Intelligent dosing optimization system and method for waterworks based on data driving

Abstract

The invention discloses a waterworks intelligent dosing optimization system and method based on data driving, and belongs to the technical field of water treatment, and the system comprises a historical data intelligent clustering module which obtains and preprocesses historical operation data, employs a minimum distance clustering algorithm to automatically generate a clustering center representing a typical working condition, and sends the clustering center to a cloud server; calculating a corresponding expert recommended dosage, and constructing an expert experience knowledge base; the real-time parameter dynamic matching module is used for acquiring real-time operation parameters, performing window processing, performing Euclidean distance matching on the real-time operation parameters and a clustering center in a knowledge base, and finding out a reference dosage corresponding to the most similar working condition; and the dosage optimization decision module is used for performing feedback correction on the reference dosage according to the comparison relationship between the real-time effluent quality and the standard value, and generating a final dosage instruction under the constraint of the minimum dosage. According to the invention, accurate and dynamic control of the dosage is realized, and the medicament consumption is obviously reduced on the premise of ensuring the stability of the effluent quality.

Description

Technical Field

[0001] This invention belongs to the field of water treatment technology, specifically relating to a data-driven intelligent dosing optimization system and method for waterworks. Background Technology

[0002] Coagulation and sedimentation in water treatment plants is a core step in water purification, and the dosage of coagulant is a key factor determining treatment effectiveness and operating costs. Currently, many water plants still heavily rely on the personal experience of operators for chemical dosing control. This traditional approach has significant drawbacks: First, human decision-making tends to be conservative, often leading to excessive dosing of chemicals to ensure absolute compliance with effluent quality standards, resulting in waste and directly increasing operating costs. Second, parameters such as raw water quality (e.g., turbidity, temperature) and flow rate frequently undergo nonlinear and non-stationary dynamic changes, making it difficult for human experience to respond quickly and accurately to these complex changes, potentially leading to untimely or inaccurate dosing. Finally, while water plants accumulate massive amounts of historical data over long-term operation, this data is often fragmented and has not been effectively utilized to form a structured knowledge base, resulting in a lack of systematic optimization basis and data support for chemical dosing decisions.

[0003] While some technical solutions attempt to analyze historical data using data clustering methods—for example, by multi-dimensionally clustering historical water quality parameters to generate characteristic operating conditions and establishing a mapping relationship between water quality characteristics and chemical dosage—these methods typically employ algorithms that require pre-setting the number of clusters (such as K-means), making them less adaptable to scenarios with variable operating conditions and unstable data. Furthermore, their decision responses are often hierarchical, lacking smoothness and making it difficult to achieve refined continuous control.

[0004] Therefore, there is an urgent need for an intelligent dosing optimization technology that can make full use of historical data, automatically adapt to changes in operating conditions, and achieve continuous and precise control, in order to solve the problems existing in the current technology. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to address the shortcomings of existing technologies by providing a data-driven intelligent chemical dosing optimization system and method for water treatment plants. This system integrates data clustering, dynamic matching, and feedback optimization to achieve both precise control of chemical dosage and improved economic benefits.

[0006] Technical Solution: The data-driven intelligent chemical dosing optimization system for water treatment plants described in this invention includes a historical data intelligent clustering module, which is configured to: acquire historical operating data, wherein the historical operating data includes at least influent operating parameters, effluent water quality parameters, and corresponding historical chemical dosing amounts; preprocess and cluster analyze the historical operating data to generate multiple cluster centers representing typical operating conditions, and calculate the corresponding expert-recommended chemical dosing amount based on the historical chemical dosing data contained in each cluster center, thereby constructing an expert experience knowledge base;

[0007] The real-time parameter dynamic matching module is configured to: acquire real-time operating parameters and perform window processing; match the processed real-time operating parameters with the cluster centers in the expert experience knowledge base to determine the cluster center that best matches the current working condition, and use the expert recommended dosage corresponding to the cluster center as the benchmark dosage;

[0008] The dosing optimization decision module is configured to: receive the baseline dosing amount, and based on the relationship between the real-time effluent water quality parameters and the preset effluent water quality standard value, perform feedback correction on the baseline dosing amount, and simultaneously apply a minimum dosing amount constraint to generate a final dosing amount instruction and output it to the dosing equipment.

[0009] To further improve the above technical solution, the preprocessing of the historical data intelligent clustering module includes using the sliding window method to perform moving average calculation on high-frequency time series data to reduce sampling, and normalizing multi-dimensional parameters to eliminate dimensional differences.

[0010] Furthermore, the clustering analysis employs a minimum distance clustering algorithm, which automatically determines the number of cluster centers through an adaptive distance threshold. This includes randomly selecting initial centers, iteratively selecting the sample furthest from the existing centers as new centers, until the distance from all samples to the nearest center does not exceed the threshold.

[0011] Furthermore, the historical operating data includes influent turbidity, effluent turbidity, flow rate, temperature, and corresponding dosage; the construction of the expert experience knowledge base includes inverse normalization of the cluster centers and calculation of the average dosage of each cluster subset as the expert recommendation value.

[0012] Furthermore, the real-time parameter dynamic matching module determines the similarity by calculating the Euclidean distance between the real-time running parameters and each cluster center, and selects the cluster center with the smallest Euclidean distance as the best matching cluster center.

[0013] Furthermore, the real-time parameter dynamic matching module uses a moving average filter to perform window processing on the real-time operating parameters, with a time interval of 5 minutes.

[0014] Furthermore, the dosage optimization decision module uses the following formula to determine the final dosage. :

[0015]

[0016] in, Based on the baseline dosage, For real-time effluent water quality parameters, Here, k represents the standard value for effluent water quality, and k is a correction factor less than 1. The minimum dosage constraint is defined as follows: The constraint coefficient is greater than 0 and less than 1.

[0017] Furthermore, the correction coefficient k is set to 0.9, and the constraint coefficient... The value is 0.8.

[0018] Furthermore, the historical data intelligent clustering module is also configured to periodically incorporate newly collected operational data into historical operational data and re-execute clustering analysis to dynamically update the expert experience knowledge base.

[0019] The method for optimizing intelligent chemical dosing in water treatment plants using the aforementioned data-driven intelligent dosing optimization system includes the following steps:

[0020] Building a knowledge base: Acquire historical operating data containing influent operating condition parameters, effluent water quality parameters, and corresponding historical dosing amounts; preprocess the historical operating data, including using a sliding window method to perform moving average calculation on high-frequency time series data to reduce sampling, and normalizing multi-dimensional parameters to eliminate dimensional differences; perform cluster analysis using a minimum distance clustering algorithm, automatically determine the number of cluster centers through an adaptive distance threshold, generate multiple cluster centers representing typical operating conditions, and calculate the corresponding expert-recommended dosing amount based on the historical dosing data contained in each cluster center, thereby constructing an expert experience knowledge base;

[0021] Real-time matching: Obtain real-time operating parameters and perform window processing using a moving average filter with a 5-minute time interval; perform Euclidean distance similarity matching between the processed real-time operating parameters and each cluster center in the expert experience knowledge base, select the cluster center with the smallest Euclidean distance as the best matching cluster center, and use the expert recommended dosage corresponding to this cluster center as the baseline dosage;

[0022] Optimization decision: Receive the baseline dosage and, based on the relationship between real-time effluent water quality parameters and preset effluent water quality standard values, use a piecewise function to correct the baseline dosage, while applying a minimum dosage constraint to generate the final dosage instruction;

[0023] Dynamic update: Periodically incorporate newly collected operational data into historical operational data and re-execute the steps of building the knowledge base to dynamically update the expert experience knowledge base.

[0024] Beneficial effects: Compared with the prior art, the advantages of the present invention are as follows:

[0025] This invention transforms the implicit experience of operators into an explicit, queryable expert knowledge base by adaptively clustering historical data, thereby automating and intelligentizing drug dosing decisions.

[0026] This invention employs a minimum distance clustering algorithm, which can automatically determine the operating mode based on the characteristics of the data itself, avoiding the subjectivity of manually setting parameters; combined with Euclidean distance matching of real-time parameters, it can quickly and accurately identify the current operating condition, providing a reliable basis for decision-making.

[0027] By introducing a dynamic correction mechanism based on effluent turbidity feedback and a minimum dosage constraint, a closed-loop control of "small-step fine-tuning" is achieved. This enables the continuous exploration of the optimal solution for chemical consumption while ensuring the safety of effluent water quality, thus significantly reducing chemical costs.

[0028] This invention has the ability to periodically update the knowledge base, continuously learn new operating data, and enable the system model to always maintain its ability to adapt to changes in the water plant's operating status. Attached Figure Description

[0029] Figure 1 It is a data-driven intelligent chemical dosing optimization system for water treatment plants.

[0030] Figure 2 It is a data-driven intelligent chemical dosing optimization method for water treatment plants.

[0031] Figure 3 This is a comparison chart of historical alum addition and matching alum addition.

[0032] Figure 4 This is a comparison chart of historical alum addition and intelligent alum addition. Detailed Implementation

[0033] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0034] Example 1: As Figure 1 The data-driven intelligent dosing optimization system for water treatment plants shown here consists of three main modules: a historical data intelligent clustering module, a real-time parameter dynamic matching module, and a dosing optimization decision module.

[0035] 1. Historical Data Intelligent Clustering Module

[0036] Step 1: Acquire and preprocess historical running data

[0037] Historical dosing data primarily originates from human experience-based decision-making. Its decision-making logic is often based on conservative principles accumulated through long-term operation and maintenance, thus exhibiting strong robustness and reliability in most cases. Although this type of data may tend to be conservative in terms of response speed and optimization scope, failing to fully explore potential system optimization space, it embodies a wealth of experiential knowledge from real-world operating scenarios, including the identification of abnormal operating conditions, safety control of process boundaries, and implicit considerations of complex environmental factors. This experience not only reflects the operator's deep understanding of system characteristics but also the wisdom of making reasonable decisions based on limited information under historical conditions. Therefore, historical dosing data not only possesses significant reference value but can also provide key domain knowledge support when building data-driven models and formulating intelligent dosing strategies, becoming an effective basis for achieving refined and intelligent operational decision-making.

[0038] Retrieve historical operating data (including influent turbidity, effluent turbidity, flow rate, temperature, and corresponding chemical dosage) from the database over a period of time (e.g., one week). This data constitutes a multidimensional dataset containing at least four dimensions of operating parameters, such as influent turbidity. Flow rate Q, temperature W, and effluent water quality parameters of the sedimentation tank (effluent turbidity) ) and the corresponding historical dosage D.

[0039] To improve data quality and subsequent clustering results, preprocessing is necessary. Preprocessing includes two sub-steps:

[0040] Downsampling: For high-frequency, minute-level time-series data, downsampling is used to reduce data granularity, noise, and highlight macroscopic trends. This invention employs a sliding window method, using 5-minute intervals to divide continuous time series into non-overlapping, equal-length windows. Within each window, the arithmetic mean of the data points is calculated to determine the moving average, thereby extracting the central tendency features of the data within that 5-minute time slice. This method not only effectively smooths short-term fluctuations and outlier interference but also transforms the original high-frequency data into more interpretable low-frequency features, providing a well-structured and robust data foundation for subsequent time-series analysis, system performance evaluation, and decision optimization.

[0041] Normalization: refers to historical operating parameters (including influent turbidity). effluent turbidity ,flow ,temperature and dosage The parameters are normalized to map the values ​​of each dimension to the interval [0, 1], as shown in the following formula:

[0042]

[0043] in, Indicates the first The first sample Dimensional original parameter values ​​(such as wait); Indicates the first The historical minimum and maximum values ​​of the dimension parameter; This represents the normalized parameter value, which takes the range of... Within the interval.

[0044] Step 2: Cluster Analysis

[0045] An improved minimum distance algorithm is used to automatically determine the number of cluster centers and optimize the initial center distribution. An initial center is randomly selected. Then, iteratively select the sample farthest from the existing center as the new center, until the distance from all samples to the nearest center does not exceed a threshold. Finally, each sample is assigned to the nearest center, forming a cluster. Detailed algorithm description:

[0046] 1) Threshold setting: Set the distance threshold (like, ).

[0047] 2) Cluster center generation

[0048] Step 1: Randomly select a sample point as the initial center. ;

[0049] Step 2: Calculate the distance from all sample points to all existing cluster centers. The Euclidean distance is used to select the farthest sample as the second center. ;

[0050] Step 3: Iteratively calculate the minimum distance from the sample to the existing center. If it exists If so, then the corresponding sample is set as the new center;

[0051] Termination condition: Repeat step 3 until the distance from all sample points to their nearest cluster center is no greater than the threshold. When all The iteration stops at a certain point, eventually generating a set of cluster centers. ;

[0052] 3) Sample allocation: Each sample is assigned to the nearest cluster center, forming a cluster subset:

[0053]

[0054] The minimum distance clustering method can automatically determine the number of cluster centers based on the data. In practical applications, there is no need to pre-set parameters such as the number and size of cluster centers. The number and size of the cluster centers are determined by the data itself, which can adapt to situations where the data is not robust.

[0055] Step 3: Building an expert experience knowledge base

[0056] The clustering results are processed to form a knowledge base that can be directly queried.

[0057] Inverse normalization: normalizing the cluster centers The coordinates are restored to the original physical parameter range, so that the processed cluster centers directly correspond to the actual operating parameters. The formula is:

[0058]

[0059] in Indicates the first The cluster centers at the in The actual physical values ​​of the parameters (such as actual turbidity, flow rate, etc.).

[0060] Calculate the expert-recommended dosage: for each cluster subset The average dosage was calculated and used as the expert recommendation value for this cluster center.

[0061]

[0062] Recommended dosage for each cluster center Based on historical operational data, preserving expert decision-making logic. Generating a mapping table. This will create a searchable expert knowledge base.

[0063] 2. Real-time parameter dynamic matching module

[0064] The current influent parameters are matched with the historical operating condition model library to generate an initial dosing recommendation.

[0065] Step 1: Real-time data window processing

[0066] The real-time operating parameters (influent turbidity, effluent turbidity, flow rate, and temperature) for the five minutes preceding the current moment are obtained. Using the same moving average filtering method as historical data preprocessing, the average value of each parameter within this time window is calculated as a representative of the current stable operating conditions. This step has a clearly defined scope and does not involve any dimension related to the dosage. The dosage is not included in the treatment scope of this step; only the influent parameters are processed.

[0067] For the current time window ( The influent parameters within the window are processed using a moving average filter to eliminate instantaneous fluctuations and extract data features. The arithmetic mean of each parameter within the window is calculated using the following formula:

[0068]

[0069] in, This indicates the number of data points within a 5-minute window (1 sampling point per minute). The 5-minute window length balances the requirements for response speed and data stability. Representing historical moments Measured values ​​of influent turbidity, effluent turbidity, flow rate, and temperature; This represents the mean of the filtered parameters, which serves as the input for the dosing decision.

[0070] Step 2: Similarity Matching

[0071] Filtered new data With historical cluster center Perform a match:

[0072] 1) Euclidean distance calculation:

[0073]

[0074] 2) Optimal matching selection: Select the cluster center with the smallest distance. The corresponding expert dosage As a benchmark value:

[0075]

[0076] in, Representing cluster centers The corresponding historical average dosage reflects expert decision-making experience under typical operating conditions. This represents the optimal historical operating condition pattern obtained through Euclidean distance matching.

[0077] 3. Dosage Optimization Decision Module

[0078] Step 1: Obtain the baseline dosage

[0079] Extract the best matching cluster centers The corresponding expert recommended dosage And use it as the benchmark dosage. The output is then passed to the next module.

[0080] Figure 3This is a comparison chart of historical alum addition and matched alum addition. Orange sample points represent historical alum addition, and green sample points represent matched alum addition. The overall trend of matched alum addition (green points) and historical alum addition (orange points) is highly consistent, indicating that the expert knowledge base generated by the system through historical data clustering can accurately capture the patterns of manual operation, verifying the effectiveness of the clustering algorithm. At peak alum addition (e.g., alum addition of 275) and trough values ​​(e.g., alum addition of 160), the deviation between matched alum addition and historical values ​​is minimal, indicating that the system has a strong ability to match extreme working conditions and has not lost key features due to data compression or clustering.

[0081] Step Two: Dynamic Optimization and Adjustment

[0082] Based on the current measured values ​​of effluent turbidity The ratio to the standard value is used to correct the baseline dosage, while a lower limit is set to prevent excessive reduction and ensure system robustness. The formula is as follows:

[0083]

[0084] in: It is the baseline dosage output by the real-time parameter dynamic matching module; These are real-time effluent water quality parameters (such as effluent turbidity). It is the preset standard value for effluent water quality; This indicates a strict constraint on reducing the dosage, ensuring that the dosage is not lower than 80% of the baseline value. The final generated dosage instruction will be sent to the dosing equipment via PLC for execution.

[0085] When the real-time effluent turbidity Exceeding the standard When the system is in use, the dosage will be increased proportionally; when Less than or equal to When this happens, the system will attempt to reduce the dosage proportionally, but the reduced value cannot be lower than 80% of the baseline dosage. This rigid constraint prevents the risk of water quality fluctuations due to over-optimization, ensuring the robustness of the system. This "small-step fine-tuning" strategy drives the system state to gradually approach the target stable value.

[0086] Figure 4 This is a comparison chart of historical alum dosage and intelligent alum dosage. Orange sample points represent historical alum dosage, and blue sample points represent intelligent alum dosage. Because the effluent turbidity of the test water plant is low, the trends of intelligent and historical alum dosages are similar. The intelligent alum dosage curve shows that almost all sample points are lower than the historical alum dosage, resulting in a 20% reduction in chemical consumption. This indicates that the optimization decision module of this invention, through dynamic matching, feedback correction, and lower limit constraint mechanisms, achieves a significant reduction in chemical consumption while ensuring water quality.

[0087] Example 2: As Figure 2 As shown, the intelligent dosing optimization method for waterworks based on the data-driven intelligent dosing optimization system provided in Example 1 includes the following steps:

[0088] Building a knowledge base: Acquire historical operating data containing influent operating condition parameters, effluent water quality parameters, and corresponding historical dosing amounts; preprocess the historical operating data, including using a sliding window method to perform moving average calculation on high-frequency time series data to reduce sampling, and normalizing multi-dimensional parameters to eliminate dimensional differences; perform cluster analysis using a minimum distance clustering algorithm, automatically determine the number of cluster centers through an adaptive distance threshold, generate multiple cluster centers representing typical operating conditions, and calculate the corresponding expert-recommended dosing amount based on the historical dosing data contained in each cluster center, thereby constructing an expert experience knowledge base;

[0089] Real-time matching: Obtain real-time operating parameters and perform window processing using a moving average filter with a 5-minute time interval; perform Euclidean distance similarity matching between the processed real-time operating parameters and each cluster center in the expert experience knowledge base, select the cluster center with the smallest Euclidean distance as the best matching cluster center, and use the expert recommended dosage corresponding to this cluster center as the baseline dosage;

[0090] Optimization decision: Receive the baseline dosage and, based on the relationship between real-time effluent water quality parameters and preset effluent water quality standard values, use a piecewise function to correct the baseline dosage, while applying a minimum dosage constraint to generate the final dosage instruction;

[0091] Dynamic update: Periodically incorporate newly collected operational data into historical operational data and re-execute the steps of building the knowledge base to dynamically update the expert experience knowledge base.

[0092] Example 3: The data-driven intelligent dosing optimization system for waterworks provided in Example 1 is applied to a waterworks and deployed in a hardware environment that includes sedimentation tanks, data instruments, PLCs (programmable logic controllers), dosing equipment, databases, and an industrial control platform.

[0093] Before the raw water flows into the sedimentation tank, data instruments collect various parameters of the raw water, including influent turbidity. The system collects data on flow rate (Q) and temperature (W) and sends this data to the PLC. The PLC communicates with the data instruments and dosing equipment, and stores the data in a database via the OPC communication module. The industrial control platform integrates the core software modules of this invention. The dosing equipment communicates bidirectionally with the PLC, sending real-time dosing values ​​and performing dosing according to the PLC's settings. Data instruments are also installed at the sedimentation tank effluent outlet to collect the turbidity of the treated water. The PLC stores real-time data in a database through the OPC system for later calculation and data retrieval.

[0094] After obtaining historical data from the water plant (usually one week), calculate the threshold if it is less than the set threshold. Expert-recommended dosage values ​​for cluster centers and corresponding cluster subsets:

[0095]

[0096] After the matching is completed, the matched data will be stored in the corresponding file.

[0097] In actual operation, the current time window is obtained from the database. The influent parameters (influent turbidity, effluent turbidity, flow rate, and temperature) are measured and then subjected to moving average filtering to eliminate instantaneous fluctuation interference and extract data features.

[0098] Then the filtered new data With historical cluster center Perform a match. Using the formula:

[0099]

[0100] Select the cluster center with the smallest distance. The corresponding expert dosage As a benchmark value, combined with the current measured value of effluent turbidity. The ratio to the standard value is used to correct the baseline dosage, while a lower limit is set to prevent excessive reduction and ensure system robustness. The formula is as follows:

[0101]

[0102] The actual dosage determined The signal is sent to the on-site PLC, which converts it into a corresponding analog signal to control the operation of the dosing equipment.

[0103] The industrial control platform includes a human-machine interface, a historical data intelligent clustering module, a real-time parameter dynamic matching module, a dosing optimization decision module, and an OPC communication module. The function of the human-machine interface is to display the calculation results of each module and various data of the sedimentation tank operation in real time.

[0104] The function of the historical data intelligent clustering module is to re-match historical clustering data with new data every two days (which can be manually set) and update the cluster centers. This ensures that the model can keep up with changes in operating conditions in real time.

[0105] The real-time parameter matching module can obtain a matching value based on the current real-time data within 2-3 seconds and send the result to the dosing optimization decision module to obtain the real-time dosing amount. In addition, it has a good response strategy for abnormal fluctuations or sudden increases in influent turbidity, and further ensures that the turbidity of the sedimentation tank effluent remains stable and meets the set requirements.

[0106] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A data-driven intelligent chemical dosing optimization system for water treatment plants, characterized in that, include: The historical data intelligent clustering module is configured to: acquire historical operating data, which includes at least influent operating parameters, effluent water quality parameters and corresponding historical dosage; preprocess and cluster analyze the historical operating data to generate multiple cluster centers representing typical operating conditions; and calculate the corresponding expert recommended dosage based on the historical dosage data contained in each cluster center, thereby constructing an expert experience knowledge base. The real-time parameter dynamic matching module is configured to: acquire real-time operating parameters and perform window processing; match the processed real-time operating parameters with the cluster centers in the expert experience knowledge base to determine the cluster center that best matches the current working condition, and use the expert recommended dosage corresponding to the cluster center as the benchmark dosage; The dosing optimization decision module is configured to: receive the baseline dosing amount, and based on the relationship between the real-time effluent water quality parameters and the preset effluent water quality standard value, perform feedback correction on the baseline dosing amount, and simultaneously apply a minimum dosing amount constraint to generate a final dosing amount instruction and output it to the dosing equipment.

2. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The preprocessing of the historical data intelligent clustering module includes using the sliding window method to perform moving average calculation on high-frequency time series data to reduce sampling, and normalizing multi-dimensional parameters to eliminate dimensional differences.

3. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The clustering analysis employs the minimum distance clustering algorithm, which automatically determines the number of cluster centers through an adaptive distance threshold. This includes randomly selecting initial centers, iteratively selecting the sample furthest from the existing centers as new centers, until the distance from all samples to the nearest center does not exceed the threshold.

4. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The historical operating data includes influent turbidity, effluent turbidity, flow rate, temperature, and corresponding dosage; the construction of the expert experience knowledge base includes inverse normalization of the cluster centers and calculation of the average dosage of each cluster subset as the expert recommendation value.

5. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The real-time parameter dynamic matching module determines the similarity by calculating the Euclidean distance between the real-time running parameters and each cluster center, and selects the cluster center with the smallest Euclidean distance as the best matching cluster center.

6. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 5, characterized in that, The real-time parameter dynamic matching module uses a moving average filter to perform window processing on the real-time operating parameters, with a time interval of 5 minutes.

7. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The dosage optimization decision module determines the final dosage using the following formula. : ； in, Based on the baseline dosage, For real-time effluent water quality parameters, Here, k represents the standard value for effluent water quality, and k is a correction factor less than 1. The minimum dosage constraint is defined as follows: The constraint coefficient is greater than 0 and less than 1.

8. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 7, characterized in that, The correction coefficient k is 0.9, and the constraint coefficient The value is 0.

8.

9. The data-driven intelligent chemical dosing optimization system for water treatment plants according to claim 1, characterized in that, The historical data intelligent clustering module is also configured to periodically incorporate newly collected operational data into historical operational data and re-execute clustering analysis to dynamically update the expert experience knowledge base.

10. The intelligent dosing optimization method for waterworks based on a data-driven intelligent dosing optimization system for waterworks according to claim 1, characterized in that, Includes the following steps: Building a knowledge base: Acquire historical operating data containing influent operating condition parameters, effluent water quality parameters, and corresponding historical dosing amounts; preprocess the historical operating data, including using a sliding window method to perform moving average calculation on high-frequency time series data to reduce sampling, and normalizing multi-dimensional parameters to eliminate dimensional differences; perform cluster analysis using a minimum distance clustering algorithm, automatically determine the number of cluster centers through an adaptive distance threshold, generate multiple cluster centers representing typical operating conditions, and calculate the corresponding expert-recommended dosing amount based on the historical dosing data contained in each cluster center, thereby constructing an expert experience knowledge base; Real-time matching: Obtain real-time operating parameters and perform window processing using a moving average filter with a 5-minute time interval; The processed real-time operating parameters are matched with each cluster center in the expert experience knowledge base by Euclidean distance similarity. The cluster center with the smallest Euclidean distance is selected as the best matching cluster center, and the expert recommended dosage corresponding to the cluster center is used as the benchmark dosage. Optimization decision: Receive the baseline dosage and, based on the relationship between real-time effluent water quality parameters and preset effluent water quality standard values, use a piecewise function to correct the baseline dosage, while applying a minimum dosage constraint to generate the final dosage instruction; Dynamic update: Periodically incorporate newly collected operational data into historical operational data and re-execute the steps of building the knowledge base to dynamically update the expert experience knowledge base.

Images