the pollutant intrusion, the better the scheme.

[0144] The technical solution of the present invention is described in detail below with a specific example:

[0145] Step 1: Use the example pipe network model C-town to simulate and obtain the situation of pollutant invasion at each node of the pipe network.

[0146] like Figure 3 As shown in the figure, the C-town example network has a total of 407 nodes, 443 pipes, 1 constant head water source, 7 water tanks, 11 pumps and 5 valves. The simulation parameters are set as follows: total simulation time is 96h, hydraulic time step is 10min, water quality time step is 5min, and mode time step is 10min. Sensors are arranged on the nodes of the network, and there are a total of 407 initial candidate locations. Pollutant intrusion occurs at each node of the network. The pollution simulation starts at 24 times a day, corresponding to 24 demand patterns. Therefore, the total number of pollution events is 9768 (i.e. 24×407, where 407 is the total number of nodes). The pollutant injection concentration is 100mg / L, and the pollutant injection is continuous before the pollutant intrusion is detected. The lowest concentration of pollutants detected by the sensor is 0.01mg / L (i.e. C min =0.01mg / L), if the pollutant diffused to the node where the sensor is located is lower than this concentration, it cannot be detected. Considering that the pollutant invasion should be detected in the shortest possible time, the present invention limits the maximum detection time of pollutant invasion at a single node to 5 hours. If the detection time exceeds 5 hours, it is considered that the pollutant invasion of the node has not been detected.

[0147] Step 1 is as follows:

[0148] Step 1.1: Simulate 407 pollution events to obtain the impact of pollution on each node of the water supply network;

[0149]

[0150] l ij is the impact of pollutant intrusion at node j on node i. If there is an impact (the concentration of pollutants at node i is greater than 0), then l ij =1, otherwise, l ij =0;

[0151] Step 1.2: Simulate 407 pollution events to obtain the impact of pollution on each pipeline in the water supply network when pollution occurs at each node of the network;

[0152]

[0153] l p,i is the impact of pollutant intrusion at node i on pipeline p. If there is an impact (the concentration of pollutants at pipeline p is greater than 0), then l p,i =1, otherwise, l p,i =0;

[0154] Step 1.3: Simulate 9768 pollution events to obtain the pollution detection matrix;

[0155]

[0156] d j,i is the detection status of the sensor at node j after pollution event i occurs. If the pollutant concentration at node j is greater than 0.01 mg / L after pollution event i occurs, it can be detected, otherwise it cannot be detected.

[0157] Step 2: According to the impact of pollutant intrusion on water supply network nodes and pipelines, the probability of pollutant intrusion at each node of the network is obtained. Step 2 is as follows:

[0158] Step 2.1: The pollution probability P1 of each node in the pipeline network is obtained based on the number of affected nodes when pollution occurs at each node in the pipeline network. The pollution probability distribution of each node based on the number of affected nodes is as follows: Figure 5 (a)

[0159] Step 2.2: The pollution probability P2 of each node in the pipeline network is obtained based on the impact of pollutant intrusion on water demand at each node in the pipeline network. The pollution probability distribution of each node in the pipeline network based on the affected water demand is as follows: Figure 5 (b)

[0160] Step 2.3: The pollution probability P3 of each node in the pipeline network is obtained based on the number of affected pipelines when pollutants invade each node in the pipeline network. The pollution probability distribution of each node in the pipeline network based on the number of affected pipelines is as follows: Figure 5 (c)

[0161] Step 2.4: The pollution probability P4 of each node in the pipeline network is obtained based on the pollutant diffusion distance when the pollutant intrusion occurs at each node in the pipeline network. The Microsoft probability distribution of each node in the pipeline network based on the pollutant diffusion distance is as follows: Figure 5 (d)

[0162] Figure 4 The figure shows the pollution probability distribution of each node according to the impact of pollutant intrusion on each node. Each node has different distribution under different pollution probabilities, and is divided into three levels according to the probability of pollution at each node. Pollution intrusion monitoring is a mitigation measure aimed at minimizing the harm caused by pollutant intrusion. In the case of a limited number of sensors, it is necessary to give priority to the detection and identification of pollution events that cause serious pollution consequences. Figure 4 As shown in the figure, the impact of pollution on each node of the pipeline network is different. Therefore, it is reasonable to consider the change of pollution probability at each node when optimizing the layout of sensors.

[0163] After pollutants enter the network through nodes, they diffuse through pipes in the network, causing harm to the water supply network. Therefore, the number of affected nodes, affected water demand, affected pipe number and pollutant diffusion distance after pollutants invade each node of the network can represent the harm caused by pollutant invasion. The higher the harm, the higher the probability of pollution. Although some previous methods have considered the change of pollution probability, the pollution probability is more based on the attributes of each node itself, and does not consider the impact caused by it, which may be unreasonable. For example, node N11 ( Figure 4 a) is a zero demand node. If it is based on the node’s own demand, the probability of contamination at this node is zero. However, if pollutants invade this node, it will affect a large number of pipelines ( Figure 4 c); and node 156 ( Figure 4 a) The length of the directly connected pipeline is short, and the pollution probability is very low according to the length of the directly connected pipeline. However, if the probability node is invaded by pollutants, it will affect a large number of nodes and pipelines, and also cause the pollutants to spread widely in the pipeline network; the flow rate of pipeline P319 is very low, but if the downstream node of the pipeline is invaded by pollution, it will also cause serious consequences (similar to node 156). Therefore, the attributes of the node itself cannot accurately reflect the pollution probability distribution of each node in the pipeline network.

[0164] Although the pollution probability distribution of each node is obtained based on the impact of pollutant invasion on the pipeline network, the various impacts caused by the same node are not the same. For example, after the pollutant invasion occurs at node 156, the pollutant diffusion distance is large, which affects a large number of nodes and pipelines, but the water demand of these nodes is small, so the impact on water demand is small. After the pollution occurs at node 123, the pollutant diffusion distance is small, so the number of nodes and pipelines affected is small, but the water demand of the affected nodes is large, so the impact on water demand is large. The pipeline is directly connected to the node, but the number of pipelines affected by the pollution invasion is not the same as the number of nodes, for example Figure 4 In (c), the number of nodes connected to pipelines P1 and P347 is different, and the number of pipelines connected to nodes 110 and 167 is also different. When the number of affected pipelines is the same, due to the pipeline length and pipeline flow rate, the pollutant diffusion distance is different from the number of affected pipelines, e.g. Figure 4 In (d), if the flow rates of pipes P8 and P3 are the same, the pollutants take longer to pass through pipe 8. The four pollution probabilities consider the impact of pollutant invasion at each node on the water supply network from different dimensions, but show different distributions under different pollution probabilities. Therefore, the present invention considers a joint pollution probability (P5) to comprehensively consider various impacts.

[0165] Combine the contamination probability P5 with the four contamination probabilities P1-P4 mentioned above, and substitute P1-P5 into the solution of the multi-objective optimization function in step 3. The importance of each contamination event detected by each sensor layout scheme is represented by the change of the contamination probability of each node. When a contamination event with a higher contamination probability is detected, the corresponding optimization objective function value is higher, thereby ensuring that the contamination event with a high contamination probability is detected first.

[0166] Step 3: Based on the multi-objective optimization method, the Pareto frontier solution under various pollution probabilities is obtained. Step 3 is as follows:

[0167] Step 3.1: Randomly select a group of water supply network nodes as the initial plan for sensor optimization layout;

[0168] Step 3.2: Based on the selected initial sensor solution, the four optimization objective function values ​​of each solution are calculated according to the pollution detection matrix and the five pollution probabilities;

[0169] Step 3.3: Select another group of water supply network nodes as the evolutionary solution, optimize the existing solution based on the new objective function value calculated in step 3.2, and repeat the selection of network nodes until the existing objective function value cannot be optimized. Output the sensor layout solution and the corresponding objective function value as the Pareto front solution;

[0170] The NSGA-II method was used to solve the optimization objective function of five contamination probabilities, and 800 Pareto front solutions were obtained under the five contamination probabilities.

[0171] Step 4: Use the multi-criteria decision-making cluster analysis method to cluster and sort the Pareto front solutions under various pollution probabilities, obtain the optimal solutions for each cluster under different situations, and compare and analyze various optimal solutions based on the impact of pollutant intrusion on the pipeline network;

[0172] Step 3.2: Use the multi-criteria decision analysis method to cluster the Pareto front solution and obtain the cluster distribution of the sensor optimization layout scheme under various situations. The present invention uses four optimization objectives (F1, F2, F3, F4) to compare the various schemes and considers three situations: situation 1, the four optimization objectives use the same weight; situation 2, give priority to the number of sensors (the weights of F1, F2 and F4 are 0.2, and the weight of F3 is 0.6); situation 3, give priority to the number of pollution event detections (the weights of F1, F3 and F4 are 0.2, and the weight of F2 is 0.4). No matter what weights are used for the four optimization objectives, they are within the protection scope of the present invention. Figure 2 The Pareto solution clustering diagram is shown. First, it is calculated according to PROMETHEE. And sort them from small to large, select the largest and the smallest As the center point of the two clusters, and then calculate the and and The absolute difference of , select the cluster with the smaller absolute difference. After obtaining two new clusters, recalculate the absolute difference of each point in the cluster Then the clustering is repeated until the difference is less than the threshold, the clustering is stopped, and the optimal solution within each cluster is obtained. Figure 5 , Figure 6 and Figure 7 The Pareto solutions of the optimal sensor layout under different pollution probabilities in case 1, case 2, and case 3 are clustered distributions. As shown in the figure, under the same situation, different pollution probabilities result in different Pareto distributions. For the Pareto solutions with the same pollution probability, different clustered distributions will be obtained when different weights are used for the optimization objectives. Figure 5 As shown in , for case 1, each contamination probability obtained 5 clusters (excellent, suboptimal, good, qualified and poor). Figure 6 and Figure 7 As shown, cases 2 and 3 respectively obtain three clusters (excellent, qualified, and poor).

[0173] like Figure 5As shown in the figure, the optimal solutions (cluster 1) for the distribution of each probability Pareto solution obtained under situation 1 are concentrated in the middle. The number of pollution event detections, the number of sensors, the detection time and the joint information entropy of this type of solution are in the middle. Clusters 2, 3, 4 and 5 are distributed at the two ends of cluster 1. The number of clusters under each pollution probability is not equal. The preferences between the various solutions in the same cluster are within a certain range, and the performance differences of these solutions are small. The points in cluster 1 are more concentrated, while the points in cluster 5 are the most dispersed. This is because some solutions in cluster 5 mostly represent some extreme cases, such as a very small or very large number of sensors. These solutions only dominate a single optimization target and do not consider various optimization targets. Therefore, they are ranked at the back. Based on the optimization target of multi-criteria decision-making, a solution that satisfies the decision maker's preferences can be obtained, rather than just considering a certain optimization target.

[0174] If the number of sensors is given priority, the Pareto solution clustering of each pollution probability is obtained as follows Figure 6 As shown in Figure 2, the scheme with fewer sensors is preferred. Although the number of sensors is given priority, it is not the scheme with the least number of sensors that ranks highest.

[0175] Figure 8 The figure shows the pollution event detection rate, detection time, number of sensors and joint information entropy of each scheme under different pollution probabilities. In order to compare various schemes under the condition of changing pollution probability, the present invention calculates the numerical detection rate and detection time (P0) as well as the true detection rate and detection time (P1-P5). The true detection rate of various schemes is higher than the actual detection rate. This is because when considering the change of node pollution probability, each scheme pays more attention to the detection of high-probability pollution events. Under the condition of equal probability, the detection rate of these schemes is low, and some pollution events with lower pollution hazards are not detected. Therefore, the true detection rate of various schemes is high, and the corresponding pollution detection time is also longer.

[0176] like Figure 8 As shown in , when the number of sensors is large enough, the numerical detection rate (detection time) and the true detection rate (detection time) of pollution events are not much different, as shown in Figure 8 (a) Schemes 5, 14 and 15. When the number of sensors is large enough, the sensor network can detect most pollution events in a timely manner regardless of which node in the pipeline network pollutants invade. Therefore, increasing the number of sensors can significantly improve the detection rate of pollution events. However, due to the installation and maintenance costs of sensors, the number of sensors cannot be infinite.

[0177] When clustering Pareto solutions, if the number of sensors is given priority, 15 solutions are obtained, such as Figure 8(b) shows that schemes 1, 4, 7, 10, and 13 are the optimal schemes based on probabilities P1, P2, P3, P4, and P5. Compared with other schemes, these schemes have fewer sensors. Figure 8 As shown in (b), when the pollution probability change is not considered, the pollution event detection rate of these schemes is less than 60%, and nearly half of the pollution events are not detected. However, if the pollution event probability change is considered, the pollution event detection rate of these schemes exceeds 80%. When the number of sensors is limited, a better sensor layout scheme can be obtained based on the pollution probability change, which can reduce the harm caused by pollutant invasion.

[0178] Step 4.2: Compare and analyze various solutions based on the impact of pollutant intrusion on the water supply network at each node of the network;

[0179] Fig. 9 , Fig.10 and Fig.11 The impact of the optimal solutions based on each pollution probability in case 1, case 2 and case 3 after the invasion of pollutants occurs. The present invention considers the impact of three moments, namely 1 hour, 3 hours and 5 hours. At the same detection time, the impact of each solution in case 2 is the largest, and that in case 3 is the smallest.

[0180] like Fig. 9 As shown, compared with I3 and I4, I1 and I2 are relatively small (10 -1 ). This is because after a pollutant invades a certain node in the pipe network, the pollution must spread to each node through the pipe. The shorter the pollution invasion time (for example, 1 hour), the impact of pollutant invasion on the pipe is greater than that on the node; and as time goes by (for example, 5 hours), the difference in the impact of pollutant invasion on the pipe and on the node decreases. At 1 hour, the impact of pollutant invasion on the pipe after the sensor layout schemes obtained based on the five pollution probabilities (I3 and I4) is basically the same, while the impact of each scheme on the node (I1 and I2) shows a large difference. The schemes based on probabilities P4 and P5 have an I2 that is smaller than the case where the pollution probability changes under equal probability. This may be because these two probabilities pay more attention to the spread of pollutants in the water supply network, mainly focusing on the detection of water pipelines, and the number of network nodes in the water pipeline is relatively small, and even if some nodes exist, they may be zero-demand nodes.

[0181] At 3h, as time goes by, more pollution events are detected, and the various impacts caused by pollutant invasion continue to decrease. I1 is still the highest under equal probability for various schemes. Fig. 9 As shown in (b), the scheme based on P2 rapidly reduces I2 under various pollution probabilities and has almost no effect, which indicates that compared with several other pollution probabilities, the scheme based on P2 pays more attention to reducing the impact on water demand.

[0182] At 5h, the schemes based on probabilities P2 and P3 have less impact on the water supply network under various pollution probabilities, which indicates that the scheme with the least impact on the water supply network may be obtained based on pollution probabilities P2 and P3 when optimizing the sensor layout. Although P5 takes into account the four impacts, the schemes based on P1 and P4 still have a greater impact on the water supply network at 5h, so it may be more reasonable to use smaller weights for P1 and P4 when considering the joint pollution probability.

[0183] like Fig. 9 As shown in Figure 2, the impact of various schemes obtained in Case 2 on the water supply network at each time is within the same order of magnitude. This is because the number of sensors is small, and the number of pollution events detected is relatively small regardless of the pollution probability. At 1 hour, Fig.10 (a) shows that I1 is higher than other pollution probabilities when the pollution probability is P0. Fig.10 I1 shown in (be) is lower than other pollution probabilities when the pollution probability is P0. This is because compared with other pollution probabilities, P1 pays more attention to detecting pollution events that affect more nodes. These events are detected first in a short time (e.g., 1h). Therefore, the scheme based on P1 has lower I1 than the case of pollution probability P0 when the pollution probability changes (P2-P5). However, as time goes by (3h and 5h), as Fig.10 As shown in the figure, I1 of various schemes is lower than that of P0 when the pollution probability changes (P2-P5). This shows that considering the change of node pollution probability when optimizing the sensor layout can significantly reduce the harm caused by pollution. At 5h, the schemes based on P2 and P4 have less impact on the water supply network under various pollution probabilities.

[0184] Fig.11 The figure shows the impact of each scheme on the water supply network at different pollution probabilities under scenario 3. Compared with scenarios 1 and 2, the number of sensors in scenario 3 is more. Therefore, I1(10 -2 ) and I2(10 -3) are very small. At 1h and 3h, the schemes are not much different under different pollution probabilities, but at 5h, the impact of each scheme under the change of pollution probability is significantly smaller than that under the condition of equal probability. This is because when the number of sensors is large enough, most pollution events can be detected even without considering the change of node pollution probability. Although most pollution events can be detected, some pollution events take a long time to detect. Therefore, the scheme obtained by considering the change of node pollution probability at 3h is more affected by pollution under the condition of equal probability. When the number of sensors is sufficient, the scheme obtained by considering the change of node pollution probability has little effect on the change of pollution event detection rate, but can shorten the detection time of pollution events, thereby reducing the impact of pollutant invasion on the water supply network.

[0185] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A multi-target arrangement method of pollution monitoring sensors for water supply network based on pollution probability, characterized by: The steps include: Step 1: simulate the situation of pollutant invasion at each node of the water supply network to obtain the impact of various pollution events on the water supply network; Step 2: According to the impact of pollution at each node on the water supply network, the pollution probability of each node is obtained; Step 3: Based on the multi-objective optimization method, the Pareto frontier solution under each pollution probability is obtained; Step 4: Use the multi-criteria decision-making cluster analysis method to cluster and sort the Pareto front solutions under various pollution probabilities, obtain the optimal solutions for each cluster under different situations, and compare and analyze various optimal solutions based on the impact of pollutant intrusion on the pipeline network; In step 2, five pollution probabilities are obtained: pollution probability based on the number of affected nodes, pollution probability based on the water demand, pollution probability based on the number of affected pipelines, pollution probability based on the diffusion distance of pollutants, and joint pollution probability that comprehensively considers the above four influences; The step 2 specifically includes: Step 2.1: Based on the number of impacts on the pipeline network nodes caused by pollutant intrusion at each node, the probability P1 of pollution occurring at each node of the pipeline is obtained: Where P1(i) represents the probability of pollutant intrusion at node i, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij The impact of pollution in node i on node j; Step 2.2: Based on the impact of pollutant intrusion on water demand at each node of the pipeline network, the probability P2 of pollution occurring at each node of the pipeline network is obtained: Where P2(i) is the pollution probability of node i based on the affected water demand, j = 1, 2, ..., N n , N n is the total number of water supply network nodes; l ij is the impact of pollution in node i on node j; d jt is the water demand of node j at time t; It represents the total water demand of the water supply network during the simulation period; Step 2.3: Based on the number of pipelines affected by the intrusion of pollutants at each node in the pipeline network, the probability P3 of pollution occurring at each node in the pipeline network is obtained: Where P3(i) is the contamination probability of node i based on the number of affected pipelines, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ip is the impact of pollution on pipeline P after node i occurs; Step 2.4: According to the diffusion distance of pollutants after pollution occurs at each node, the probability P4 of pollution occurring at each node in the pipeline network is obtained: Where P4(i) represents the pollution probability of node i based on the pollution diffusion distance, P = 1, 2, ..., N P , N P is the total number of water supply network pipes; l ip is the impact of pollution on pipeline P at node i; L p is the length of the pipe P; Step 2.5: Combine the four pollution probabilities to obtain the joint probability P5 of pollutant intrusion at each node in the pipeline network: P5(i)=w1·P1(i)+w2·P2(i)+w3·P3(i)+w4·P4(i); (8) Wherein, P5(i) represents the joint pollution probability obtained based on the four pollution probabilities, w1, w2, w3 and w4 are the weights of each pollution probability, satisfying w1+w2+w3+w4=1.

2. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability according to claim 1 is characterized in that: The impact on the water supply network in step 1 includes the impact on the network nodes and pipelines.

3. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability as claimed in claim 2 is characterized by: The step 1 specifically includes: Step 1.1: Simulate the situation of pollutant invasion at each node of the pipeline network to obtain the pollution impact matrix when pollutants invade each node; The pollution impact matrix includes the node impact matrix and the pipeline impact matrix, which refers to the impact of pollutant invasion on the pipeline nodes and pipelines at each node of the pipeline network. Assuming that the pollutant invasion occurs at the pipeline node, the number of pipeline nodes is N. n Based on the hydraulic model of the water supply network, the influence matrix I of the network nodes and pipelines after the intrusion of pollutants at each node of the network is obtained. t (i, j) and I t (i, p), as shown in equations (1) and (2) respectively: In the formula, and They represent the impact of pollutant intrusion on node j and pipeline p at time t, respectively. t=Δt, 2Δt, …, AΔt, Δt represents the water quality simulation time interval, AΔt represents the total simulation time. Assuming that the pollutant intrusion occurs at time t=0, the pollution is continuously injected during the entire simulation time period. j=1, 2, …, N n , N n is the total number of nodes in the water supply network, p=1,2,…,N p , N p is the total number of water supply network pipes. If the pollutant concentration at node j is greater than zero at time t, then otherwise, If the pollutant concentration at pipe p is greater than zero at time t, then otherwise, Step 1.2: Simulate the situation where pollutants invade each node of the pipe network to obtain the pollution detection matrix when pollutants invade each node; The network hydraulic model is used to simulate the intrusion of pollutants at each node of the water supply network, and the pollution detection matrix D under different pollution situations is obtained. t (i, j), as shown in formula (3): In the formula, It represents the detection of the sensor at node j after the pollution intrusion at node i at time t. If the pollutant concentration at node j at time t is greater than or equal to C min ,but otherwise, t=Δt',2Δt',…,BΔt',Δt' is the sampling time interval of the sensor, BΔt' is the longest detection time of the sensor network for pollutant intrusion on a single node, C min This is the lowest concentration that the sensor can detect.

4. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability as claimed in claim 1, characterized in that: In step 3, based on the multi-objective optimization method, four optimization objective functions are used to obtain the Pareto frontier solution under each pollution probability: shortest detection time, maximum number of pollution event detections, minimum number of sensors, and maximum joint information entropy.

5. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability as claimed in claim 4, characterized in that: The step 3 specifically includes: Step 3.1: Consider optimizing the objective function F1 to minimize the pollutant intrusion detection time; Where F1(s) represents the detection time of pollutant intrusion at each node of the pipe network by the sensor arrangement scheme s, γ(i,t) represents the shortest detection time of pollutant intrusion at node i by the sensor network, if no sensor alarms the pollutant intrusion at node i, then γ(i,t)=BΔt', α is an indicator, if the sensor network can detect the pollutant intrusion at node i, then α=1, otherwise α=0, p(i) is the probability of pollution at node i; Step 3.2: Consider optimizing the objective function F2 to maximize the number of pollution event detections; Where p(i) is the pollution probability of node i; Step 3.3: Consider optimizing the objective function F3 to minimize the number of sensors: F3=min(N s ) Where N s is the number of sensors placed in the network; Step 3.4: Consider optimizing the objective function F4 to maximize the joint information entropy of the sensor network: Where j = 1,…,N I,i , N I,i is the sensor S at node i i The number of intervals divided for various pollution events, P(y i,j )=m i / N I,i is the probability of the pollution event being distributed in the jth interval, m i is the number of pollution events in the jth interval; Step 3.5: Use the NSGA-II multi-objective optimization model to solve and obtain the Pareto frontier solutions under five pollution probabilities.

6. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability as claimed in claim 5, characterized in that: Step 3.5 is as follows: First, randomly select a group of pipe network nodes as sensor deployment nodes, and select the pollution detection matrix D according to these pipe network nodes. t (i,j) Select the corresponding row and add it to the leakage detection matrix of the sensor layout S Then, the objective function value and crowding distance of the sensor layout scheme are calculated; then, a new set of pipe network nodes are selected as to improve the objective function value and evolve to the next generation; finally, the pipe network nodes are continuously selected until no better solution than the existing generation can be obtained, and the sensor optimization layout scheme and its objective function value are output in the form of Pareto front solution.

7. The method for multi-target arrangement of pollution monitoring sensors for water supply network based on pollution probability according to claim 1, characterized in that: The step 4 specifically includes: Step 4.1: Cluster and sort the Pareto solutions under each pollution probability based on multi-criteria decision analysis to obtain the optimal sensor solution under various circumstances; Step 4.2: Calculate the impact of pollutant intrusion on the water supply network at each node of the network at each time for each scheme, and compare and analyze various schemes; the impact caused by pollutant intrusion is shown in equations (12)-(15): Where I1(S), I2(S), I3(S) and I4(S) respectively represent the impact of pollutant intrusion on the number of nodes, water demand, number of pipes and pollutant diffusion distance at each node of the water supply network under the sensor layout scheme S, and N S (t) represents the number of pollution events that are not detected by Scheme S at time t, N(t) represents the total number of pollution events, p(i) represents the probability of pollutant intrusion at node i, n(i), d(i), Pipe n (i) and l(i) represent the number of nodes affected, the affected water demand, the number of affected pipes and the pollutant diffusion distance at time t after the pollutant invades node i, respectively. For a scheme, the smaller the various impacts after the pollutant invades, the better the scheme is.