A combined system overflow pollution real-time control method based on rainfall data

Abstract

The application discloses a kind of based on rainfall data's combined flow system overflow pollution real-time control method, including current water rain condition data acquisition;Overflow category construction and cluster analysis;Current water rain condition category identification and control strategy call, real-time acquisition current water rain condition set, calculate and the distance between each cluster center, identify its belonging overflow category, and extract corresponding control strategy set, for guiding the real-time control of combined flow system drainage facility.The application method only uses water rain condition to combine clustering algorithm, simplifies calculation process;Get future a period of time to multiple combined flow system pipe network drainage control facilities for real-time control, without too frequent data update.

Description

Technical Field

[0001] This invention belongs to the field of overflow control technology, and in particular relates to a real-time control method for combined overflow pollution based on rainfall data. Background Technology

[0002] Combined sewer systems are characterized by rainwater and sewage sharing the same pipe network. During light to moderate rainfall, initial rainwater carries surface pollutants into the pipes and is diverted to sewage treatment plants via interception facilities. However, during periods of heavy rainfall, when the pipe network flow exceeds the interception capacity or the treatment capacity of the sewage treatment plant, the mixed rainwater and sewage are directly discharged into receiving water bodies through overflow outlets, forming combined sewer overflows. While this overflow behavior has a flood control function, it also directly discharges untreated pollutants (such as chemical oxygen demand (COD), ammonia nitrogen (NH3-N), and total suspended solids (SS)) into environmental water bodies, causing pollution to the urban water environment.

[0003] Combined sewer overflow pollution control is one of the core challenges in urban water environment management. Current methods do not effectively utilize historical data resources. While a large amount of historical rainfall, water level, and water quality monitoring data has been accumulated, most systems have not conducted in-depth analysis and modeling, failing to extract typical scenarios to support real-time control strategies. The operation of combined sewer systems is influenced by multiple factors, including rainfall duration, water distribution, and the capacity of each node. In actual operation, scenarios are complex and variable, making fixed rules insufficient to cope with diverse events, and lacking multi-scenario modeling and classification mechanisms.

[0004] There is an urgent need for a context recognition and classification method based on historical data, which can fully combine rainfall data, water level fullness and water quality monitoring information to construct typical combined sewer overflow categories, and establish a set of corresponding pollution control strategies based on the clustering results, so as to achieve more refined and efficient real-time overflow control and reduce the risk of urban non-point source pollution. Summary of the Invention

[0005] This invention aims to utilize multi-source water and rainfall data and cluster analysis technology to identify the overflow category of the current rainfall event, and implement differentiated real-time control strategies accordingly to achieve precise pollution reduction and drainage scheduling.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A real-time control method for combined sewer overflow pollution based on rainfall data includes the following steps:

[0008] S1. Current Water and Rainfall Data Collection:

[0009] The current rainfall observation data, future rainfall forecast data, and real-time water level filling values ​​of multiple key nodes in the pipeline network are obtained to form the current water and rainfall situation set;

[0010] S2. Overflow category construction and cluster analysis:

[0011] A set of historical water and rainfall events and a set of historical water quality events are constructed. A clustering algorithm is used to perform cluster analysis on the set of historical water and rainfall events to generate a set of overflow categories.

[0012] By mapping between cluster centers and historical event points, a correspondence is established between each overflow category and its corresponding set of historical water quality events, thereby extracting water quality characteristics and constructing a set of real-time combined overflow control strategies.

[0013] S3. Current water and rainfall situation category identification and control strategy retrieval:

[0014] The system acquires real-time data on water and rainfall conditions, calculates the distance to each cluster center, identifies the overflow category to which it belongs, and extracts the corresponding set of control strategies to guide the real-time control of combined sewer systems.

[0015] Step S1 generates a set of water and rainfall data reflecting the characteristics of the current rainfall process, ensuring that the control logic is based on the latest state. Step S2 uses a clustering algorithm to classify historical water and rainfall events, forming a set of overflow categories, and associates them with historical water quality events to extract the pollution characteristics and control requirements corresponding to different categories. Step S3 calculates the distance between the current water and rainfall data and each cluster center, determines its category, and calls the corresponding control strategy, realizing a full-process control system from "observation-identification-decision," significantly improving response speed and decision accuracy.

[0016] Furthermore, the current water and rainfall data set includes at least one of the following data types:

[0017] a) Hourly rainfall at the current time and in the preceding hours;

[0018] b) Rainfall forecast data for the next few hours;

[0019] c) Water level filling data for multiple key nodes at various times.

[0020] Multi-source data fusion improves the accuracy of current water and rainfall conditions description and provides more reliable data support for category identification.

[0021] Furthermore, the set of historical water quality events includes at least one of the following water quality parameters: chemical oxygen demand, ammonia nitrogen, and dissolved oxygen.

[0022] To ensure that pollution intensity and water quality characteristics can be extracted under different overflow categories, thereby formulating differentiated and implementable pollution control strategies.

[0023] Furthermore, the clustering algorithm is a K-means clustering algorithm, which includes the following sub-steps:

[0024] a) Randomly select k historical water and rainfall event samples as initial cluster centers;

[0025] b) Assign historical events to their corresponding cluster centers based on the minimum Euclidean distance;

[0026] c) Update cluster centers based on the mean of each data sample;

[0027] d) Repeat steps b and c until the change in cluster centers is less than the threshold or the maximum number of iterations is reached.

[0028] Ensure that the overflow classification results are stable, representative, and physically meaningful, and enhance the correlation between the categories and pollution characteristics.

[0029] Furthermore, the real-time merging overflow control strategy includes at least one of the following control operations:

[0030] a) Open or close a specific overflow outlet;

[0031] b) Start or stop the operation of the storage tank;

[0032] c) Adjust the pump station operation strategy.

[0033] It provides a variety of executable control actions to adapt to different response needs and achieve precise and automated control.

[0034] Furthermore, the cluster category to which the current water and rainfall situation belongs is determined by calculating the Euclidean distance between the current water and rainfall situation and each cluster center, and selecting the category corresponding to the smallest distance.

[0035] It employs efficient classification criteria to quickly locate categories, ensuring real-time performance while also considering cluster stability.

[0036] Furthermore, the control strategy set has differentiated response rules for different categories, including:

[0037] a) In the event of light rain and high pollution, close all overflow outlets to reduce the pollution load;

[0038] b) In the case of moderate rain and a single node about to be full of water, open the overflow outlet of the corresponding node locally;

[0039] c) In cases of moderate rain where multiple nodes are not fully flooded but pollution concentrations are high, close the overflow outlets;

[0040] d) In the event of heavy rain, high water level, and low pollution, all overflow outlets should be opened immediately for priority drainage.

[0041] By leveraging context awareness to enable differentiated strategy configuration, the overall pollution reduction capacity and operational resilience of urban drainage systems can be enhanced.

[0042] The present invention has the following beneficial effects:

[0043] Traditional combined sewer overflow control relies on static thresholds or experience-based scheduling, which cannot respond to complex and ever-changing rainfall scenarios in real time, and struggles to balance drainage safety and water environment protection goals. This invention fully leverages the high correlation between rainfall data and combined sewer overflow pollution control. The proposed technical solution constructs a dynamic hydrological and rainfall data set containing both observational and predictive information; extracts typical overflow categories from historical data and associates them with pollution characteristics; quickly identifies the category during the current rainfall process and invokes the matching pollution control strategy; thus realizing a multi-scenario, multi-strategy intelligent response control system. Through the method of this invention, only hydrological and rainfall data combined with clustering algorithms are used, simplifying the calculation process; real-time control of multiple combined sewer drainage facilities over a future period is achieved without requiring overly frequent data updates. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the method.

[0045] Figure 2 The example shows the sequence of pre-rainfall observations, future rainfall forecasts, and real-time water level filling at multiple key nodes in the pipeline network.

[0046] Figure 3 The example is a historical water and rainfall sequence diagram.

[0047] Figure 4 This is a sequence diagram of historical water quality events for an example.

[0048] Figure 5 The data shows the rainfall and water level filling sequence of multiple key nodes in the pipeline network, representing the first clustering feature after clustering.

[0049] Figure 6 The rainfall and water quality sequence plots show the first clustering features after clustering.

[0050] Figure 7 The data shows the rainfall and water level filling sequence of multiple key nodes in the pipeline network, based on the third clustering feature after clustering.

[0051] Figure 8 The rainfall and water quality sequence diagrams show the third clustering features after clustering. Detailed Implementation

[0052] A real-time control method for combined sewer overflow pollution based on rainfall data, the overall implementation process is as follows: Figure 1 This includes the following steps S1 to S3:

[0053] S1. Current Water and Rainfall Data Collection

[0054] The current rainfall observation data, future rainfall forecast data, and real-time water level filling values ​​of multiple key nodes in the pipeline network are obtained to form the current water and rainfall situation set.

[0055] in:

[0056] Rainfall observation data: collected by on-site rainfall monitoring equipment and transmitted via the Internet of Things;

[0057] Rainfall forecast data can be obtained from meteorological departments, commercial meteorological companies, or through simulations using fundamental / non-fundamental models.

[0058] Water level fullness data: collected in real time by water level monitoring devices deployed at key nodes via the Internet of Things.

[0059] The current water and rainfall data is expressed in the following form:

[0060]

[0061] in, This represents the current water and rainfall data set; n represents different data dimensions.

[0062] Taking a real-time control project for overflow pollution from a combined sewer system in a large city as an example, the current water and rainfall information is as follows: Figure 2 As shown, the data includes historical rainfall observations at time t+0 and from t-1 to t-5 hours, future rainfall forecasts from t+1 to t+6 hours, and the water level filling degree of three key pipeline nodes from t-5 to t+0, comprising a total of 12 types of data features.

[0063] S2. Overflow Category Construction and Cluster Analysis

[0064] This step uses clustering algorithms to analyze historical water and rainfall events and water quality events, constructing several representative overflow categories to support the formulation of subsequent control strategies.

[0065] S21. Historical Data Acquisition and Construction

[0066] Collect the following two types of historical data:

[0067] Collection of historical flood and rainfall events X:

[0068] X = {(x1, x2, ..., x} n )1,(x1,x2,…,x n )2,…,(x1,x2,…,x n ) m1}

[0069] Where X is the set of historical water and rainfall events; m1 represents the number of historical water and rainfall events; and n is the data dimension contained in each event, consistent with step S1.

[0070] Historical water quality event set Y:

[0071] Y = {(y1, y2, ..., y} p )1,(y1,y2,…,y p )2,…,(y1,y2,…,y p ) m2}

[0072] Where Y is the set of historical water quality events for water quality observation data of multiple key nodes in the pipeline network; m2 represents the number of historical water quality events; and p represents the dimension of water quality parameters (such as COD, ammonia nitrogen, dissolved oxygen, etc.).

[0073] Historical water and rainfall conditions, historical water quality events, etc. Figure 3 , 4 As shown, the content includes the m′th event (t+0) at a historical time point, rainfall observations from 1 hour (t-1) to 5 hours (t-5) before the historical time point, and from 1 hour (t+1) to 6 hours (t+6) before the historical time point, water level filling status of multiple key nodes in the pipeline network, and real-time water quality (COD, dissolved oxygen, ammonia nitrogen) data. Specifically, the m′th event data in set X includes rainfall data from t-5 to t+0 and real-time water level filling status of three key nodes in the pipeline network from t-5 to t+6. The m′th event data in set Y includes three types of water quality data from t-5 to t+6.

[0074] S22. Cluster analysis to establish overflow category set C

[0075] The historical rainfall and water level data set X is analyzed using the K-means clustering algorithm. The specific process includes:

[0076] Step S211, initialize cluster centers

[0077] At the start of the K-means model algorithm, the number of clusters is set to k, and k event samples are randomly selected as initial cluster centers, forming the initial cluster center set Q0:

[0078]

[0079] Q0 is the initial set of cluster centers; This represents the initial cluster center value on the nth dimension of water and rainfall characteristics; This represents the initial center of the k-th cluster category; the initial cluster center set is used in step S212.

[0080] Step S212, the event is assigned to the nearest cluster category.

[0081] For each historical event point x∈X in the set of historical water and rainfall events X, calculate its Euclidean distance to all cluster centers and assign it to the nearest cluster C. k :

[0082]

[0083] Step S213, Update cluster centers

[0084] For each cluster category C k A new cluster center is calculated based on all data points within the cluster, and the mean is used as the updated cluster center Q. t :

[0085]

[0086]

[0087] Q t Let be the set of cluster centers obtained in the t-th iteration; Let x be the new cluster center for the nth type of water and rainfall data in the tth iteration; n It represents the data value of the nth feature dimension; |C k | represents the number of events in the k-th cluster category.

[0088] Step S214, iterate until termination

[0089] Repeat steps S212 to S213, repeatedly assigning cluster categories to the data set and updating cluster centers, until the termination condition is met. The termination condition includes that the cluster centers no longer change significantly, the maximum number of iterations is reached, and the final cluster center set Q is obtained.

[0090] Q = {(q1,q2,…,q} n )1,(q1,q2,…,q n )2,…,(q1,q2,…,q n ) k}

[0091] Furthermore, based on the distance between the event point and the cluster center, the cluster category to which each historical water and rainfall event belongs can be obtained, forming a set of event classification results:

[0092] C = {c1, c2, ..., c} m}

[0093] Where C is the set of cluster categories; c m ∈{1,2,...,k} represents the category to which the m-th historical event belongs; multiple events may belong to the same category or may be distributed in different categories.

[0094] In this embodiment, the number of clusters is set to 4, and the clustering results are as follows: Figure 5 , 6 As shown in 7 and 8, Figure 5 , 6 Tables 7 and 8 only show the clustering results for two categories, displaying the distribution characteristics of each category using both bar charts and solid line curves. Each cluster category contains the following time series data dimensions:

[0095] Rainfall data: hourly rainfall from t-5 to t+0 hours;

[0096] Pipeline status: Water level filling changes at 3 key nodes from t-5 to t+6 hours;

[0097] Water quality data: parameters such as dissolved oxygen, COD, and ammonia nitrogen;

[0098] Future trends: Rainfall forecast trends for each cluster category from t+1 to t+6 hours.

[0099] The clustering categories may consist of 2 to k categories, therefore the categories within each subset of a clustering category set may be the same or different from each other. The clustering analysis results are used to support the construction of the overflow pollution risk assessment model and the formulation of early warning strategies in subsequent step S3.

[0100] S23. Construct a set of real-time overflow control strategies Z

[0101] Since there is a one-to-one correspondence or dual mapping relationship between historical water and rainfall events X and historical water quality events Y, the cluster category C can be synchronously mapped to the water quality event set Y, thereby analyzing the typical water quality characteristics of various overflow scenarios and providing a basis for control strategies.

[0102] Based on the above mapping relationship, water quality data corresponding to events contained in each cluster can be extracted, and feature profiles of each cluster can be constructed. Then, through scenario-based simulation methods (such as establishing simulation models to simulate rainfall response) or based on empirical rules, targeted combined sewer overflow control strategies can be formulated for each category, thereby forming a set of real-time control strategies.

[0103] Z = {(z1, z2, ..., z} r )1,(z1,z2,…,z r )2,…,(z1,z2,…,z r ) k}

[0104] Where Z is the set of real-time merging overflow control strategies; (z1, z2, ..., z r ) k z represents a control policy group containing r strategies established for the k-th cluster category. rThe representative indicates specific combined sewer control strategy items, such as opening or closing specific overflow outlets, starting and stopping storage tanks, and scheduling pump station operations.

[0105] For example, Figure 5 The control strategies for each of the four clustering categories shown can be formulated based on the following judgment principles:

[0106] Category 1: This is a light rain scenario. The water level at each pipeline node will not reach the full water threshold, but the pollutant concentration will be high. The proposed strategy is to close all overflow outlets to reduce the risk of pollutant discharge.

[0107] Category 2: This is a moderate rain scenario. Only node 1 will reach full water level in the next hour, and the pollution concentration is low. The proposed strategy is to open the overflow outlet of the pipeline network at node 1 on t+1 to alleviate local hydraulic pressure.

[0108] Category 3: Moderate rain scenario, all nodes will not be full of water in the future, but the pollutant concentration is high, the proposed strategy is to close the overflow outlet;

[0109] Category 4: Heavy rain scenario. All nodes are currently at full water level with low pollutant concentration. The proposed strategy is to immediately open all overflow outlets (t+0) to prioritize drainage safety.

[0110] S3. Current water and rainfall situation category identification and control strategy retrieval

[0111] Real-time acquisition of current water and rainfall conditions By calculating the distance to the cluster center Q, the overflow category k′ to which it belongs is identified, and the corresponding control strategy Z is extracted. k ′:

[0112] Z k′ =(z1,z2,…,z) r ) k′

[0113] Z k′ It is a set of real-time combined sewer overflow control strategies that can control multiple combined sewer drainage control facilities in real time.

[0114] like Figure 3 The current water and rainfall situation is shown. and Figure 5 The second category is the closest, so it is identified as a moderate rain scenario. It is predicted that the first node will overflow soon and the pollution concentration will be low. The system will automatically activate the first node overflow device at t+1.

Claims

1. A real-time control method for combined sewer overflow pollution based on rainfall data, characterized in that, The steps include the following: S1. Current Water and Rainfall Data Collection: The current rainfall observation data, future rainfall forecast data, and real-time water level filling values ​​of multiple key nodes in the pipeline network are obtained to form the current water and rainfall situation set; S2. Overflow category construction and cluster analysis: A set of historical water and rainfall events and a set of historical water quality events are constructed. A clustering algorithm is used to perform cluster analysis on the set of historical water and rainfall events to generate a set of overflow categories. By mapping between cluster centers and historical event points, a correspondence is established between each overflow category and its corresponding set of historical water quality events, thereby extracting water quality characteristics and constructing a set of real-time combined overflow control strategies. S3. Current water and rainfall situation category identification and control strategy retrieval: The system acquires real-time data on water and rainfall conditions, calculates the distance to each cluster center, identifies the overflow category to which it belongs, and extracts the corresponding set of control strategies to guide the real-time control of combined sewer systems.

2. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The current water and rainfall data set includes at least one of the following data types: a) Hourly rainfall at the current time and in the preceding hours; b) Rainfall forecast data for the next few hours; c) Water level filling data for multiple key nodes at various times.

3. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The set of historical water quality events includes at least one of the following water quality parameters: chemical oxygen demand, ammonia nitrogen, and dissolved oxygen.

4. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The clustering algorithm is the K-means clustering algorithm, which includes the following sub-steps: a) Randomly select k historical water and rainfall event samples as initial cluster centers; b) Assign historical events to their corresponding cluster centers based on the minimum Euclidean distance; c) Update cluster centers based on the mean of each data sample; d) Repeat steps b and c until the change in cluster centers is less than the threshold or the maximum number of iterations is reached.

5. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The real-time merging overflow control strategy includes at least one of the following control operations: a) Open or close a specific overflow outlet; b) Start or stop the operation of the storage tank; c) Adjust the pump station operation strategy.

6. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The current water and rainfall situation is assigned to a cluster category by calculating the Euclidean distance between the current water and rainfall situation and each cluster center, and then selecting the category corresponding to the smallest distance.

7. The real-time control method for combined sewer overflow pollution based on rainfall data according to claim 1, characterized in that, The set of control strategies has differentiated response rules for different categories, including: a) In the event of light rain and high pollution, close all overflow outlets to reduce the pollution load; b) In the case of moderate rain and a single node about to be full of water, open the overflow outlet of the corresponding node locally; c) In cases of moderate rain where multiple nodes are not fully flooded but pollution concentrations are high, close the overflow outlets; d) In the event of heavy rain, high water level, and low pollution, all overflow outlets should be opened immediately for priority drainage.

Images